# Huzaifa Saif-ur-Rehman — Full Content > Huzaifa Saif-ur-Rehman is a staff software engineer specializing in Laravel, PHP, JavaScript, DevOps, SaaS architecture, and API integrations. ## About Huzaifa Saif-ur-Rehman Huzaifa Saif-ur-Rehman is a staff software engineer from Pakistan with 15 years of professional experience in PHP, Laravel, JavaScript, and Node.js. He has collaborated with international clients from the USA, India, Dubai, France, Australia, and Canada across industries including Traveling, Education, Logistics, and Import/Export. Certifications and Recognition: - JSNAD: OpenJS Certified Node.js Application Developer — OpenJS Foundation - LFW211: Node.js Application Development — The Linux Foundation (LiFT Scholarship recipient) - Laravel Live Pakistan 2024 Speaker — presented on Laravel Telescope Guzzle Watcher and Laravel Pulse Guzzle Recorder - Full Stack Laracasts Laravel Architect Open Source: - Laravel Telescope Guzzle Watcher: Top 9 Laravel package of 2023, 270K+ downloads - Laravel Pulse Guzzle Recorder: Monitors Guzzle HTTP calls in Laravel Pulse dashboard Contact: info@muhammadhuzaifa.pro LinkedIn: https://www.linkedin.com/in/mhuzaifacs/ GitHub: https://github.com/huzaifaarain ## Services ### SaaS Product Development & Engineering URL: https://muhammadhuzaifa.pro/saas-product-development Senior Laravel SaaS product engineering. Marketplaces and multi-party payments with Stripe Connect (provider onboarding, split payments, platform fees, payouts, webhook-driven reliability). Multi-tenant SaaS platforms (team accounts, roles, billing, feature flags). Rescue and modernization of stalled or fragile SaaS, including legacy-to-modern Laravel rebuilds with hardened payment flows and test coverage. Built with Laravel, Livewire, Vue, Inertia, Tailwind, PostgreSQL, and Stripe. ### DevOps Services URL: https://muhammadhuzaifa.pro/devops-services Linux server setup (Ubuntu, Debian, CentOS), Docker, Nginx, Apache, CI/CD (GitHub Actions, GitLab), AWS, Google Cloud, DigitalOcean. ### API & Data Integrations URL: https://muhammadhuzaifa.pro/integrations HubSpot, Salesforce, SAP Business ByDesign, SAP Business One, Oracle NetSuite, Stripe, PayPal, Braintree, GraphQL. ### AI Engineering & LLM Development URL: https://muhammadhuzaifa.pro/ai-engineering LLM feature integration (summarize, classify, extract, chat, copilots) into Laravel and Node products with streaming and cost controls. Retrieval-augmented generation (RAG) grounded in your own data with embeddings and vector search. Agentic workflows with multi-step tool use and explicit guardrails. Prompt engineering, evaluation harnesses, and observability so AI features fail loudly and safely. Frontier hosted models (Claude, OpenAI) and open-source or self-hosted models, including private self-hosted models in Docker for sensitive data. ### Get a Quote URL: https://muhammadhuzaifa.pro/quote Request a quote for Laravel integrations, payment flows, SaaS rescue, or production engineering. Include the project context, budget band, timeline, and platforms involved. ## Portfolio ### Marketplace SaaS Platform With Stripe Connect URL: https://muhammadhuzaifa.pro/portfolio/marketplace-saas-platform-with-stripe-connect Category: Full-Stack Tech: Laravel, PostgreSQL, Livewire, Stripe Connect, HubSpot, S3, Laravel Pennant This is a multi-sided marketplace SaaS that connects service providers with the clients who hire them, and handles the money between them. I led development on the platform: a Laravel application where providers onboard, invoice their clients, and get paid out, while the platform collects its fee on every transaction. The hard part of a product like this is never the marketing page. It is moving other people's money correctly, every time, between three parties. Multi-party payments with Stripe Connect The core is Stripe Connect done the real way: not a single charge, but separate charges and transfers. A client pays an invoice, the platform collects its application fee, and the remainder is transferred to the provider's connected Stripe account. Providers see a live payout estimate before they send an invoice, with the platform fee and the per-method processing cost (card versus ACH) broken out, so there is no surprise about what they will actually receive. Provider onboarding, connected accounts, fees, and payouts are all handled in code. Payments that do not fail silently Payment state is driven by Stripe webhooks, not by a browser redirect. Events such as invoice paid and invoice voided are received as webhook events and processed through queued jobs, so the platform's record of who paid stays correct even if a user closes the tab, the network drops, or the same event arrives twice. Getting the money state always right is the part that separates a marketplace that works from one that quietly loses or mis-attributes payments. Built as a product, not a script The platform is a modular Laravel monolith with clear boundaries between payments, CRM synchronization, jobs, and user management. It is multi-tenant through team-based accounts, uses typed data objects to keep payment payloads explicit and safe, runs work on a real queue, and ships with automated test coverage and production observability. Feature flags let new capabilities roll out safely, and HubSpot CRM workflows keep the business side in sync. The stack Laravel, PostgreSQL, Livewire, Tailwind, Alpine, Laravel Cashier and Stripe Connect, HubSpot, S3, feature flags, and production monitoring, on PHP 8.4 and Laravel 12. My role Led development across the product: the payments architecture, the marketplace flows, the modular structure, and the reliability of the money path. ### Best Price Market Ecommerce Marketplace URL: https://muhammadhuzaifa.pro/portfolio/best-price-market-ecommerce-marketplace Client: Best Price Market Category: Full-Stack Tech: Laravel, Livewire, Ecommerce, Inertia.js, Vue 3, SEO, Payments Best Price Market is a full ecommerce marketplace built on Laravel, where the catalog, the checkout, and the money all have to be right. I built the platform end to end: a typed, modern Laravel application with an Inertia and Vue front end, a multi-currency catalog, and a payment pipeline engineered so an order is only ever marked paid when the payment is genuinely confirmed. The product A complete storefront and admin platform: products with variants and custom fields, categories and brands, inventory, cart and checkout, orders, shipping carriers, discounts, and multi-currency pricing. Customers browse and buy, while staff manage the catalog, orders, and fulfillment through role-based admin dashboards. Persistent carts, abuse protection, content, and merchandising are part of the product, not afterthoughts. Payments that are only "paid" when they are really paid Checkout runs through the Zen payment gateway, and order status is driven by Zen's IPN, the gateway's webhook, not by the customer's browser. Every IPN is stored as its own record, then processed by a queued, monitored job that re-validates the notification directly with Zen before trusting it, confirms the order by its merchant transaction id, marks it paid, and notifies the customer. Each event's outcome is tracked as succeeded or failed, failures are recorded with a reason, and the job retries with backoff. The result is a payment record that stays correct even if a browser closes, a network drops, or a notification arrives twice, with a full audit trail of every payment event. Built as a typed, maintainable product The codebase uses typed data objects that generate TypeScript types for the front end, so the contract between Laravel and Vue is explicit and safe. It is organized into service, job, observer, and data layers, with activity logging for audit trails, role-based permissions, queue monitoring, and media handling. SEO is engineered in: sitemaps, feeds, schema.org structured data, breadcrumbs, and Open Graph metadata. The stack Laravel, Inertia, Vue, Livewire, Spatie Data with TypeScript generation, money handling for multi-currency, the Zen payment gateway, and queue monitoring. My role Built the platform: the catalog and checkout, the reliable payments pipeline, the typed architecture, and the admin and SEO systems. ### GZPCHS Housing Society Management System URL: https://muhammadhuzaifa.pro/portfolio/gzpchs-housing-society-management-system Client: Gulshan-e-Zeal Pak Cooperative Housing Society Category: Full-Stack Tech: Laravel, Inertia.js, Vue 3, Tailwind CSS, MySQL, PDF, Excel Built a comprehensive property management system for Gulshan-e-Zeal Pak Cooperative Housing Society covering plots, blocks, shops, bookings, vouchers, allotment letters, announcements, reporting, and user management. The application includes automated monthly revenue voucher generation, ledger reporting, executive dashboard metrics, data entry analytics, audit trails, document management, role-based access, backups, PDF generation, and Excel exports. The current version uses Laravel, MySQL, Inertia.js, Vue 3, Tailwind CSS, ApexCharts, Vite, Spatie settings, activity logs, media library, and permissions. ### GEL Virtual Mailbox & Logistics Platform URL: https://muhammadhuzaifa.pro/portfolio/gel-virtual-mailbox-logistics-platform Client: Global Express Limited Category: Full-Stack Tech: Laravel, MySQL, PHP, Stripe, Vue.js, Logistics, DataTables Global Express Limited is a Caribbean freight-forwarding and virtual mailbox business. Customers across Barbados, the Cayman Islands, Guyana, Saint Kitts and Nevis, and other islands ship to a US address; GEL receives, processes, and forwards their packages, and bills them for it. I have been the sole engineer on their Laravel platform since 2016, building and maintaining the system that runs that operation end to end. The domain This is not a CRUD app. A package moves through a real lifecycle, from intake, barcode, and media attachments to freight-rate calculation, invoicing, payment, and forwarding, with payment and claim statuses at each step. Shipping eligibility depends on the customer's country, since some destinations are allowed and others are blocked by business and carrier rules. Staff, agents, and customers each see a different slice of the system through role-based permissions. Invoices can be issued per package or merged in bulk into a single payment. What I built A Laravel platform covering that whole lifecycle: package intake and barcode generation, a media library for package photos and documents, country-aware freight-rate calculation, per-package and bulk-merged invoicing, Stripe checkout for payment, customer and staff notifications, internal ticketing, Excel exports, and the admin-heavy DataTables interfaces the team runs the business on. The code is organized into service, repository, query, and action layers with per-domain events and observers, so the business rules live in one place and the package lifecycle stays consistent wherever it is triggered. Telescope provides production visibility. Engineering for the long run The part most worth noting is longevity: this system has run a live business for years, through changing requirements. I am now modernizing it to Laravel 13 with an Inertia and Vue 3 frontend, including a hardened, webhook-driven billing path and test coverage around the money flow. Knowing where a long-lived payment integration can fail silently, and improving it without disrupting a running business, is the same work I do for integration clients. My role Sole engineer: architecture, build, and ongoing maintenance since 2016, and now the full modernization. ### HubSpot & Oracle NetSuite Bidirectional Sync URL: https://muhammadhuzaifa.pro/portfolio/hubspot-oracle-netsuite-bidirectional-sync Category: Integration Tech: Laravel, HubSpot, Oracle NetSuite, SOAP API, REST API, Queues, Data Sync Built a Laravel integration synchronizing HubSpot contacts and companies with Oracle NetSuite customers, while pushing NetSuite sales orders back into HubSpot deals. Handled bidirectional sync concerns including ID mapping, field transformations, retry handling, partial failures, webhook loop prevention, and operational visibility for reprocessing failed records. The implementation used Laravel, HubSpot APIs, Oracle NetSuite REST and SuiteTalk SOAP APIs, queues, scheduled jobs, Guzzle observability, DataTables, Bootstrap, and MySQL. ### HubSpot & SAP Business One Integration URL: https://muhammadhuzaifa.pro/portfolio/hubspot-sap-business-one-integration Category: Integration Tech: Laravel, HubSpot, Data Sync, SAP Business One, ERP Integration, Modules Built a Laravel integration moving customer, contact, and order data from HubSpot into SAP Business One with clean module boundaries around source, target, and orchestration logic. Mapped nested contact structures, preserved ordering dependencies between company and contact records, and handled partial failures without blocking the wider synchronization flow. The stack included Laravel modules, HubSpot APIs, SAP Business One APIs, Telescope, outbound HTTP request monitoring, DataTables, Bootstrap, settings management, and MySQL. ### Bidirectional HubSpot/NetSuite Integration URL: https://muhammadhuzaifa.pro/portfolio/bidirectional-hubspotnetsuite-integration Client: Australian shelter industry client Category: Integration Project: https://www.upwork.com/freelancers/mhuzaifacs Tech: Laravel, HubSpot, MySQL, PHP, SOAP API, Queues, NetSuite Built a bidirectional HubSpot and NetSuite integration for Australia's shelter industry using Laravel, PHP, MySQL, queues, scheduler commands, and a service-layer architecture. The system synchronizes contacts and companies in both directions, sends NetSuite sales orders into HubSpot, and keeps marketing data current through automated queue jobs and scheduled Artisan commands. NetSuite communication used SOAP/WSDL and API-user authentication, while HubSpot communication used a private app. The admin interface used Alpine.js, Livewire, and Tailwind CSS for monitoring and control. ### Laravel HubSpot & Childcare Management Integration URL: https://muhammadhuzaifa.pro/portfolio/laravel-hubspot-childcare-management-integration Client: Australian childcare provider Category: Integration Project: https://www.upwork.com/freelancers/mhuzaifacs Tech: Laravel, Livewire, HubSpot, Tailwind CSS, Queues, API Integration, Scheduler Developed and deployed a Laravel integration between HubSpot and a custom childcare management system for an Australian client. The solution synchronizes campuses, children, parents, family trees, enrollments, and bookings into HubSpot so the client can run more accurate marketing automation. The implementation used Laravel queues, scheduled Artisan commands, service classes, HubSpot private app authentication, basic authentication for the childcare system, and a Livewire/Tailwind interface. ### Integrated Custom Payment Gateway URL: https://muhammadhuzaifa.pro/portfolio/integrated-custom-payment-gateway Client: E-commerce client Category: Integration Project: https://www.upwork.com/freelancers/mhuzaifacs Tech: Laravel, MySQL, PHP, API Integration, Payment Gateway, Laravel Nova Built a ZEN.COM payment gateway integration for a Laravel e-commerce checkout running on Laravel Nova. Created the payment service responsible for communicating with the ZEN.COM Payment API and wired it into the checkout flow so customer payments could be handled inside the existing Laravel application. ### Laravel Passport OAuth 2.0 Server & Client URL: https://muhammadhuzaifa.pro/portfolio/laravel-passport-oauth-20-server-client Category: Full-Stack Project: https://www.upwork.com/freelancers/mhuzaifacs Tech: Laravel, MySQL, OAuth 2.0, Laravel Passport, Docker, API Created a Docker-based Laravel Passport example showing both an OAuth 2.0 server and a consuming client application. The setup includes PHP-FPM containers for the server and client, MySQL, Nginx, and phpMyAdmin, giving developers a complete reference implementation for issuing and consuming OAuth tokens in Laravel. ### Tombstone Orders Laravel Application API URL: https://muhammadhuzaifa.pro/portfolio/tombstone-orders-laravel-application-api Category: Full-Stack Project: https://www.upwork.com/freelancers/mhuzaifacs Tech: Laravel, MySQL, PHP, Docker, API, Testing Built a Laravel API for managing tombstone orders and related catalog data. The implementation included migrations, models, seeders, resource controllers, and unit coverage for orders, tombstones, fonts, text colors, and icons. ### HubSpot, Ostendo & Mindbody Integration URL: https://muhammadhuzaifa.pro/portfolio/hubspot-ostendo-mindbody-integration Category: Integration Tech: Laravel, HubSpot, Queues, Ostendo, Mindbody, AWS, CSV Built a Laravel integration across HubSpot, Ostendo, and Mindbody, combining API synchronization with CSV import and export workflows for bulk data movement. Created service-layer boundaries for HubSpot and Ostendo, controller flows for operational actions, and queue-backed processing for long-running or failure-prone integration work. The implementation used Laravel, HubSpot APIs, Ostendo APIs, Mindbody data flows, AWS SDK, League CSV, Sanctum, Bootstrap, log viewing, and database queues. ### Funraisin to HubSpot Fundraising Sync URL: https://muhammadhuzaifa.pro/portfolio/funraisin-to-hubspot-fundraising-sync Category: Integration Tech: Laravel, HubSpot, DataTables, Funraisin, CRM Integration, Monitoring Built a Laravel synchronization layer that moved fundraising platform data into HubSpot contacts and deals for CRM automation and reporting. Implemented scheduled sync workflows, API client services, operational dashboards, log visibility, authentication, and HTTP request monitoring for production support. The stack included Laravel, HubSpot APIs, Funraisin data flows, Telescope, outbound request monitoring, Sanctum, DataTables, Bootstrap, jQuery, Sass, and MySQL. ### PowerPay Payment Processing Integration URL: https://muhammadhuzaifa.pro/portfolio/powerpay-payment-processing-integration Category: Integration Tech: Laravel, HubSpot, Inertia.js, Payments, PowerPay, Rollbar Built a Laravel payment-processing integration using a modern Inertia.js and Jetstream frontend with production error tracking and API-driven workflows. Integrated PowerPay flows with HubSpot-related business processes, JavaScript route generation, Sanctum authentication, and operational monitoring for failed payment or API interactions. The stack included Laravel, Inertia.js, Jetstream, HubSpot APIs, PowerPay, Rollbar, Doctrine DBAL, Ziggy, Sanctum, PHP, and MySQL. ### Paycom & Zuper Field Operations Integration URL: https://muhammadhuzaifa.pro/portfolio/paycom-zuper-field-operations-integration Category: Integration Tech: Laravel, Livewire, Tailwind CSS, Paycom, Zuper, Flux, Spatie Data Built a Laravel integration between Paycom and Zuper for synchronizing HR, payroll, and field service operations data through modular service boundaries. Used modern Laravel application structure with separate modules for each platform, typed data objects, Fortify authentication, Livewire Flux interfaces, Tailwind CSS, and Vite. The implementation focused on maintainable platform adapters, clear data transfer objects, authenticated admin workflows, and reliable synchronization between operational systems. ### Laravel Telescope Guzzle Watcher URL: https://muhammadhuzaifa.pro/portfolio/laravel-telescope-guzzle-watcher Client: Open source Category: Open Source Project: https://github.com/huzaifaarain/telescope-guzzle-watcher Tech: Laravel, API, Telescope, Guzzle, Open Source, Observability Open-source Laravel package for observing outbound HTTP requests made through guzzlehttp/guzzle inside Laravel Telescope. The package uses Guzzle's on_stats request option to capture request timing and details, giving Laravel teams better visibility into API integrations. The package has crossed 270K+ downloads according to the site settings. ### Laravel Pulse Guzzle Recorder URL: https://muhammadhuzaifa.pro/portfolio/laravel-pulse-guzzle-recorder Client: Open source Category: Open Source Project: https://github.com/huzaifaarain/laravel-pulse-guzzle-recorder Tech: Laravel, API, Guzzle, Open Source, Observability, Pulse Open-source Laravel package for recording outbound Guzzle HTTP calls in Laravel Pulse. The recorder helps teams inspect slow or high-volume API calls from the Pulse dashboard, making API-heavy Laravel applications easier to monitor in production. ## Blog Posts ### Future of Software Engineering in the Age of AI Agents at Cyber Nexus 2.0 - CSS - MUET URL: https://muhammadhuzaifa.pro/the-future-of-software-engineering-at-cyber-nexus-css-muet Published: 2026-11-07 ![Software engineering is more than coding](https://muhammadhuzaifa.pro/storage/photos/1/muet-hackathon-linkedin-1.jpg) The future of software engineering is not about typing code faster. It is about directing AI agents, checking their work and owning the result. I shared this view in a 60-minute talk at Cyber Nexus 2.0, the hackathon run by the Cyber Security Society at Mehran University of Engineering and Technology, MUET. The talk moved from vibe coding to agentic engineering. This post covers the main points of the talk, with the stories, the safety rules and a plan for students. ## Why software engineering is more than coding Coding is one box on a big map. The map also holds requirements, design, testing, quality checks, project management, release management, maintenance, security and ethics. A working engineer touches all of them. The classic software development life cycle has eight stages. They are feasibility, planning, requirements, design, implementation, testing, deployment and maintenance. This process grew from one belief. Writing the code is the slow and costly part. That belief is now weak, because agents write code fast. But people still decide what to build, review the result, test it, release it and keep it running. Those steps still move at human speed. So faster coding alone does not finish a project sooner. An Anthropic playbook made the same point, that build gets fast and the human-speed steps around it do not. Those steps become the limit. The lesson is to learn the whole map, not only the code. ## Vibe coding, AI engineering and agentic engineering Vibe coding is a way of building where you forget that the code even exists. Andrej Karpathy named it in February 2025, and Collins Dictionary made it Word of the Year for 2025. Tools like Lovable, Bolt.new and v0 by Vercel let you describe an app and get a working one. That is why it is fun. The trade-off is simple. If you stop reading the code, nobody checks it. Karpathy said he accepts every change and no longer reads the diffs, which are the lists of changes. This suits throwaway projects and demos. It is a poor choice for anything people will depend on. The talk then showed three ways of working. In AI-assisted development, you write the code and the AI suggests, completes and explains. In AI engineering, you build applications on foundation models, which are large pretrained models, using prompts, RAG, tools, agents and evaluation. In agentic engineering, agents write the code and you orchestrate them and act as oversight. A year after his first post, Karpathy said professionals now work this way, with more oversight. ## The building blocks of AI engineering **The API.** A model is reached over plain HTTPS. You send a message with your key and a version header, and you get a response. Official kits exist for Python, TypeScript, Go, Java and more. **Prompts.** A weak prompt says only, summarize this report. A strong one says you are an analyst, write five bullets for a non-technical reader, quote figures exactly and say so when unsure. Be specific about the role, the audience, the format and the limits. **Evals.** An eval tests a prompt, the way a unit test checks code. Code-based grading is fast, reliable and scales well, but it lacks nuance. Model-based grading lets a model score the output, for example from one to ten. **Tools and MCP.** With tool use, the model asks to call a tool, your code runs it and returns the result, and the model answers. The Model Context Protocol, or MCP, is an open standard that Anthropic published in November 2024. You build a connection once, and many agents can use it. **The harness.** This is the runtime around the model. It holds the loop, memory, skills, subagents, tools, a sandbox and an interface. The same model in a different harness is a different product. ## Workflow or agent, and which steps do I control? A workflow is a system where models and tools follow paths written in code. An agent is a system where the model directs its own process and tool use. Anthropic's guide on building effective agents names five workflow patterns. They are prompt chaining, routing, parallelization, orchestrator-workers and evaluator-optimizer. The real question is how much freedom to give. I keep the steps where a mistake costs the most. In plan mode, the agent first researches in read-only mode and proposes a plan. I approve it, and only then does the agent run. The talk also gave an order for adopting these habits. First, write down your intent, set up feedback loops and use plan mode. Next, add skills, subagents and evals. Then wire in pull request review and automatic build and test checks. Files such as intent, spec and plan move between stages in version control, with review gates. Human judgment stays central. ## What goes into an agent's context The model sees only what is in its context window, the limited space for everything it reads in one run. It holds the system prompt, tools, memory, chat history and retrieved documents. Deciding what goes in is the job. RAG cuts documents into chunks, turns them into searchable numbers, retrieves the best chunks and generates an answer. Agentic search lets the agent think, search, read and repeat. The creator of Claude Code said early versions used RAG with a local vector database, but agentic search generally worked better. My first story was a lab plan. I asked how many students fit if an empty room became a lab. No Autodesk MCP was available, so I used an open source one called AIBlueprint, with FreeCAD to preview the drawing. I gave it my hand sketch with exact measurements and made changes by chat. In version five, the room was about 16 feet by 15.5 feet, with two tables, desks on both walls, a clear instructor zone and room for the door to swing. Twenty-one students fit, nine at wall desks and twelve at tables. My second story was my tax return. The agent first read Pakistan's tax laws and rules into memory. Then I gave it my documents, and it filled the FBR tax portal through browser MCP while I watched. It did better than my lawyer the year before and taught me a few things. My finances were simple, and I reviewed everything. The lesson is to make the agent understand first. ## One rulebook for every AI tool I keep one global rulebook in a Git repository. One small tool renders it into Claude Code, Codex and OpenCode. Another syncs 29 skills to six agent tools. A skill is a folder with an instruction file, and many tools support the format. One rule says what to do when an instruction looks non-standard or suboptimal. The agent must stop, name the problem, give a better option with the reason, and let me decide. When an agent makes the same mistake twice, I fix the rulebook, not the prompt. I edit once, and every agent reads it. This saves time, because I do not repeat the same advice in every chat with every tool. Agents can also write their own memory, but the rules that matter stay in my hands. ## Basic safety for agents Agents read the outside world, so everything they read is untrusted input. In May 2025, Invariant Labs described a flaw in the official GitHub MCP server. A malicious issue in a public repository carried hidden instructions. This is called prompt injection. A user asked the agent to check the open issues. The agent then read private repository data and opened a public pull request that contained it. Secure connections are not enough on their own. I follow three habits. First, add guardrails at each layer the talk describes: access, control, gateway, runtime and sandbox. Second, use a sandbox, a closed space where a mistake cannot reach your machine. A container shares the host's kernel, the core of the operating system. A microVM, a tiny virtual machine, has its own kernel, so the agent cannot touch host files or the outside network. Third, treat Git as the undo button. Give each parallel agent its own worktree, a separate checkout, so agents do not collide and bad work is easy to discard. ## Computer science basics still matter A tool that runs slowly keeps the agent waiting. The MCP TypeScript kit waits 60 seconds for a request by default. Complexity decides whether a call finishes. A tool whose work grows with the cube of the input can time out. Agents also live in the terminal, so Linux and shell skills are now AI skills. The talk ended with a warning about relying on AI for thinking. The first source is a 2025 preprint, a paper shared before formal review, from MIT Media Lab. It had a small sample. Fifty-four people finished the first three sessions in three groups, which were language model, search and brain only. Brain activity was measured. The language model users showed the weakest connectivity and struggled to quote their own work. The second source is a 2026 preprint with a hypothesis. The more we delegate, the less we practise. Its human figures are an analogy, and the 2026 one is an estimate. I treat both as signals, not proof. Use AI, and keep thinking. ## A call to action for MUET students Every department uses costly tools, and agents can reach them through MCP servers. My search found community servers for civil tools like ETABS, SAP2000, Revit and AutoCAD. QGIS has some, and Esri has a beta for ArcGIS. SolidWorks and ANSYS have community servers, and Ansys publishes its own. Electrical has an official one for MATLAB. Petroleum is the gap. I found none for IPM, CMG, tNavigator or KAPPA, which is not proof that none exists. These are common industry tools, not MUET's own labs. Many servers already exist, so the value is often in checking them, fitting local standards, adding safety and making teaching material. The plan has four steps. Pick a tool another department uses. Check whether an MCP server exists. Team up with that department and build it, or validate it and fit it to local standards. Then make it a final year project. Start with one tool and one department. Build for real users, such as a shop, your society, an NGO or another department. I built my housing society's ERP, a system to run its daily work, for free. A few shipped projects, each with a live link, a clear README file and the problem you solved, beat a pile of certificates. ## The future engineer stands above the loop Software work is moving from a line to a loop. In the old line, one slow trip back meant a new release cycle. In the loop, a cycle takes hours, not weeks. The engineer stands above it, to instigate, direct and govern. The loop keeps running, and human judgment stays above it. Special thanks to my wonderful student Ameer Ali for inviting me. I’m grateful to Engineer Saqib Hussain for presenting me with the shield and mentoring the student organizers. Thank you also to Muhammad Raza Abbas, President of the Cyber Security Society and rest of the team. ![receiving an appreciation award at Cyber Nexus 2.0](https://muhammadhuzaifa.pro/storage/photos/1/muet-hackathon-linkedin-3.jpg) ### Laravel Pulse MCP Lets AI Agents Read Production Data URL: https://muhammadhuzaifa.pro/laravel-pulse-mcp-lets-ai-agents-read-production-data Published: 2026-09-29 Your AI coding agent can read every file in your project. But it has no idea what is breaking in production right now. Laravel Pulse MCP fixes that gap. ## The copy and paste loop This was my routine. Open the Pulse dashboard. Find the slow query. Copy it. Paste it into the chat. Go back for the error. Copy again. Paste again. The agent was smart. It was just blind. Every fix started with me acting as a messenger between two screens. ## What changes Laravel Pulse MCP is a free, open source package. You install it in your Laravel app. Then you connect the AI agent you already use, like Claude Code, Codex or Cursor. Now you just ask. Ask what is slow in production today. Ask which errors started this morning. Ask if the queue is backing up. The agent looks at your real Pulse data on its own. Then it opens the exact file and line behind each problem and suggests a fix. No dashboard tabs. No copying. ## Safe by design Production data needs care, so the package is strict by default. It is off until you turn it on. It can only read, never change anything. Only people you approve can connect, and they sign in through your own app. Sensitive values like tokens in URLs and email addresses are hidden before the agent sees them. ## Why this matters Laravel's hosted monitoring service, Nightwatch, already gives AI agents this kind of access. Pulse is free and self-hosted, and until now it had nothing like it. This package brings the same idea to every app that runs Pulse. I tested it in my own production app before sharing it, and it works great. ## Try it Setup takes a few minutes. All technical steps are in the README at [https://github.com/huzaifaarain/laravel-pulse-mcp](https://github.com/huzaifaarain/laravel-pulse-mcp). The package is on Packagist at [https://packagist.org/packages/huzaifaarain/laravel-pulse-mcp](https://packagist.org/packages/huzaifaarain/laravel-pulse-mcp). If it saves you time, a star on GitHub helps other Laravel developers find it. ### Building a HubSpot MCP Server for Lists and Segments URL: https://muhammadhuzaifa.pro/building-a-hubspot-mcp-server-for-lists-and-segments Published: 2026-09-12 ## A gap in the native HubSpot connector HubSpot's official MCP connector lets an AI agent call HubSpot directly, but it does not cover every endpoint. Lists and segments are missing. I built a small Laravel MCP server to fill that gap, with three tools for fetching list folders, fetching lists, and searching across them. Each tool is a thin class that calls a service class, and each service talks to HubSpot through the official hubspot/api-client SDK instead of raw HTTP. Every request runs through token auth first, and every tool call gets logged, so a failure days later is still traceable back to its exact input. ## An unexpected reuse The HubSpot SDK makes its requests through Guzzle under the hood. To see those calls in my existing debugging tools, I wired in two packages I had already built for regular web apps, [telescope-guzzle-watcher](https://github.com/huzaifaarain/telescope-guzzle-watcher) and [laravel-pulse-guzzle-recorder](https://github.com/huzaifaarain/laravel-pulse-guzzle-recorder). The first shows every outgoing HubSpot call in Telescope. The second tracks the same raw Guzzle traffic in Pulse, catching calls that Pulse's own HTTP client watcher does not see. Neither package was written with AI agents in mind. I built them to debug regular API integrations. But an MCP server is still just a Laravel app making Guzzle calls underneath, so the same tools gave me full visibility into what the agent was actually doing on HubSpot's side. It was a good reminder that solid observability tooling pays off in places I did not plan for. ### OpenAI Navier-Stokes Math Claim, What Really Happened URL: https://muhammadhuzaifa.pro/openai-navier-stokes-math-claim-what-really-happened Published: 2026-09-10 Last updated: 2026-11-13 # OpenAI and the Navier-Stokes Problem, What Really Happened In early September 2026, OpenAI said one of its AI systems had produced a proof for a version of the Navier-Stokes problem, a famous puzzle in mathematics that had stayed open for about 90 years. The claim is real, but it is also disputed, it has not been checked by independent experts, and it is now tangled up in a fight about credit and data. This post tells the story in simple words. You will learn what the problem is, how people tried to solve it over the years, what OpenAI did differently, what it is claiming, and why some respected mathematicians are uneasy. At the end, I share my own view, shaped by years in software, about what really happens to the data you hand these companies. ## What is the Navier-Stokes problem in plain words Think about water flowing in a river, smoke rising from a candle, or air moving over a car. All of that motion is fluid motion. Around 180 years ago, two scientists named Claude-Louis Navier and George Stokes wrote down a set of equations that describe how fluids move. An equation here just means a math rule that connects speed, pressure, and force. These rules work very well. Engineers use them every day to design planes, weather models, and pipelines. But there is a deep question that no one has been able to answer with full proof. The question is this. If you start with a smooth, calm fluid, do the equations always keep giving smooth, sensible answers forever. Or can the math suddenly break, with the speed of the fluid shooting up to infinity at a single point in a finite amount of time. That sudden break is called a singularity. A finite-time singularity means the break happens after a limited stretch of time, not in some far-off forever. Nobody has ever seen this happen in real water. But real water is not the point. The point is whether the equations themselves are always well behaved. This exact question is one of the seven Millennium Prize Problems. These are seven famous unsolved problems named in the year 2000 by the Clay Mathematics Institute. Each one carries a prize of one million dollars. So far only one of the seven has been solved. ## Why this puzzle stayed open for 90 years ### The chalkboard era Serious work on this started in 1934. A French mathematician named Jean Leray studied what happens when the equations might break down. He built a weaker kind of solution that always exists but might not stay smooth. For decades after that, mathematicians pushed at the problem with pen, paper, and chalkboards. They proved many partial results. They could not close the gap. ### The computer era From the 1980s onward, faster computers let researchers run detailed simulations of fluids. Some simulations seemed to show the fluid speed climbing very fast, which hinted a singularity might be possible. Others suggested the fluid always smooths itself out. A simulation can suggest an answer. It cannot prove one. A proof has to rule out every possible case with logic, not just show one example on a screen. ### The recent era of clever workarounds In the last several years, a few mathematicians found a side door. Instead of a fluid left alone, they studied a fluid with a gentle outside push added to it. In math this push is called a forcing term. With the right push, some of them managed to build examples where related fluid equations do break down in finite time. Researchers such as Diego Cordoba and Luis Martinez-Zoroa were central to these forcing methods. This work is respected, but most mathematicians see the pushed version as an easier cousin of the real problem, which asks about a fluid with no outside push at all. ## What OpenAI did differently Past attempts were made by small teams of humans, sometimes helped by a computer simulation. OpenAI took a different route. It pointed a very large number of AI agents at the problem at the same time and let them work together. According to OpenAI and news reports, the effort ran from about 1 September to 5 September 2026, roughly 88 hours of core work. At its peak it used around 10,000 AI agents running at once. These agents exchanged about 2.7 million messages and produced around 130 billion words of output. The model behind them was an internal system that OpenAI has not released to the public, described in reports as part of its next generation line beyond the current public model. The cost has been put in the millions of dollars, with some reports suggesting tens of millions. After the agents produced an argument, OpenAI says the proof was written out in Lean. Lean is a software tool that checks every logical step of a proof and complains if any step is wrong. That checking step took about another 17 hours. ## OpenAI's main claims Here is what OpenAI is saying, in plain terms. First, its AI system produced a proof that a version of the three-dimensional Navier-Stokes equations can develop a finite-time singularity. In everyday words, the fluid forms a spinning vortex that gets stretched thinner and thinner, like a strand of spaghetti, while its speed climbs without limit, even though the total energy stays finite. Second, this version of the problem uses a forcing term, that gentle outside push described earlier. Third, the argument has been formalized and checked in Lean, so the logic inside the written proof holds together. Fourth, OpenAI says it is not claiming the one million dollar prize. Fifth, OpenAI says its work began on 1 September after it heard a rumor that another group was making progress on a related fluid problem. ## The case against the claim The reaction from the math world has been careful and, in places, sharp. ### A fight over who got there first A mathematician named Tristan Buckmaster, who works at New York University, had been working with Levent Alpoge, a researcher at Anthropic, which is a direct competitor of OpenAI. The two of them had spent about a year on a closely related fluid problem. Around 15 August 2026 they reached their own blow-up results with a smooth forcing term. They published their paper on 7 September 2026. Buckmaster has said in public that OpenAI only turned its full attention to this exact problem after hearing that his team was close, and that OpenAI's route to the answer looked similar to their own. He argues that this is not the kind of path a model finds in a few days just from being handed the problem. ### The question about Codex sessions This is the part that matters most for the wider point of this article. Buckmaster and Alpoge had been putting their drafts and working notes into OpenAI Codex, which is OpenAI's coding assistant, throughout their project. Alpoge later said he received tips that word of their progress had reached OpenAI. Buckmaster says he asked OpenAI directly whether its model had been trained on, or had access to, their Codex sessions. He says he was told the model did not look up user data, but that he did not get a clear answer about training data. There was also a dispute about authorship. Buckmaster says he was offered a choice that involved publishing a paper crediting OpenAI while leaving Alpoge's name off, and that the reason given was that Alpoge works for a rival. Sebastien Bubeck, an OpenAI researcher, has denied asking for Alpoge's name to be removed. Sam Altman, the head of OpenAI, has also responded on social media. ### The math is not confirmed yet The Clay Mathematics Institute, which owns the official problem and the prize, has not accepted the result. Its president, Martin Bridson, said the review would be unhurried and fully rigorous. The problem is still listed as unsolved on the institute's website. Other mathematicians point out that OpenAI's proof needs the outside push to work. There is no public sign that the same result holds for a fluid with no push, which is the real target. Terence Tao, one of the most respected living mathematicians, noted something odd about the whole episode. There seems to be a growing gap between getting an answer and actually understanding why the answer is true. So the honest status today is this. OpenAI has a claimed proof of a forced, easier version of the problem, checked by software, not yet checked by independent human experts, and not recognized by the body that awards the prize. ## How people reacted on X Most of this fight has played out in public posts on X, formerly Twitter, and on Mathstodon, a social site used by mathematicians. The reaction split roughly three ways. Some people were excited about the technical feat. Some were skeptical about how it was framed. Many were focused on the drama over credit. Sebastien Bubeck, the OpenAI researcher at the center of it, posted that a set of false and inflammatory claims about him was going around. He said he entered the discussion following normal academic manners and was let down that it had reached this point. He said OpenAI did not use the other team's prompts or proofs to direct its agents. He also said he never asked for Levent Alpoge to be removed from the authorship of his own work, and that a screenshot being shared was simply him reaching out to line up the timing of both releases. Sam Altman, the head of OpenAI, posted that he would have much preferred coordination. He said OpenAI did not rush to publish even though the other team was not communicating with them. He said that once both sides could see each other's work, the two approaches looked different. He also said the team acted with integrity and generosity. Tristan Buckmaster's public statement is what set the story off. In it he laid out the timing, the similar route, the authorship offer, and the remark he took as a threat to his career. Terence Tao weighed in from the math side. Before the OpenAI announcement he warned on Mathstodon about the cost of turning long-standing open problems into viral benchmark moments. He praised the Buckmaster and Alpoge work as a real achievement. He also described a strange split in this whole episode, between getting an answer fast and actually understanding why it is true. Skeptical users made a sharper point. A short line like 10,000 agents solved Navier-Stokes hides how much of the work was human framing, filtering, and checking. Some also argued that OpenAI could have used patterns in Codex activity to spot which research directions looked promising, even without reading any single private file. None of that is proven. It is the kind of doubt that lingers when the systems involved cannot be inspected from outside. ## What OpenAI says about your data OpenAI has addressed the data question head on. Its statement is that the researchers and the agents did not see the other team's work by any means before it was published, and that no specific user data was accessed to solve the problem. Then comes the careful part. OpenAI also said that, while it thinks it is unlikely, it cannot fully rule out that de-identified data drawn from people using its products helped improve its models. De-identified means the obvious name tags are stripped off. It does not mean the underlying content never existed in the system. It is also worth knowing that OpenAI has said content from individual ChatGPT and Codex accounts may be used to train its models unless a user opts out. There was also a separate incident in June 2026 where the Codex tool, after a smaller screenshot failed, took a picture of a user's entire desktop with no clear warning. That was a bug, not a policy. But it shows how much these tools can see. ## My view after years in software Now the personal part. I entered this industry in 2011. Once I started to understand how software and large systems are actually built, one thing became hard to accept. A company tells you it runs a service for you, and in the same breath it tells you it does not hold or touch your data. Those two statements rarely fit together. Privacy policies, compliance badges, and trust pages are mostly the front of the shop. They are written for users to read and feel calm. Behind that front sit the servers, the logs, the backups, the training pipelines, and the internal tools that staff can use. Almost none of that back area is open to outsiders. If these companies allowed real independent audits of their servers, we could see who stores what, for how long, and who inside the company can reach it. Without that kind of open audit, we are asked to take their word for it. So I work from a simple assumption. Anything I type into Codex or ChatGPT, including a conversation like this one, may sit somewhere inside those systems in some form. The same goes for my documents. Some of them may be with OpenAI or another cloud provider today. Before that they lived on Google Drive. Before that they sat on some other cloud service. The address keeps changing. The basic bargain does not. I hand over content, and I lose the ability to prove what happens to it next. The Navier-Stokes story is a sharp example of why this matters. Two researchers used a tool for their daily work. Later they were left asking whether that same tool's owner had gained an edge from their private drafts. Whether or not that turns out to be true, the fact that a serious person cannot rule it out is the problem. ## What this means for you You do not need to stop using these tools. Most people cannot, and the benefits are real. But a few habits help. Treat any chat with an AI tool as something that could be stored and read later. Keep truly sensitive material, such as legal papers, medical records, and unreleased ideas, out of general chat tools, or use business plans that offer stronger data terms in writing. Turn off training use in your account settings if the option exists. And press for the thing that would actually fix the trust gap, which is independent, public audits of how these systems handle data. ## Frequently asked questions ### Did OpenAI solve the Navier-Stokes Millennium Prize problem Not in the way the prize defines it. OpenAI claims a proof for an easier version that includes an outside push on the fluid. The Clay Mathematics Institute still lists the problem as unsolved and is reviewing the work. ### Is OpenAI getting the one million dollars OpenAI has said it is not claiming the prize. ### What is a finite-time singularity It is the moment when the equations stop giving sensible answers. The fluid speed shoots up to infinity at a point after only a limited stretch of time. ### Why does the forcing term matter so much The real problem asks about a fluid left completely alone. Adding a gentle push makes the math easier. A result with a push does not automatically settle the harder question. ### Did OpenAI use private Codex data to get ahead There is no proof of that. A researcher raised the question, OpenAI denied accessing specific user data, and OpenAI also said it cannot fully rule out that stripped-down usage data helped train its models. ## Closing thought The headline says a machine cracked a 90-year-old puzzle in 88 hours. The real story is smaller and more human. A claimed proof of an easier case, waiting on expert review, arriving in the middle of a dispute about credit and private data. That last part is the one worth keeping in mind, long after this news cycle ends. ## Sources - [Quanta Magazine, 8 September 2026: reporting on the OpenAI announcement](https://quantamagazine.org/ai-has-solved-one-of-maths-1-million-millennium-prize-problems-20260908). The 88 hours of work, the group of about 10,000 AI agents, the extra 17 hours for the Lean check, and the competing work by Buckmaster and Alpoge. ### NYC Banned AI for Kids, Here Is Why That Is Right URL: https://muhammadhuzaifa.pro/nyc-banned-ai-for-kids-here-is-why-that-is-right Published: 2026-09-05 Last updated: 2026-11-13 # NYC Banned AI for Kids, Here Is Why That Is Right New York City Public Schools, the largest school system in the United States, announced on September 2, 2026 that it is banning generative AI tools for students from 2-K through eighth grade. I think this is the right call, and I want to explain why, using both the facts of the policy and my own view on what children lose when a machine does their thinking for them. ## What New York City Public Schools Actually Announced Mayor Zohran Mamdani and Schools Chancellor Kamar Samuels announced a one year pause on student facing generative AI. Generative AI means tools that write, answer, or create content on request, such as chatbots and AI tutors. The pause covers close to 600,000 students, about two out of every three children in the district. The rule blocks chatbots, AI tutors, AI powered instructional programs, and companion chatbots meant for emotional support. The district will turn off the AI features in more than 38 programs that schools had already approved, because those programs did not meet the new safety standards. The policy also sets limits by age. Children from pre kindergarten through second grade will not use individual devices at all, though a shared smart board in the classroom is still allowed. Students in third through fifth grade get a recommended screen time cap of 30 minutes a day. Students in sixth through eighth grade get a recommended cap of 45 minutes a day. High school is treated differently. A small number of closely watched pilot classrooms will keep testing AI tools, and every high school student will get two lessons a year on how AI works, its biases, and how to think about its answers critically. There are exceptions. Students with disabilities who rely on assistive technology, English language learners, and students in career and computer science programs are not covered by the same limits. Teachers are also treated differently from students. They can still use AI to help plan lessons, translate material, or draft messages to parents, but not to grade work, monitor behavior, offer counseling, or write a student's special education plan. This is also a reversal. New York City schools banned ChatGPT in January 2023, only two months after it launched, then lifted that ban in May 2023. The schools chief at the time admitted the first ban came from what he called knee jerk fear. An early draft of new rules in March 2026 drew pushback from parents and city council members, and this September policy is the revised version. ## Why I Agree With the Direction NYC Has Taken I want to be clear about where I stand before I go further. I am not against AI. I believe AI has already become a core part of daily life, and it is going to touch every profession in some form, whether that profession is medicine, law, engineering, or teaching itself. That much I am fairly certain about. But being certain that AI will be everywhere is not the same as saying every age group should use it in every way, right now, without limits. A tool can be useful for an adult and still be the wrong thing to hand to a five year old who is still learning how to hold a pencil and form a thought. NYC's policy draws exactly that line, and I think that line is drawn in roughly the right place. ## AI Is Not Free, Not Even for One Word There is a cost to AI that most people never see, and it belongs in this conversation. When a person solves a problem in their head, the cost is a bit of glucose and a few minutes of attention. When an AI model answers the same question, even something as short as a single word, the request still passes through a neural network, a system of billions of connected calculations built to guess the next likely word. That guess costs real resources. It costs computing time, measured in units called tokens, which are small pieces of text the model processes one at a time. It costs electricity to run the servers doing that computation. Estimates of how much electricity a single AI response uses vary widely, and companies and researchers do not always agree on the numbers, some figures are ten to fifty times higher than what AI companies themselves have reported. Water is also used to cool the data centers that house these servers, though the exact amount depends heavily on the cooling method and location, and is genuinely disputed between sources. What is not disputed is the direction, global electricity use by AI data centers is expected to grow sharply by the end of this decade. I raise this not to argue against using AI, since I use it myself every day. I raise it because a child asking a chatbot to solve their math homework is not a free action just because no money changed hands. It draws on real electricity and real infrastructure, for a task a human brain could often do at a fraction of the cost, using nothing more than a pencil and a few minutes of effort. ## The Gym Analogy, You Cannot Rent Someone Else's Muscles Think about a person trying to build muscle at the gym. No matter how advanced the machines are, that person still has to lift the weight themselves. A machine that lifted the weight for them would leave the gym looking busy, but their body would never actually get stronger. Critical thinking works the same way. No matter how powerful AI and large language models become, a person's own ability to reason, judge, and solve problems only grows if they use it. That ability is a kind of mental muscle, and it only strengthens through use, not through watching something else do the work. This matters most for children, because childhood is exactly the period when the brain is building these habits for the first time. If a child is handed a tool that solves every problem before they have struggled with it, they are not skipping a chore, they are skipping the repetition that builds reasoning itself. ## What I Am Seeing in Schools Right Now I recently went to my children's school to pick up my daughter, and one of the main points in the school's pitch to parents was its AI curriculum. The students this curriculum targets include nursery children, kindergarten students, fifth graders, and eighth graders. This is not a distant trend, it is happening now. If an eighth or ninth grade student is getting most of their homework done through AI, where does their own thinking actually happen. A child can upload a photo of a math problem to a chatbot and get a full solution back in seconds, without ever sitting with the problem long enough to feel stuck. Human intelligence naturally varies from person to person. That variation is exactly what produces scientists, inventors, and people who can solve problems nobody else can. If every child is taught that their reasoning should simply defer to whatever a language model says, that variation stops mattering, because everyone is leaning on the same layer underneath them, and the difference between someone who reasons well and someone who does not quietly disappears. ## The Cigarette Packet Problem I do not accept the argument that says, we will teach children how to use AI properly, tell them what they should and should not do, and if they ignore that advice, the responsibility is no longer ours. That argument sounds a lot like selling cigarettes, taxing them, and putting a warning label on the pack, then treating the responsibility as fulfilled. That is not how responsibility works. If a school or a company introduces a tool into a system full of children, it also takes on the responsibility to make sure that tool is used in the most positive way possible, while actively limiting its downsides. A warning label is not the same as a safeguard. Realistically, nobody can maintain constant supervision over how a child uses AI, not the school and not even the parents at home. A father is often busy with work, a mother is often busy running the household, and the child sits down to do homework alone. The child types a question into a chatbot, maybe uploads a photo of the problem, and gets a full answer back almost instantly. In that moment, where did the child's own thinking go. ## What a Child Loses When the Struggle Disappears That child was supposed to sit with the problem, feel a little stuck, and push through that discomfort. That struggle is not wasted time, it is the process that builds real thinking. Early research on AI reliance and critical thinking, while still developing and not fully settled, has started to find a pattern, people who lean on AI tools heavily for thinking tasks tend to score lower on critical thinking measures, and the effect looks strongest in younger users. Researchers have also started describing something they call digital amnesia, where students struggle to recall material from work they only recently completed with heavy AI help. There is also a more basic problem. If a child never builds their own knowledge and judgment, what will they use to check whether an AI's answer is even correct. A person with no foundation of their own has no way to catch a mistake, because they have nothing to compare the answer against. ## What Schools Should Actually Do This is where I think NYC's approach, even with its limits, points in a useful direction. It does not pretend AI does not exist. High schoolers still get regular lessons on how AI works, where it can be biased, and how to question its answers, which is a real form of AI literacy rather than a blanket ban with nothing to replace it. Younger children, meanwhile, get more hands on learning, more play, and more direct teacher guidance, at the exact age when those things build the foundation everything else stands on later. That is the balance worth copying, protect the years when a child's reasoning is still forming, and once that foundation exists, teach older students to use AI with judgment instead of blind trust. AI is not going anywhere, and I do not think it should. But a child's own mind is not something a machine can build for them, no matter how advanced that machine becomes. ## Sources - [NYC Mayor's Office press release, 2 September 2026](https://www.nyc.gov/mayors-office/news/2026/09/mayor-mamdani-and-chancellor-samuels-put-students-first-with-nat.html). The policy, the grades it covers, the number of students affected, the screen time caps, and the two yearly AI literacy lessons in high school. ### RAG vs OKF, A Guide to Claude, OpenAI, and Open Source URL: https://muhammadhuzaifa.pro/rag-vs-okf-a-guide-to-claude-openai-and-open-source Published: 2026-08-24 Last updated: 2026-11-13 Retrieval Augmented Generation, or RAG, and the Open Knowledge Format, or OKF, solve two different problems. RAG is a technique. It lets an AI model search outside documents and pull relevant pieces into its answer. OKF is a format. It is a standard way to write down and organize knowledge so any AI system can read it. You do not have to choose only one. Most real systems will end up using both. This guide explains each one in plain terms, then walks through how they show up inside three ecosystems people actually build on, Anthropic's Claude, OpenAI's models, and the open source RAG tools like LangChain and LlamaIndex. ## What RAG actually is Retrieval Augmented Generation is a pattern, not a product. A model gets a question. Before it answers, a search step runs. That search step looks through your documents, finds the most relevant pieces, and adds them to the prompt. The model then answers using both its own training and the fresh material it was just handed. The search step usually works through embeddings. An embedding is a list of numbers that captures the meaning of a piece of text. Documents get cut into chunks, each chunk gets turned into an embedding, and those embeddings get stored in a vector database. At query time, the question also gets turned into an embedding, and the system finds the stored chunks whose embeddings are closest in meaning. RAG became popular because it is cheaper and faster than retraining a model on new data. It also lets an answer point back to a real source, which matters for trust and for catching mistakes. RAG has known weak spots. Chunking can cut a sentence away from the context that explained it. A search step can return the wrong chunk, or an outdated one, if the index has not been refreshed. Keeping thousands of embeddings in sync with fast-changing source data is real ongoing work, not a one-time setup task. ## What OKF actually is Google Cloud announced the Open Knowledge Format in June 2026. OKF is an open specification for writing down knowledge in a shape that any AI agent can read, without a proprietary database, SDK, or vendor account. An OKF bundle is a folder of plain markdown files. Each file starts with a short block of YAML frontmatter, a small structured header, followed by normal markdown text. The specification requires exactly one field in that header, called `type`. Everything else, like a title, a description, or tags, is optional and left to whoever is writing the bundle. Here is what one file looks like, based on Google Cloud's own example. ``` --- type: BigQuery Table title: Orders description: One row per completed customer order. resource: https://console.cloud.google.com/bigquery?p=acme&d=sales&t=orders tags: [sales, revenue] timestamp: 2026-05-28T14:30:00Z --- # Schema | Column | Type | Description | |--------|------|-------------| | order_id | STRING | Globally unique order identifier. | | customer_id | STRING | FK to customers. | # Joins Joined with customers on customer_id. ``` This example uses the version 0.1 `timestamp` field. Version 0.2 renames it to `generated.at`. Files link to each other with ordinary markdown links, the same way a wiki page links to another wiki page. That web of links forms a graph, which is richer than a plain folder structure. Google's team has pointed to a line from AI researcher Andrej Karpathy as part of the thinking behind this, the idea that keeping a personal wiki up to date is exactly the kind of bookkeeping task an LLM is good at, even when a human would find it tedious. OKF launched as version 0.1, and Google Cloud called it a starting point, not a finished standard. Version 0.2 has since been published. It still requires only the `type` field, but it renames `timestamp` to `generated.at` and adds optional fields for sources, provenance, and status. The spec, a reference agent that drafts OKF files from BigQuery data, a visualizer, and sample bundles are public on GitHub, and OKF now has its own repository, GoogleCloudPlatform/open-knowledge-format. ## The real difference between RAG and OKF Here is the distinction that most headlines skip past. RAG is a retrieval technique. It answers the question, how does a model find the right piece of text at the right moment. OKF is a packaging format. It answers a different question, what shape should that text be in, so it stays accurate, linkable, and readable by more than one system. Google Cloud has said directly that OKF is not a replacement for RAG. Most real setups will use OKF to organize and clean up the source knowledge, then still run a RAG pipeline, or another retrieval method, on top of that OKF bundle to actually serve answers at scale. OKF suits knowledge that is curated and relatively stable, things like a table schema, a runbook, or an internal policy. RAG still wins for searching large, messy, unstructured collections, like years of support tickets or a stack of PDFs. Think of it this way. RAG is the search engine. OKF is the well organized library that search engine is searching through. A better organized library does not remove the need for a search engine once that library gets large. It just means the search engine has cleaner material to work with. ## RAG and OKF in the Anthropic ecosystem ### How Claude does RAG today Anthropic's own engineering team has published a technique called contextual retrieval. Standard RAG chunks a document and embeds each chunk on its own, which strips away surrounding context. Contextual retrieval fixes that by using Claude to write a short piece of explanatory context for each chunk before it gets embedded, so the chunk still carries the information that explains what it is about. Anthropic combines this with a traditional keyword search method called BM25, and with a reranking step, and reports that the combination cuts retrieval failures by up to 67 percent compared to plain embedding search. Anthropic also points out a case where RAG is not needed at all. For a knowledge base under about 200,000 tokens, Claude's large context window and prompt caching feature let you paste the entire knowledge base into the prompt directly. Prompt caching keeps that content ready between calls at a lower cost, so there is no retrieval step to build or maintain for smaller knowledge sets. ### Where OKF fits next to Claude Skills and MCP Anthropic already ships a markdown based system that looks close in spirit to OKF, called Agent Skills. A skill is a folder with a file named SKILL.md at its root. That file has YAML frontmatter with a name and a description, followed by markdown instructions, plus optional scripts and reference files. Claude loads a skill only when the task calls for it, instead of holding every skill in context all the time. Anthropic's Model Context Protocol, or MCP, is the separate piece that lets Claude connect to outside tools and data sources at runtime. An OKF bundle is a natural thing to expose through an MCP connector, since both are built around plain files and both avoid a proprietary storage layer. As of this writing, Anthropic has not announced first-class OKF support, so pairing the two today means writing that connector yourself, or waiting for a community integration. ## RAG and OKF in the OpenAI ecosystem ### File search and vector stores as OpenAI's RAG layer OpenAI's Assistants API was shut down on August 26, 2026, after the Responses API reached the same feature set. The Responses API now carries File Search, OpenAI's built in RAG tool. You upload files into a vector store, OpenAI parses and chunks them, generates embeddings, and stores everything for you. At query time, File Search runs both vector search and keyword search together and hands the model the matching passages. File Search accepts a range of formats, including PDF, DOCX, and plain markdown, and supports metadata filtering so you can scope a search to a subset of files. This is a fully managed RAG pipeline. You do not choose a chunking strategy or an embedding model by hand, and pricing is based on storage plus the number of search calls. ### What OKF would add on top Because File Search already accepts markdown files, an OKF bundle can be uploaded into an OpenAI vector store today without any special adapter, since a `.md` file with YAML frontmatter is still just a markdown file to that system. What OpenAI's stack does not give you is the format's real benefit, portability. Files uploaded into an OpenAI vector store are indexed and stored inside OpenAI's infrastructure. An OKF bundle kept in a plain git repository stays readable, linkable, and reusable by any other tool, including a different vendor's RAG pipeline, without re-exporting anything. Feeding OKF bundles into OpenAI's File Search is a reasonable way to get OpenAI's managed retrieval on top of vendor neutral source files, but the neutrality lives in the bundle, not in OpenAI's index. ## RAG and OKF in the open source ecosystem ### LangChain, LlamaIndex, and the RAG framework field The open source RAG landscape is large and active heading into 2026. LangChain remains the biggest and most general framework, with the widest set of integrations for models, embeddings, and vector stores, and is popular for building full agent applications, not just retrieval. LlamaIndex is more focused, built specifically around indexing and querying private data. RAGFlow and Haystack are two more actively used options, each with a large open source following. None of these frameworks define a shared, standard file format for the knowledge they index. Each project tends to bring its own loaders, its own chunking defaults, and its own document objects. ### OKF as a shared format across open source tools This is exactly the gap OKF targets. A LangChain pipeline, a LlamaIndex pipeline, and a completely different tool can all point at the same OKF bundle and load it with their own existing markdown loaders, since OKF adds nothing beyond markdown and YAML frontmatter on top of the file. Because an OKF bundle is just files in a folder, it fits naturally into a git repository, which means it gets version history, pull request review, and diffing for free, the same way source code does. A team using LlamaIndex for retrieval today could adopt OKF purely as the way source documents are organized, without changing their retrieval framework at all, since OKF sits underneath the framework rather than replacing it. Because OKF is brand new, most open source RAG frameworks have no dedicated OKF loader yet. Treat it, for now, as a well organized markdown source you point your existing loader at, not as a framework feature you can turn on with a flag. ## When to use RAG, OKF, or both Use RAG alone when your knowledge is large, unstructured, and changes constantly, like a support ticket archive or a growing collection of PDFs, and no one is going to hand curate it into a clean format. Use OKF alone, without a full RAG pipeline, when your knowledge base is small enough to fit in a model's context window, and you mainly need it well organized and version controlled rather than searched at scale. Use both together for the common middle case, a knowledge base that is curated and structured, like table schemas, runbooks, or internal policies, but too large or too fast changing to paste into every prompt. Store the source material as an OKF bundle for portability and version control, then run a RAG pipeline, whether that is Anthropic's contextual retrieval, OpenAI's File Search, or an open source framework, on top of that bundle to actually serve answers. ## Common mistakes to avoid Do not treat OKF as a search engine. It has no query layer of its own, so an OKF bundle by itself will not answer a question, it only organizes the material that a separate retrieval step will search. Do not assume every ecosystem already supports OKF out of the box. As of this writing, only Google Cloud has shipped first-class tooling. Claude, OpenAI, and most open source frameworks can consume OKF files as plain markdown, but none of them have announced dedicated OKF integrations yet. Do not skip contextual chunking when you build RAG on top of curated documents. A well organized OKF bundle still loses meaning if a chunking step slices a table away from the heading that explains it, so pair a good source format with a chunking approach that keeps context intact. Do not treat OKF as a stable, unchanging standard. It launched as version 0.1, Google Cloud called it explicitly a starting point, and version 0.2 has already renamed a field, so expect the spec to keep evolving. ## Frequently asked questions **Is OKF a replacement for RAG?** No. Google Cloud has stated directly that OKF is not a replacement for RAG. It is a format for organizing the knowledge that a RAG pipeline, or another method, later retrieves. **Who created OKF?** Google Cloud announced OKF in June 2026, through its Data Analytics and BigQuery engineering team. The specification, reference tools, and sample bundles are public on GitHub. **Does Claude or OpenAI support OKF natively?** Not as of this writing. Both can read an OKF bundle as plain markdown files, since neither format requires special tooling to parse, but neither vendor has announced dedicated OKF support. **Can I use OKF with LangChain or LlamaIndex today?** Yes, through each framework's existing markdown and YAML loaders. There is no dedicated OKF loader yet, since the specification is only a few months old. ## Sources - [Google Cloud blog post announcing the Open Knowledge Format](https://cloud.google.com/blog/products/data-analytics/how-the-open-knowledge-format-can-improve-data-sharing). The announcement, the single required `type` field, and the "starting point" wording for version 0.1. - [Open Knowledge Format on GitHub](https://github.com/GoogleCloudPlatform/open-knowledge-format). The version 0.2 specification and the change from `timestamp` to `generated.at`. - [Anthropic: Contextual Retrieval](https://www.anthropic.com/news/contextual-retrieval). The 67 percent drop in retrieval failures with reranking, and the 200,000 token guidance. - [OpenAI: Assistants migration guide](https://developers.openai.com/api/docs/assistants/migration). The shutdown of the Assistants API on August 26, 2026. ### Evaluating Bezier Curves With GPU Texture Lookups URL: https://muhammadhuzaifa.pro/evaluating-bezier-curves-with-gpu-texture-lookups Published: 2026-08-21 A new technique lets GPUs evaluate Bezier curves using the same hardware that samples textures, instead of running the usual shader math. The paper behind it, A Texture Lookup Approach to Bezier Curve Evaluation on the GPU by Muhammad Anas and Alan Wolfe, published in the Journal of Computer Graphics Techniques, shows how to pack a curve into a tiny texture and read points off it with a single hardware accelerated lookup. This article walks through how the technique works, why it can be faster, where it falls short, and what it costs in numerical precision. ## The problem with evaluating curves on a GPU A **Bezier curve** is a smooth curve defined by a small number of points called **control points**. Games and design tools use Bezier curves everywhere, for font outlines, camera paths, motion trails, and particle movement. To find a point on the curve at some position, usually written as a value called `t` between `0` and `1`, a GPU normally runs the **de Casteljau algorithm**. This method repeatedly blends pairs of control points together, a process called **interpolation**, until only one point is left. For a cubic Bezier curve, the most common type, this takes several rounds of blending done directly in shader code. A **shader core** is the part of the GPU that runs small programs to compute colors, positions, and other per pixel or per vertex values. Running de Casteljau there means every curve evaluation competes with all the other math the shader is already doing, lighting, shading, physics, and anything else in the frame. ## The core idea, use the texture unit instead of the shader GPUs contain a second, mostly unrelated piece of hardware built for a completely different job, sampling textures. A texture is an image mapped onto a surface, and the **texture sampling unit** is dedicated circuitry that reads pixels from that image and blends nearby ones together, a process called **bilinear interpolation**, extremely fast and almost for free from the shader's point of view. Anas and Wolfe's insight is that bilinear interpolation is, mathematically, doing the same kind of blending the de Casteljau algorithm needs. So instead of asking the shader to compute the blending steps directly, they encode a curve's control points into a small texture. A single texture lookup, sent to hardware that is normally busy with unrelated image sampling, then returns a point on the curve. Traditional approach. Control points feed into shader arithmetic. Several interpolation steps run one after another inside the shader. The output is a point on the curve. Texture lookup approach. Control points get encoded into a small texture ahead of time. The GPU's texture sampling hardware performs the interpolation during a normal texture read. The output is a point on the curve. ## Packing a cubic Bezier curve into a 2 by 2 texture The paper shows that a cubic Bezier curve, defined by four control points, can be encoded using a method called **Seiler interpolation** into a texture as small as two pixels by two pixels. Seiler interpolation is a way of arranging values in a small grid so that standard bilinear texture sampling reproduces the intended blend when the texture is read at the right coordinates. With the control points packed into that tiny 2 by 2 texture, a single bilinear texture lookup at the appropriate texture coordinate returns the correct point on the curve. No loop, no repeated shader side blending, just one hardware texture read. This is a genuinely different way of framing the problem. Instead of treating curve evaluation as a math operation, it treats curve evaluation as an image sampling operation, and lets the GPU's existing image sampling hardware do the work it was built to do. ## Why this works, shader cores and texture units are separate hardware The reason this can be a real performance win, and not just a curiosity, comes down to GPU architecture. Shader cores and texture sampling units are physically separate resources inside a GPU. They can, to a meaningful degree, do work in parallel. If an application already keeps its shader cores fully busy, and the texture units are sitting comparatively idle, moving curve evaluation from the shader to the texture unit does not compete for the same resource. It shifts work to hardware that has spare capacity, rather than adding more work to hardware that is already the bottleneck. This also means the technique's benefit depends entirely on the balance of a given workload. A scene that is already texture heavy, with texture units doing most of the work, will not see the same gains as a scene that is shader heavy with texture units mostly idle. The benchmarks later in this article show exactly that pattern. ## Beyond one curve, surfaces, volumes, and splines The paper does not stop at a single cubic Bezier curve. The same texture lookup idea extends to several related shapes used in graphics and computer aided design. **Bezier surfaces**, two dimensional grids of control points that define curved surfaces, and **Bezier volumes**, the three dimensional equivalent, can both be encoded into larger textures and read with the same kind of hardware accelerated lookup. **Piecewise curves**, made of several Bezier segments joined together, rational polynomials, and general polynomial evaluation are also covered. The authors also address **B-splines** and **NURBS**, two curve and surface types widely used in industrial design, animation, and CAD software. A B-spline is a flexible curve built from multiple smaller curve segments joined smoothly together, and NURBS extends that same idea with additional weighting for complex three dimensional shapes. ## Turning a B-spline into pieces a texture can hold B-splines and NURBS are not natively in the same simple form as a single Bezier curve, so the paper describes converting them first. It uses a known method called **Boehm's algorithm** to break a B-spline into a series of ordinary Bezier segments. Once the B-spline exists as a sequence of Bezier pieces, each piece can be encoded into its own small texture and evaluated with the same texture lookup approach used for a single cubic curve. This means the technique is not limited to simple curves used for camera paths or motion. It can, in principle, extend into more demanding surface modeling work as well, though the paper's own performance testing focuses mainly on simpler curve cases. ## Benchmark one, a compute heavy path tracer To measure real performance, the authors tested a compute heavy **path tracer**, a rendering method that simulates the path of individual light rays through a scene, at a resolution of 1920 by 1080 pixels with 16 rays traced per pixel. On an RTX 4090 graphics card, the standard shader based method took 1.952 milliseconds per frame, and the texture lookup method took 1.938 milliseconds. On an RTX 3070, the shader method took 12.662 milliseconds, and the texture method took 12.496 milliseconds. Both results favor the texture lookup method, but only slightly, around 0.7 percent faster on the RTX 4090 and around 1.3 percent faster on the RTX 3070. In a workload like this path tracer, where a huge amount of other compute work is already happening per pixel, the curve evaluation itself is a small enough slice of total work that shifting it to texture hardware barely moves the needle. ## Benchmark two, a large particle simulation The second benchmark tells a much clearer story. The authors simulated large numbers of particles moving along cubic Bezier curves on an RTX 4050, varying the particle count. | Particle count | Shader based method | Texture lookup method | |---|---|---| | 1 million | 6.9 milliseconds | 6.9 milliseconds | | 5 million | 13.9 milliseconds | 10.3 milliseconds | | 10 million | 23.2 milliseconds | 18.6 milliseconds | At 5 million particles, the texture lookup method cut frame time by roughly 26 percent. At 10 million particles, it cut frame time by roughly 20 percent. In this workload, curve evaluation is a much larger share of the total work per frame, since the entire simulation is essentially millions of repeated curve lookups. That is exactly the kind of workload where freeing up the shader cores by moving work to the texture unit pays off. ## Why the method does not always win The paper is careful not to present the texture lookup method as a universal speedup. The authors report cases where it actually made performance worse, because the additional memory access needed to read the encoding texture could not overlap well with the rest of the computation already happening on the GPU. This matters for anyone considering the technique for a real project. It is not a drop in replacement that helps everywhere. It helps specifically when shader cores are the bottleneck and texture units have spare capacity, and it can hurt when memory bandwidth or texture unit load is already the limiting factor. The right move is to profile a specific workload before assuming the technique will help. ## The precision cost and the hybrid fix Texture sampling hardware does not always operate at the same numerical precision as shader math. **Texture coordinates**, the values that specify where in the texture to read, can be stored and processed using a more compressed, **fixed point** format inside the GPU rather than full **floating point** precision. Fixed point storage uses a fixed number of digits after the decimal point, rather than the flexible, wider range that floating point numbers allow. The paper measures this directly and shows that relying entirely on texture hardware interpolation can introduce small but visible errors in the shape of the resulting curve, particularly for cases needing high precision. Their solution is a **hybrid approach**. Part of the interpolation happens through the texture hardware, and the remaining stage is finished in the shader using full 32 bit floating point precision, the standard high precision number format used throughout graphics programming. This limits the exposure to fixed point error while still capturing most of the performance benefit from offloading part of the calculation to the texture unit. In short, the real design choice is between full shader based computation with full precision, and partially texture based computation with a small, correctable precision cost. Which one makes sense depends on how much precision a given application actually needs and how much shader headroom it has to spare. ## What this means for real graphics engines The technique is most useful in **shader bound** scenarios with high curve evaluation volume, such as large particle systems, procedural animation, or any pipeline generating many curve samples per frame. It is less useful, and can even hurt performance, in workloads that are already **memory bound** or texture unit heavy. The paper includes supporting implementations covering both **DirectX 12** and **WebGPU**, two of the graphics APIs used to program modern GPUs directly, along with the path tracer and particle simulation used for testing. That gives engineers a concrete starting point for trying the technique against their own workloads rather than working from the math alone. The broader lesson generalizes past Bezier curves specifically. GPUs contain several kinds of **fixed function hardware** built for narrow purposes, texture sampling among them, and workloads that are bottlenecked on general purpose shader cores can sometimes find real headroom by reframing part of their math as an operation that fixed function hardware already does well. ### DeepSeek Harness Explained, the Open Source Agent Framework URL: https://muhammadhuzaifa.pro/deepseek-harness-explained-the-open-source-agent-framework Published: 2026-08-17 DeepSeek Harness is a free, open source framework for building AI coding agents. DeepSeek AI released it on August 13, 2026, as a developer preview under the MIT license. It gives developers a way to build agents like Claude Code or OpenAI Codex, but with every part of the system open for them to change. ### The core idea, everything is a plugin Most coding agents ship as one fixed package. The model, the tools, the file editor, and the user interface all come bundled together. DeepSeek Harness takes a different approach. It treats every part of the agent as a plugin, a small piece of code you can swap out or replace. This covers the model the agent uses, its tools, its skills, how it saves sessions, where it runs code safely, how it stores data, its main loop, its task scheduling, and its user interface. A developer can replace any one of these pieces without touching the rest of the system. DeepSeek pairs this with a second rule, every run must be traceable. The system keeps an append only log, a record that only grows and is never edited, of what happened during a run. This log stores the system prompts, the model's reasoning steps, every tool call and its result, and any work handed off to a sub agent. Because the log is complete, a developer can resume a run, fork it into a new branch, search through it, or replay it step by step. ### Cordis, the engine under the hood DeepSeek Harness is built on Cordis, a smaller framework that handles the plugin system itself. Cordis treats a shared context as a kind of directory. A plugin can register itself under a name, such as `tools` or `llm`, and other parts of the system find it by that name instead of importing it directly. Cordis also manages load order. If one plugin needs another, it declares that dependency, and Cordis works out the correct order to start everything. This is a workflow heuristic worth flagging, not a Google ranking rule, but it matters here because it is the actual technical reason DeepSeek Harness can be this modular without breaking. ### The four runtime modes DeepSeek Harness ships with four modes, each suited to a different kind of task. **Standard mode** is the full coding agent. It reads and edits files, runs shell commands, searches your code and the web, keeps a plan, and can hand off work to sub agents. **Code mode** has the model write TypeScript code to carry out multi step operations, instead of making many separate tool calls. **Minimal mode** strips the agent down to just a shell and a file editor. DeepSeek used this exact mode to benchmark its new V4-Pro model on public coding agent tasks, which is worth knowing if you compare those benchmark numbers to other tools. **Creator mode** is for building and testing your own custom presets, with tools to inspect the system while it runs. ### How to install and run it The fastest way to try DeepSeek Harness is through npm, which requires Node.js installed on your machine. ``` npx @deepseek-ai/dsh web ``` This command starts a local web interface, by default at `http://127.0.0.1:3080`, where you pick a workspace folder and approve any sensitive actions the agent wants to take. Developers who want to build from source can clone the GitHub repository and build it with pnpm instead. Because this is a v0.1 developer preview, the project's own documentation warns that there will be compatibility breaking changes. Treat any code or workflow you build on it today as likely to need updates. ### What it can already do Even in preview, DeepSeek Harness already inspects repositories, edits files, runs shell commands, searches files and the web, keeps a running plan, invokes skills, delegates tasks to sub agents, and enforces approval policies before risky actions run. These are the same core abilities that make Claude Code and Codex full coding agents rather than simple autocomplete tools. ### DeepSeek Harness vs Claude Code and Codex DeepSeek Harness is not a finished replacement for Claude Code or Codex yet. It is best understood as an open, model agnostic alternative to the agent infrastructure that sits underneath them. Here is how the three compare, based on public reporting and each project's own documentation as of August 2026. | Area | DeepSeek Harness | Claude Code | OpenAI Codex | |---|---|---|---| | Read, edit, and test a repo | Yes | Yes | Yes | | Shell and dev tools | Yes | Yes | Yes | | Planning and sub agents | Yes | Yes | Yes | | Permissions and sandboxing | Yes, set through plugins | Yes, mature built in system | Yes, granular controls | | Interfaces | Local web app, headless command, Python SDK | Terminal, VS Code, JetBrains, desktop, browser, mobile, Slack | CLI, IDE extension, desktop app, web, other integrations | | Hosted background agents | Not offered as a DeepSeek managed service | Yes | Yes | | GitHub pull request workflow | Not a finished integration yet | GitHub Actions, automatic reviews, issue to PR flows | Cloud tasks, automatic reviews, PR fixes, GitHub Action | | Model choice | DeepSeek, Anthropic, OpenAI, and other compatible endpoints | Mainly Claude models | Mainly OpenAI models, configurable in the open source CLI | | Maturity | Developer preview, breaking changes expected | Established commercial product | Established commercial product plus an open source CLI | | License | MIT | Commercial, with extension points | CLI is open source, cloud and app services are managed | The clearest gap today is around finished products built on top of the agent loop, things like hosted background agents and a polished GitHub pull request workflow. Claude Code and Codex have both built those out. DeepSeek Harness has the underlying pieces, approval policies, sub agents, tool calling, but has not yet packaged them the same way. The clearest strength is openness. Because the model itself is just another plugin, a team can point DeepSeek Harness at DeepSeek's own models, Anthropic's models, OpenAI's models, or a custom endpoint, without being tied to one vendor. ### Why this launch matters DeepSeek Harness arrived alongside the general release of DeepSeek-V4-Pro, an updated model built for agent workloads. That timing is not a coincidence. Until now, competition between AI labs was mostly about model quality and token price. Harness moves DeepSeek into a different layer, the software that decides how a model calls tools, edits files, manages sessions, and runs long, multi step tasks. That layer is harder for a developer to swap than a model. You can often change which model an application calls with a small config edit. Replacing the whole system that manages your agent's tools, memory, and workflow is a much bigger job. By open sourcing that layer, DeepSeek is competing for a stickier, more durable part of the developer relationship, not just for the top spot on a benchmark chart. ### Current limits, it is still a preview Be clear eyed about where the project stands today. It is a v0.1 developer preview. The maintainers say plainly that breaking changes are coming. It does not yet offer a DeepSeek managed hosted agent service, and its GitHub pull request workflow is not yet a finished feature the way it is in Claude Code or Codex. If you plan to build production tools on it, expect to update your code as the framework changes. ### Should you try it If you want to experiment with a fully open, swappable agent framework, or you want to run agents against more than one model provider without being locked into one vendor's stack, DeepSeek Harness is worth a look today. If you need a stable, fully supported product with hosted background agents and a mature GitHub workflow right now, Claude Code or Codex remain the safer choice until Harness matures past preview. ### Pakistan at 79, What We Are Teaching Our Civil Servants URL: https://muhammadhuzaifa.pro/pakistan-at-79-what-we-are-teaching-our-civil-servants Published: 2026-08-14 Last updated: 2026-11-13 Pakistan turns 79 today, and one of the more serious things the country built this year is a four-level artificial intelligence curriculum for its civil servants. The Civil Services Academy in Lahore, working with atomcamp and two federal ministries, now runs AI training as a required part of officer education. I read the manuals. This post explains what is in them, and where I think they are strong and weak. ## Happy Independence Day To every Pakistani reading this, at home or abroad, happy Independence Day. 79 years is not a long time for a country. It is long enough to see what we are good at, and we are good at building things under constraint. I want to mark the day by writing about something concrete rather than something sentimental. National pride is easier to feel when you can point at work. So let me point at some work. ## What the four AI manuals actually are The Government of Pakistan, through the Ministry of Information Technology and Telecommunication and the Ministry of Planning, Development and Special Initiatives, partnered with the Civil Services Academy and the training company atomcamp to produce a structured AI course for public servants. It is not a single workshop. It is a 4 level series. The levels are AI 101, AI 201, AI 301, and AI 401. Each is a two-day course with its own manual. The first volume covers 101 and 201. The second volume covers 301 and 401, and was reviewed and edited by Dr Momina Moetesum. The progression is deliberate. AI 101 teaches what AI is. AI 201 teaches how to govern an AI system you did not build. AI 301 teaches how to actually deploy one inside a ministry. AI 401 teaches how a state controls the hardware and data infrastructure underneath all of it. According to reporting from the launch, AI 101 is now mandatory at the Civil Services Academy. Around 150 probationary officers went through a two-day programme, and 30 faculty members from civil service training institutions were trained as master trainers so the course can run without the original vendor. That last detail matters more than the headline number. Training the trainers is the difference between a programme and an event. ## AI 101 and AI 201, the foundation AI 101 is a beginner course and it reads like one, in a good way. It defines AI as a system that learns from data and uses that learning to predict, recommend, or act. It walks through the three layers of an AI system, which are computing infrastructure, data and data management, and processing and modelling. It also makes a point I wish more corporate AI decks made. AI is not new. The manual traces it through mathematics, statistics, programming, and logic, notes John McCarthy naming the field in 1955, and says plainly that what changed recently is accessibility, not the underlying idea. The strongest line in AI 101 is the one it repeats. AI does not think and does not exercise judgement. It produces outputs from data, design assumptions, and the objectives it is given. Responsibility for the decision stays with the human official. AI 201 turns that principle into a job description. It tells civil servants they are AI product owners and system architects, not developers. They define what the system should achieve. They ensure it matches law and public interest. They oversee vendors. Somebody else writes the code. The module gives officers five questions to ask about any proposed AI system. Where is data stored. Is sensitive data sent to external cloud services. How is privacy protected. What happens if the AI makes an error. Is there a human review step. Those five questions are better procurement guidance than most enterprise buyers in the private sector operate with. ## AI 301, where the engineering starts AI 301 is the point where I stopped skimming. It is a genuine technical manual, split into two lanes. Lane A is for governance and policy officers. Lane B is for technical delivery teams. Both lanes share the modules, but Lane B sections go into transformer internals, retrieval engineering, and deployment mechanics. The course sets six control objectives that every implementation must satisfy before going live. Data sovereignty by design. Grounding outputs on approved sources. Role-based access with full auditability. Human approval gates on anything affecting rights, penalties, or entitlements. Protection against prompt injection and insecure output handling. Operational limits on tokens and cost. ### Why smaller models, not bigger ones The core argument of AI 301 is that model size is not a proxy for value, and governance is. It says frontier models operated by a handful of foreign firms are the wrong default answer for most Pakistani government workloads. Not because they lack capability, but because sending sensitive government data to them creates sovereignty risk, uncontrollable cost, and accountability gaps. So the manual pushes small language models, which are compact models that can run on hardware a ministry can actually own. It gives a model selection matrix covering Llama 3, Phi-3, Gemma 2, and Qwen 2.5, with minimum VRAM figures, licence types, and Urdu capability ratings for each. VRAM is the dedicated memory on a graphics card, and a model has to fit inside it to run well. The manual also insists that open weights do not mean unrestricted use, and that legal and procurement teams must review every licence before deployment. That is a small line with large consequences, and it is correct. The recommended techniques are RAG and LoRA. RAG, short for retrieval augmented generation, connects a model to an approved document store so answers are grounded in real sources and can cite them. LoRA is a fine-tuning method that adapts a model's tone and format without retraining all of its weights. QLoRA adds compression so the work fits on a single government server. This is a realistic stack. It is what a competent team with a limited budget would actually choose. ### The Urdu problem nobody can skip The section I found most valuable is the one on Urdu and multilingual AI, because it is the part no imported curriculum would contain. Pakistan's government runs in Urdu and English. Citizens write in Urdu, in Roman Urdu, and in Sindhi, Punjabi, Pashto, and Balochi. The manual states directly that no existing open-source model handles that full range well. It then lists the failure modes with their governance consequences. Code-switching inconsistency, where a model mixes Urdu and English inside one reply, produces informal official communication. Register mismatch, where the model writes casual Urdu in a formal notice, damages institutional authority. Transliteration inconsistency breaks record searchability, because the same word gets spelled differently across responses. Hallucination rates are higher in low resource languages than in English. Script confusion can garble Nastaliq against Arabic, Farsi, or Shahmukhi. There is a cost dimension too. Under standard tokenization, Urdu script produces roughly 40 to 80 percent more tokens per word than English. Tokens are the units a model is billed and limited by. That means the same document costs meaningfully more to process in Urdu, and government budgeting has to account for it. The manual's conclusion on regional languages is the honest one. For Sindhi, Pashto, and Balochi, current systems should be limited to classification and routing, not direct citizen communication, until better models exist and are tested. I have not seen a clearer public statement of this constraint from any Pakistani institution. ## AI 401, sovereignty as an infrastructure question AI 401 moves up a level. Its guiding principle is that sovereignty in AI is not determined by which model you pick. It is determined by where computation happens, how data moves, and who controls the stack. It splits sovereignty into three layers. Hardware sovereignty is where computation physically occurs. Data sovereignty is where data is stored, processed, and replicated. Operational sovereignty is who controls deployment, access, and model behaviour. The manual then audits what Pakistan actually has. The National Data Centre serves over 100 federal agencies but has no GPU infrastructure yet. NADRA runs the most operationally advanced AI in the country through biometric verification, but it is not connected to any national AI platform. Raast, the instant payment system, processes millions of daily transactions and is described as the highest return use case for real-time fraud detection, yet it has no AI inference layer. There are more than 22 commercial data centres, concentrated in Karachi, Lahore, and Islamabad, but they are judged suitable only for lower sensitivity workloads without a sovereign cloud certification framework. That is a fair assessment. It is also a more candid inventory than most official documents publish. ## The phased plan, and what it would cost AI 401 proposes a three-phase build. Phase 1 runs from 2026 to 2027 at an estimated 60 to 150 million US dollars. It deploys 200 to 500 AI-grade GPUs at the National Data Centre, commissions an Urdu foundation model fine-tuned on Pakistani legal and government text, connects NADRA to that cluster, and puts fraud detection on Raast. Phase 2 runs from 2027 to 2030 at 300 to 600 million US dollars. It expands to 1,000 to 2,000 GPUs, builds an inter-agency data exchange layer modelled on Estonia's X-Road, and launches a single government AI portal billed on usage. Phase 3 starts in 2030. It imagines a dedicated AI factory of 5,000 to 10,000 GPUs drawing 50 to 100 megawatts, plus a shared sovereign compute pool for the wider OIC bloc. The immediate actions listed for 2025 and 2026 need no new hardware at all. Enact the Personal Data Protection Bill with AI localisation clauses. Issue a sovereign compute policy. Commission an infrastructure audit. Build an AI procurement framework under existing public procurement rules. Run civil service AI literacy training. ## What the manuals get right Three things stand out. First, the governance-first posture is consistent across all four levels. AI is advisory. Humans decide. That is stated in AI 101 and still stated in AI 401, which suggests one editorial mind held the line. Second, the content is written for Pakistan's real conditions. It accounts for unreliable power, uneven broadband, scanned documents with poor OCR quality, and mixed-script corpora. It recommends hybrid keyword and vector retrieval specifically because government archives are messy. That is not generic material with local place names dropped in. Third, the use cases are specific and appropriately constrained. Policy drafting with enforced citations. Tax audit risk flagging where no AI output may ever be the sole basis for enforcement action. Beneficiary eligibility screening where the officer decides. Each one names the deployment zone, the human gate, and the failure risk. ## What still worries me The plan assumes a legal foundation that does not exist yet. As of 2026, Pakistan still has no enacted comprehensive data protection law. The Personal Data Protection Bill remains a draft. PECA 2016 is a criminal statute, not a privacy framework. Every sovereignty control in AI 401 is easier to write than to enforce without that law. Funding is the second gap. Phase 1 alone needs 60 to 150 million US dollars. Nothing in the manual explains where that comes from, and infrastructure plans in Pakistan have a long history of stalling at exactly this step. Third, capacity is thin. The manuals ask civil servants to evaluate LoRA configurations, VRAM budgets, and retrieval architectures. Two days of training gives you vocabulary, not judgement. Without a permanent technical cadre inside government, ministries will keep depending on the vendors they are supposed to be supervising. None of these make the curriculum less worthwhile. They are the reason it exists. But a training programme cannot substitute for a law, a budget line, and a hiring plan. ## What you can do with this If you build software in Pakistan, read AI 301. It is the clearest public statement of what the government will and will not accept in an AI system, which makes it the closest thing we have to a requirements document for public sector work. If you are a civil servant, the five questions in AI 201 are enough to start. Ask them of every vendor. If you work on Urdu or regional language models, the failure mode table in AI 301 is a research agenda. Somebody has to fix transliteration consistency and formal register in Urdu generation. There is no reason it should not be a Pakistani team. 79 years in, the most useful patriotic act available to most of us is to be very good at something and do it here. Happy Independence Day. ## Sources - [ProPakistani: Govt introduces new training module for civil servants (10 January 2026)](https://propakistani.pk/2026/01/10/govt-introduces-new-training-module-for-civil-servants/). The AI 101 launch, the 150 probationary officers, the 30 faculty trained as master trainers, and AI 101 becoming mandatory for future batches. - [Civil Services Academy Lahore: Artificial Intelligence for the Public Sector](https://csa.gov.pk/?p=23675). The academy's own page on the programme, run with the Ministry of IT and Telecommunication and atomcamp. - The details of AI 201, 301 and 401 come from the course manuals themselves, which I read for this post. ### HubSpot Quote Approvals, Professional Vs Enterprise Explained URL: https://muhammadhuzaifa.pro/hubspot-quote-approvals-professional-vs-enterprise-explained Published: 2026-08-11 If you are trying to block a HubSpot quote from being published until someone approves it, the tier you are on decides how much of that you can build natively. Advanced Approvals, the feature that supports multi step approval chains, requires Revenue Hub Enterprise. Standard Approvals, a single approval rule, is available on Revenue Hub Professional. That distinction sounds simple until you try to enforce a real business rule, such as blocking a quote when a line item's price falls below that specific product's floor price, across a catalog where every product has a different floor. This post walks through what each tier actually supports, the workaround that makes a per product rule possible on Professional, and one specific limitation we found only by testing it live, not by reading documentation. ### The short answer, which tier do you need If your approval logic is a single condition, someone above a certain deal size, a specific product category, one discount threshold, Standard Approvals on Professional is enough. If you need multiple approvers in sequence, different logic per step, or conditional routing such as legal only being involved after a sales manager approves, you need Advanced Approvals, which requires Revenue Hub Enterprise. Neither tier, on its own, lets a condition compare one property to a different property that varies per record, which is the gap this post focuses on. ### What Advanced Approvals actually requires Advanced Approvals is built for organizations that need multi step approval chains with sequential routing and dynamic logic. A common example is a sales manager approving first, with legal only entering the chain once the sales manager has signed off. This tier of functionality sits behind Revenue Hub Enterprise. Standard and advanced approvals cannot run at the same time in one account, you choose one model for your whole approvals setup. ### What Standard Approvals gives you on Professional Standard Approvals lets you configure one approval rule based on quote properties, line items, or deal details, and you choose whether all assigned approvers must approve or whether any single approver is enough. This is a real, usable feature, not a stripped down placeholder. The limitation is scope, one rule, not a chain, and the rule's condition builder compares a property to a value you type in, not to another property on the same or a related record. ### Why a static condition builder cannot enforce a per product floor price Here is where teams get stuck. Say you want a rule that reads, block this quote if a line item's price is below that product's floor price. The condition builder lets you filter on a line item's unit price using operators like is less than, but only against a static number you enter once. If every product in your catalog has its own floor price, a single typed number cannot represent that rule for the whole catalog. You would need a separate rule per product, which does not scale past a handful of items, let alone a catalog with tens of thousands of SKUs. ### The workaround, a workflow computed approval flag The fix is to move the comparison logic out of the condition builder and into a workflow. A quote based workflow with a custom code action can read the quote's associated line items through the CRM API, look up each line item's product to get its floor price, compare that to the quoted price, and write the result to a boolean property on the quote, something like requires approval, true or false. The Standard Approvals rule then does the simple part, it checks whether that one property is true, and blocks the quote if so. This pattern works because the hard part, per product comparison across a large catalog, happens in code where it belongs, and the approval rule itself stays a single, static condition, exactly what Standard Approvals is built to handle. We built and tested this end to end, a line item quoted below its product's floor price correctly set the flag to true and blocked publish, and raising the price back above the floor correctly cleared the flag on the workflow's next run. ### The limitation nobody mentions, editor autosave can revert the flag This is the part that only shows up once you test the full editing flow, not just the workflow in isolation. We ran a controlled test using the CRM API's property history, which records every write to a property along with a timestamp and a source. The sequence looked like this. A line item was quoted below floor, the workflow correctly set the approval flag to true. The price was then raised above floor, and the workflow correctly re ran and set the flag back to false, confirmed by timestamp. Then, an unrelated field was edited and saved in the quote editor, not the price, not the approval flag itself. That save flipped the approval flag back to true. The property history showed this write came from the quote editor's own save path, not from the workflow, meaning it was not a re run of our automation and not a manual click on the checkbox. The likely explanation, the quote editor loads the approval property into its own frontend state when the page opens, and any later save in that same editing session resubmits its cached copy of the quote, overwriting whatever the workflow wrote in the meantime. That means the flag can drift out of sync with reality during a normal editing session, not just at the moment someone clicks submit for approval. The practical effect, floor price detection itself is fully automatic and reliable, the workflow always computes the right value when it runs. What is not reliable on Professional is whether that correct value is still on the record by the time a rep actually submits the quote for approval, because routine editing in the quote UI can revert it behind the scenes. This is a real platform behavior, not a mistake in the workflow's own logic, and it is worth testing directly in your own portal before treating a workaround like this as production ready. ### Which approach fits your team If your approval need is one rule and your team can adopt a simple habit, such as refreshing the quote editor before submitting for approval, Professional plus a workflow computed flag is a workable, tested path, and it avoids an Enterprise upgrade. If your approval need genuinely requires multiple sequential approvers with different logic per step, or if the editor autosave behavior described above is not an acceptable risk for your process, Advanced Approvals on Revenue Hub Enterprise is the more direct, more reliable path, because its condition evaluation is not dependent on a workflow written property surviving an editing session. ### Frequently asked questions **Does HubSpot Standard Approvals support comparing one property to another property.** No, the condition builder compares a property to a static value you type in, not to another property on the same or a related record. A workflow with a custom code action is the way to compute a dynamic comparison and hand the result to Standard Approvals as a single boolean property. **What HubSpot plan do I need for Advanced Approvals.** Revenue Hub Enterprise. Standard and advanced approvals are mutually exclusive within one account. **Can a workflow reliably keep a custom approval property accurate on Professional.** The workflow itself computes the correct value every time it runs. Whether that value stays on the record through a normal editing session is a separate question, our testing found the quote editor's own autosave can revert a workflow written property, worth verifying in your own portal before relying on it. ### Oracle Bans AI Generated Code From OpenJDK URL: https://muhammadhuzaifa.pro/oracle-bans-ai-generated-code-from-openjdk Published: 2026-08-11 ## Oracle Bans AI Generated Code From OpenJDK Oracle's OpenJDK Governing Board has banned AI generated content from OpenJDK contributions. The policy, published in April 2026, prohibits code, text, and images produced in whole or in part by large language models, diffusion models, or similar deep learning systems from entering OpenJDK repositories, pull requests, emails, wiki pages, or bug reports. The decision stands out because Oracle is simultaneously one of the most aggressive corporate spenders on AI infrastructure in the industry, with a seventy billion dollar commitment to AI compute. The same company that wants to sell the world AI infrastructure does not want AI writing the code for its own flagship open source project. ### What the OpenJDK policy actually bans The policy text is direct. Contributions must not include content generated, in part or in full, by large language models, diffusion models, or similar deep learning systems. This applies broadly, covering source code, documentation, pull request descriptions, mailing list messages, wiki edits, and entries in the JDK Bug System. The scope leaves little room for partial use. According to the policy's own FAQ, a contributor who has an AI tool write one hundred lines of code and then edits only ten of those lines still cannot submit the result. The contribution is still considered partly AI generated, and that alone disqualifies it. ### What counts as AI generated content Any output from a large language model, a diffusion model, or a comparable generative system counts, regardless of how much a human later edits it. There is no threshold of human modification that makes AI assisted output acceptable under this policy. ### What contributors can still do Developers may still use AI tools privately. The policy allows using generative AI to debug code, review existing code, and support personal research and understanding. The restriction applies only to what gets submitted to the project, not to how a contributor works while preparing that submission. ## Why Oracle says it made this call The OpenJDK Governing Board gave three reasons for the ban. ### Reviewer burden OpenJDK relies on a limited pool of human reviewers. A flood of AI generated submissions that look plausible but contain subtle errors would consume reviewer time that the project does not have to spare. ### Safety and security The JDK underpins software running in banks, government systems, and countless production environments worldwide. The board argues that plausible looking but incorrect code creates unacceptable risk in software at that level of criticality. ### Unresolved intellectual property questions Every OpenJDK contributor signs the Oracle Contributor Agreement, which requires the contributor to own the intellectual property rights they are granting to Oracle without restriction. Whether a person actually owns the rights to output produced by an AI model is, in the board's own words, the subject of active litigation. Oracle is not willing to accept contributions built on IP ownership that courts have not yet settled. ## The contradiction, Oracle's own AI bet ### A seventy billion dollar infrastructure commitment Oracle has committed seventy billion dollars to AI infrastructure, an increase from a previous commitment of fifty five point seven billion dollars. That spending has not come without cost to the company's financial position. S&P downgraded Oracle's credit rating to BBB minus, just one notch above junk status, citing an uncertain path to making money on the investment. Oracle has also cut twenty one thousand jobs, a move the company has tied to its AI deployment strategy. ### Ellison's public comments on AI written code Oracle co-founder Larry Ellison has spoken openly about AI writing code inside the company. He has described telling AI models what the company wants built and letting the model produce the step by step process to get there. Oracle co-CEO Mike Sicilia has said AI tools let smaller engineering teams deliver more complete solutions more quickly. Neither executive has publicly explained why code produced this way is acceptable for Oracle's commercial and internal work but not acceptable for the open source project Oracle stewards for the rest of the software industry. ## GraalVM took the opposite position The contradiction gets sharper inside Oracle's own portfolio. GraalVM, a polyglot virtual machine project also backed by Oracle, adopted a policy that permits AI assisted contributions. Contributors may use AI coding assistants when preparing submissions, provided they take full accountability for the result, understand the code completely, and accept that maintainers retain full discretion to reject anything a contributor cannot demonstrate they understand. GraalVM's approach was explicitly modeled on the Linux kernel's AI coding assistants policy, which takes a similar accountability first stance rather than a blanket prohibition. Two projects under the same corporate umbrella, governed by the same Oracle Contributor Agreement, reached opposite conclusions about the same underlying legal and technical risk. ## What this means for open source contributors For anyone contributing to OpenJDK, the practical guidance is simple. Use AI tools for learning, debugging, and reviewing your own understanding of the codebase, but write and submit your actual contributions yourself. Reviewers are likely to scrutinize submissions more closely going forward, and a policy violation risks rejection regardless of how much human editing occurred afterward. For maintainers of other open source projects, OpenJDK's policy and GraalVM's contrasting one both offer a template. A project handling safety critical infrastructure with limited reviewer capacity may lean toward OpenJDK's blanket restriction. A project willing to place the accountability burden on the contributor, similar to the Linux kernel's approach, has a working alternative already in production at GraalVM. The wider industry has not settled this question yet, and Oracle's own two projects prove that even one company has not settled it internally. ### Pakistan National AI Initiatives, What Is Actually Live in 2026 URL: https://muhammadhuzaifa.pro/pakistan-national-ai-initiatives-what-is-actually-live-in-2026 Published: 2026-08-06 Pakistan launched a coordinated set of national AI initiatives through 2026, built on the National AI Policy 2025 and centered on seven new AI hubs, a public sector compute buildout, and a pledge to train one million people in AI skills. Some of this is operational today, other parts are still commitments with delivery dates ahead. This piece separates the two and explains what each initiative actually involves. ## The policy foundation, National AI Policy 2025 Everything discussed here traces back to one document. On July 30, 2025, Pakistan's federal cabinet approved the National AI Policy, the country's first formal AI strategy. The policy sets a 2035 vision of Pakistan as a leading AI hub in South Asia and organizes its approach around six pillars, an AI innovation ecosystem, mass awareness and readiness, a secure AI ecosystem, sector transformation, nationwide AI infrastructure, and international collaboration. The policy also carries specific economic projections. Government estimates cited alongside the policy suggest AI adoption could help lift GDP growth from around 7 percent toward 15 percent by 2030, with sector level gains estimated at 12 billion dollars in agriculture, 5 billion dollars in industry, and 26 billion dollars in services. These are policy projections, not measured outcomes, and should be read as the government's stated ambition rather than a verified result. The policy also created the mechanism for a National AI Fund intended to back startups and research, and it calls for AI literacy through curriculum reform. Most of what follows in 2026 is the government building out the implementation layer for this policy. ## The National AI Advancement Initiative and its seven hubs In July 2026, the Ministry of Information Technology and Telecommunication launched the National AI Advancement Initiative, referred to as NAAI, as the first major implementation vehicle for the 2025 policy. NAAI's centerpiece is a network of seven AI hubs distributed across the country rather than concentrated in one city. ### Where the hubs are and what they focus on The seven hubs are located in Islamabad, Lahore, Karachi, Peshawar, Quetta, Muzaffarabad, and Gilgit. Each hub is planned to include modern office infrastructure, high speed internet, collaboration spaces, mentorship facilities, and dedicated workspaces for founders and teams. The stated focus areas are healthcare, agriculture, education, fintech, public sector governance, and smart cities, chosen to reflect practical, sector specific AI applications rather than general purpose research. The government's framing for the geographic spread is to connect startups, academia, industry, investors, and government institutions into what it describes as a unified ecosystem, and to position Pakistan as a regional center for AI development rather than a single city cluster. ### The startup incubation and seed grant model NAAI targets incubating 560 startups over two years. Each hub is designed to induct 20 AI startups per cohort, running two cohorts a year, which works out to 40 startups annually per hub across the seven locations. Alongside incubation, the initiative includes non dilutive seed grants for 150 founders, meaning recipients do not give up equity in exchange for the funding. For early stage founders, this is a materially different offer than typical accelerator terms, and it is worth verifying current application windows directly with MoITT or the relevant hub before assuming eligibility, since program details for fast moving government initiatives can shift between announcement and rollout. ## Public sector AI infrastructure, the compute buildout A national AI strategy needs compute, and Pakistan's public sector took a concrete step on this in July 2026. The National Telecommunication Corporation, NTC, signed a three year agreement with Data Vault Pakistan to provide high performance Graphics Processing Unit as a Service and Model as a Service to government institutions. In practical terms, this is meant to give public sector bodies access to GPU capacity and hosted model infrastructure without each ministry or agency having to build and maintain its own hardware. This is a notable shift because compute access has been one of the more concrete bottlenecks separating AI policy documents from usable government AI systems in many countries. Whether this deal delivers enough capacity, at the right cost and reliability, for meaningful public sector AI workloads is something that will only be verifiable as institutions actually start using it. ## The one million person AI training pledge In June 2026, IT and Telecom Minister Shaza Fatima Khawaja detailed a government plan to train one million people in AI over three years. The plan includes free AI training for students and the recruitment of additional trainers to deliver it. This builds directly on the National AI Policy's mass awareness and readiness pillar and reflects a broader pattern across 2026, several of Pakistan's AI announcements are aimed at building a trained workforce rather than only funding startups or infrastructure. Readers evaluating this pledge should treat the one million figure as a three year target rather than a completed outcome, and should look for interim reporting on actual enrollment and completion numbers as the program matures. ## AI in government operations, the Prime Minister's Office System Separate from the hub and training initiatives, the government has also been applying AI directly to its own operations. On June 17, 2026, Federal Minister for Economic Affairs Ahad Cheema chaired a review of the Prime Minister's Office System, known as PMOS, described as an AI powered platform intended to strengthen governance, improve inter ministerial coordination, and monitor the progress of government initiatives. This is one of the more direct examples of AI being used inside Pakistan's own administrative machinery rather than only being promoted externally to industry and startups. ## Provincial programs, the Khyber Pakhtunkhwa AI Innovation Hub National initiatives are not the only activity. In July 2026, the Khyber Pakhtunkhwa Information Technology Board unveiled its own two year AI Innovation Hub project, aimed at training 1,500 government school teachers and 1,000 government officers. This provincial program runs alongside the federal NAAI hubs and points to a pattern worth watching, provinces are not waiting on federal rollout timelines to launch their own AI capacity building, which could mean faster but more fragmented progress across the country. ## What is proven versus what is still a pledge For a reader trying to act on this information, the distinction between operational and pledged matters more than the announcement date. As of mid 2026, the following are documented as launched or contractually signed, the National AI Policy approval in July 2025, the NAAI hub announcement and its seven locations, the NTC and Data Vault Pakistan compute agreement, and the Khyber Pakhtunkhwa hub launch. The following are stated targets with delivery still ahead, the 560 startup incubation figure and the 150 non dilutive seed grants, the one million person training goal, and the broader GDP and sector level economic projections tied to the 2025 policy. None of these should be treated as completed until independently reported progress data becomes available. ## How developers and founders can engage For a developer or founder deciding whether these programs are relevant, three entry points stand out. First, the NAAI hubs are the clearest startup facing mechanism, and founders working in healthcare, agriculture, education, fintech, governance, or smart city applications are the closest fit for the stated focus areas. Second, the training pledge is the most accessible entry point for individuals rather than companies, since it targets students and non IT professionals directly. Third, the compute agreement between NTC and Data Vault Pakistan is primarily aimed at government institutions rather than private founders, so it is less directly actionable for most developers, but it is a useful signal of where public sector AI capacity is headed. Before applying to or planning around any of these programs, verify current details directly with MoITT, the relevant provincial IT board, or the specific hub, since implementation specifics for government programs frequently get refined after the initial announcement. ## What to watch next The clearest signal of whether 2026's initiatives are working will come from execution data rather than further announcements, actual startup cohort numbers at the seven hubs, verified training enrollment against the one million target, and reporting on how government institutions are actually using the NTC and Data Vault compute agreement. Readers tracking this space should watch for MoITT progress updates and independent reporting rather than relying on launch announcements alone. ### From Islamabad to the Linux Kernel: Bilal's Story URL: https://muhammadhuzaifa.pro/from-islamabad-to-the-linux-kernel-bilals-story Published: 2026-07-15 > **Correction, July 15, 2026.** An earlier version of this post said Bilal may be the first Linux kernel contributor from Pakistan. That was wrong. [Muhammad Usama Anjum](https://www.linkedin.com/in/musamaanjum) has been contributing to the core kernel for years, with more than 80 upstream contributions. My sincere thanks to Maaz Khalid from the Laravelian community for the prompt correction. The post has been updated below. Does Pakistan have a Linux kernel contributors? Yes, Muhammad Usama Anjum has been contributing to the core kernel for years, and now a computer science student at Iqra University Islamabad is also a Linux kernel contributor from Pakistan, one of a small group of Pakistani developers with code merged into the core kernel. ## From Islamabad to the Linux Kernel, Bilal's Story Muhammad Bilal, a BS Computer Science student at Iqra University's H-9 Campus in Islamabad, has done something very few developers anywhere in the world ever manage. He got his code accepted into the official Linux kernel, the software at the heart of Android phones, cloud servers, supercomputers, and billions of devices worldwide. ## A proud moment for Iqra University, H-9 Campus This is exactly the kind of hands-on, real-world impact universities dream of seeing from their students. The Linux kernel is one of the largest and most rigorously reviewed open-source projects in existence. Code doesn't get in because you ask nicely. It gets in because experienced maintainers, some of whom have spent decades on the project, examine every line and decide it makes the kernel better. Bilal's patch went through that gauntlet and came out merged. Huge congratulations from the entire IU H-9 Campus family. ![muhammad-bilal-patches](https://muhammadhuzaifa.pro/storage/photos/1/muhammad-bilal-patches.png) ## Why getting code into the Linux kernel is a big deal The Linux kernel has thousands of contributors from companies like Google, Intel, Arm, and Red Hat. Every proposed change, called a patch, is posted publicly to the kernel mailing lists, where subsystem maintainers review it, question it, and often send it back for revisions, sometimes several times. Only changes that survive this review are merged into the official tree. For a working professional this is a serious achievement. For an undergraduate student doing it from Pakistan, it is remarkable. It takes real technical depth, precision, and above all, persistence. ## Pakistan already has a kernel veteran, meet Muhammad Usama Anjum When I first published this post I believed Bilal might be the first contributor to the core Linux kernel from Pakistan. I was excited, and I was wrong. Maaz Khalid from the Laravelian WhatsApp community reached out with a correction, and I am grateful he did. [Muhammad Usama Anjum](https://www.linkedin.com/in/musamaanjum) is a Linux kernel developer from Pakistan who has been doing this work upstream for years. He has more than 80 contributions in the official kernel tree, working on core kernel areas like memory management and kernel selftests. His best-known work is the PAGEMAP_SCAN ioctl, a memory-tracking interface he authored at Collabora that was merged in Linux 6.7 and covered by [LWN](https://lwn.net/Articles/940704/). He has also spoken at Open Source Summit Europe and the Linux Plumbers Conference. So Bilal is not the first, and honestly that is even better news. Pakistan does not have one kernel contributor. It has a senior one who has been quietly doing it for years, and now a student following the same path. Alhamdulillah. ## A milestone for Pakistan's open-source community Developers from Pakistan have long contributed to major open-source projects. WordPress, Laravel, and countless libraries and tools carry Pakistani names in their commit histories. The core Linux kernel is rarer territory, and Usama's years of upstream work along with Bilal's merged patch show it is firmly within reach. Kernel development is not out of reach for developers and students here ## A personal message to Bilal Bilal, you have written your name in history in a way that will never be forgotten. Your achievement will continue to inspire future generations. This was also one of my own dreams, something I could not accomplish myself, so seeing you achieve it brings me immense happiness. Honestly, I cannot fully express how proud and delighted I feel about what you have done. Thank you. My sincere prayer for you is that you complete your studies with full dedication and continue making Pakistan proud through your work. I also hope that one day we will have the opportunity to learn something from you. ## To every student reading this The Linux kernel's development happens entirely in public. The code, the mailing lists, the review discussions, all of it is open to anyone with curiosity and patience. Bilal didn't need permission, connections, or expensive tools. He needed skill, care, and the persistence to keep revising until his work met the bar. The next contribution from Pakistan doesn't have to wait years. It might be yours. If you want to get started with kernel contributions, begin with the official [kernel newbies guide](https://kernelnewbies.org/) and the kernel's own documentation on submitting patches. **Sources** - [Linux Kernel Mailing List archive, May 2026 thread index](https://lkml.iu.edu/2605.3/index.html) — Bilal's ethosu patch submissions (v1–v3, DMA length validation) - [Patch: accel/ethosu reject NPU_OP_RESIZE commands from userspace](https://lore.gitlab.freedesktop.org/drm-ai-reviews/20260523210840.92039-2-meatuni001@gmail.com/) — technical details of the WARN_ON/DoS fix - [Kernel Newbies](https://kernelnewbies.org/) — linked in the post for readers - [Muhammad Usama Anjum — LinkedIn](https://www.linkedin.com/in/musamaanjum) and [personal site](https://musamaanjum.github.io/) — 80+ upstream kernel contributions, conference talks - [LWN, An ioctl() call to detect memory writes](https://lwn.net/Articles/940704/) — coverage of his PAGEMAP_SCAN work - [Collabora, Kernel 6.7 release notes](https://www.collabora.com/news-and-blog/news-and-events/kernel-67-new-year-new-linux-kernel.html) — PAGEMAP_SCAN merged in 6.7 - Correction credit — Maaz Khalid, Laravelian WhatsApp community, 2026-07-15 ### Your AI Agent Is Not a Coworker URL: https://muhammadhuzaifa.pro/your-ai-agent-is-not-a-coworker Published: 2026-07-01 Calling your AI agent a "coworker" is making your team worse at their jobs. New HBR research confirms it. Boston University's Emma Wiles ran a study where managers reviewed work output. When the work was attributed to an "AI employee" rather than a software tool, participants caught 18% fewer errors. They were 44% more likely to escalate questionable work to a human manager instead of correcting it themselves. The employee label inverted responsibility: if Alex the agent owns the work, the human stops owning quality. The scale of this matters. Thirty-one percent of 1,261 surveyed managers say their companies frame AI agents as employees. Twenty-three percent list agents on official org charts. Every major AI vendor now markets agent management tools as "digital colleagues." Stanford researchers asked 1,500 workers across 104 jobs what AI assistance they actually wanted. The answers were often the opposite of what tech experts assumed. Law clerks wanted help tracking case progress across matters, not replacing legal reasoning. Sales reps specifically did not want AI verifying credit ratings. The people doing the work know where the tool fits. Nobel laureate Daron Acemoglu summarized it: "AI agents are being marketed as things that can replace humans ... that's just a losing proposition." For engineering leads deploying agent infrastructure: frame the agent as a tool, not a teammate. The interface should make accountability inescapable. When your team treats agents as colleagues, they catch fewer errors and offload judgment to software. That is not collaboration. That is abdication. ### Oracle NetSuite Layoffs 30000 People In 2026 URL: https://muhammadhuzaifa.pro/oracle-netsuite-layoffs-30000-people-in-2026 Published: 2026-06-29 Last updated: 2026-11-13 Oracle just laid off 30,000 people -- roughly 18% of its global workforce. The NetSuite division was among the hardest hit. And the timing is brutal: the same quarter Oracle posted record revenue of $19.2B, up 21%, with GAAP net income up 23%. The execution was a 6am email from "Oracle Leadership." No manager call. No warning. System access cut before breakfast. SuiteScript developers, implementation consultants, ACS engineers, ERP project managers -- all gone in a single morning. The "why" is the part that should worry every NetSuite customer: Oracle is trading human payroll for AI data centers. The company needs ~$50 billion to build AI infrastructure to fulfill a $638 billion contract backlog (OpenAI, Meta, xAI). The layoffs free up $8-10 billion in annual cash flow. Legacy support, SaaS maintenance, and NetSuite delivery teams are being cut. OCI and AI teams are actively hiring. For the 40,000+ organizations running on NetSuite, three things to do right now: - Verify your account team is still intact. Your primary Oracle contact may have been affected. - Review your SLA. Support quality is already slipping -- slower responses, "more AI-written replies," less skilled help. - Build a relationship with an independent NetSuite partner. The 5,000+ VARs and boutique SIs are absorbing displaced talent. Having a partner who knows your instance is a hedge against further Oracle-side disruption. For displaced NetSuite talent: the market for certified SuiteScript, SuiteFlow, and SuiteCloud architects was already undersupplied. Salaries range $115K-$321K. This is a one-time talent release into a thirsty ecosystem. The window to land is short -- once placed, these candidates stay for years. The pattern across big tech is the same: cut non-AI headcount, redirect savings to infrastructure. Oracle just did it more brutally and at larger scale than anyone else this year. ## Sources - [Oracle: Q4 and fiscal 2026 results, 10 June 2026](https://www.oracle.com/news/announcement/q4fy26-earnings-release-2026-06-10/). The revenue, net income and $638 billion backlog figures above. Oracle also filed this release with the SEC as an exhibit to its Form 8-K. ### Not Every Prompt Needs GPT-4: Smart Model Routing With a PHP Classifier URL: https://muhammadhuzaifa.pro/not-every-prompt-needs-gpt-4-smart-model-routing-with-a-php-classifier Published: 2026-06-28 ## Not Every Prompt Needs GPT-4: Smart Model Routing With a PHP Classifier If you are building anything that calls an LLM from a Laravel application, you have probably noticed the same thing I did: your bill is dominated by prompts that did not need a frontier model to answer. A user types "what are your business hours?" and it hits GPT-4. A support bot summarizes a log file through Claude Opus. A form validator fires o1 to check if an email looks valid. You are paying premium inference prices for questions a 4o-mini or Haiku could handle in half the time and a tenth of the cost. The problem is not that cheap models exist. They do, and they are good at plenty of things. The problem is that your application has no way to know, before opening a socket to a provider, whether a given prompt is going to be easy or hard. So you either route everything to the expensive model and overpay by default, or you build a routing heuristic that breaks the first time a user phrases something differently. Neither is great, and both get worse the more requests you handle. ### How we fake difficulty today (and why it fails) If you go looking, the workarounds all share the same shape. Some teams route by prompt length, on the theory that longer means harder. In practice, a one-line question about Italian contract law is short and genuinely difficult, while a two-page pasted server log that you want summarized is trivial for any model. Length measures typing, not difficulty. Others maintain a keyword dictionary: anything containing "legal," "code," or "calculate" goes to the premium tier. This works for about a week. Then you are maintaining a dictionary forever. It misses every phrasing you did not anticipate. It has no opinion about prompts in a language you did not hard-code. The most honest attempt is to ask an LLM to rate the difficulty of the prompt before answering it. The problem is that you are now paying for a model call, and waiting for it, in order to decide whether to make a model call. You have added latency and cost to the exact path you were trying to make cheaper. ### A score that comes from your own models The right question is not "is this prompt hard in the abstract?" It is "do the models I actually use find this prompt hard?" Those are different questions, and the second one is the only one that matters when you are deciding which of your providers should answer. The new package **neuron-core/llm-classifier** builds a small classifier that reads an incoming prompt and returns a difficulty score between 0 and 1, where 0 means your models find this easy and 1 means they struggle. The important word there is *your*. The score is learned from the models you actually route between -- GPT-4o-mini, GPT-4o, o1, Claude Haiku, Claude Opus -- so the classifier's idea of "hard" matches your infrastructure's actual behavior. And it runs in pure PHP. The only extension required is `ext-mbstring`. No Python sidecar, no GPU, no inference server. Training happens once, offline. Scoring runs in microseconds, in-process, before you ever open a socket to a provider. ``` composer require neuron-core/llm-classifier ``` ### Two phases, kept strictly apart **Calibration** is the offline phase. You give the calibrator a panel of your models, a corpus of example prompts, and a way to grade whether each model answered correctly. It works out which patterns trip them up and writes a single `model.bin` file. You commit it alongside your code. When your model lineup changes, re-run calibration and replace the binary. **Scoring** is the runtime phase. You load `model.bin` once -- on app boot, or inside Octane, RoadRunner, or FrankenPHP workers -- then call it on each request. Train time and run time never touch. The package ships with a ready-to-use dataset derived from RouterBench: ~36,000 prompts evaluated against 11 common LLMs. A stratified sample of ~1,845 prompts with precomputed difficulty labels is included. Download the fastText vectors, run the calibration script, and you have a working model in under two minutes. ### How the scoring actually works The package uses a free word vector dictionary from fastText -- a 300-dimensional embedding space where "buy" and "purchase" land close together and "king" and "carburetor" land far apart. Each incoming prompt is reduced to one averaged fingerprint of those vectors, and that fingerprint is the classifier's only input. The `overall()` method returns the maximum across per-capability scores, not the average. A prompt that is hard at one thing and trivial at five others should be treated as hard. An average would quietly water that down. ### Plugging it into the Neuron AI router ```php $scorer = Classifier::load('storage/model.bin'); return RouterProvider::make() ->addProvider('mini', new OpenAI(model: 'gpt-4o-mini')) ->addProvider('4o', new OpenAI(model: 'gpt-4o')) ->addProvider('o1', new OpenAI(model: 'o1')) ->setRule( (new DifficultyRule($scorer)) ->outOfDomain('o1', coverage: 0.4) ->easy('mini', maxScore: 0.33) ->medium('4o', maxScore: 0.70) ->hard('o1') ); ``` That is the entire integration. "What are your opening hours?" lands on the cheap model; "draft an NDA under Italian law" lands on the premium one -- automatically. ### Two knobs, and how to turn them Two things to tune, and you tune with data not intuition. **Difficulty cut-offs** (0.33, 0.70) decide where easy ends and hard begins. **Coverage cut-off** (0.4) decides how unfamiliar a prompt has to be before you escalate to the most capable model. Log three things for real traffic: difficulty score, coverage value, and chosen provider. Adjust until the balance is right. Cheap answers coming back wrong? Lower the hard threshold. Out-of-domain prompts leaking to cheap tier? Raise the coverage cut-off. ### Why this matters for PHP/Laravel shops Most routing tools assume a Python service somewhere in your stack -- another process to monitor, another thing that can fail at 3 AM. For a Laravel team with a clean PHP pipeline and small ops surface, that is a hard sell. The llm-classifier removes that friction entirely. It is a Composer package. It runs in the same process. Works under Octane, RoadRunner, FrankenPHP with no special config. No new infrastructure, no new runtime dependencies. If 60% of your LLM traffic is cheap-model material and you were routing everything to GPT-4o, the classifier pays for itself in days, not months. ### The takeaway For a long time the answer to "which model should answer this?" in PHP was either static or a pile of keyword matching you maintained by hand. Now there is a measured answer that costs microseconds, comes from your own models, and drops into a router that was already part of the framework. MIT licensed: [github.com/neuron-core/llm-classifier](https://github.com/neuron-core/llm-classifier) --- ### Your Hiring Pipeline Is Measuring Ghosts URL: https://muhammadhuzaifa.pro/your-hiring-pipeline-is-measuring-ghosts Published: 2026-06-24 There is a new problem showing up in engineering hiring pipelines, and it is one we built ourselves. Tom MacWright calls it "accidental anonymity." Candidates submit LLM-written resumes, link to LLM-generated portfolio sites, which point to LLM-generated GitHub projects with LLM-generated commit messages. Every artifact is polished, coherent, and completely impersonal. You learn nothing about the person behind them. The resume hits every keyword. The portfolio looks professional. The GitHub has green squares. But there is no signal -- no evidence of how this person thinks through a tradeoff, debugs something unfamiliar, or makes an architectural call when the path is unclear. The LLM filled every gap with plausible competence. Now every candidate looks the same. This is not a candidate quality problem. It is a signal problem. For years we trained hiring pipelines to scan for polished artifacts -- resumes with the right keywords, portfolios with clean projects, contribution graphs. Those were always proxies. The LLM just exposed how weak they were. The fix is not AI detection. It is designing interviews that measure what LLMs cannot fake. A live debugging session on an unfamiliar codebase. A 30-minute architecture discussion where you change the constraints three times. A judgment call on which part of a system to optimize and which to leave alone. If your top-of-funnel screening can be passed by a candidate who never touches a keyboard, your hiring pipeline is measuring the wrong thing. ### Your Model Vendor Is a Geopolitical Risk URL: https://muhammadhuzaifa.pro/your-model-vendor-is-a-geopolitical-risk Published: 2026-06-24 The US government just shut down Anthropic's Fable 5 and Mythos models via export controls, three days after Fable 5's public release. Amazon's CEO reportedly triggered the crackdown. Anthropic cut access entirely to comply. The fallout is reshaping the model supply chain. French politicians are calling it a wake-up call for European AI sovereignty. But the real shift is toward Chinese models. Zhipu's GLM-5.2 just debuted at #4 on the global LLM leaderboard at $1.73 per million tokens -- cheaper than nearly every Western frontier model, with no export controls and no access revocations. For engineering leads, the lesson is clear: model supply chain risk is not theoretical. If your production AI features are hard-coupled to a single US provider subject to export controls, your architecture has a single point of failure that is geopolitical, not technical. Multi-provider routing and open-weight fallbacks are now infrastructure resilience, not cost optimization. The deeper irony: cybersecurity experts published an open letter arguing the shutdown makes the US less safe. Restricting access to Anthropic's models while Chinese alternatives remain freely available does not reduce attack surface. It shifts market share -- and the defenders lose access while adversaries do not. This is an architecture decision with a clock on it. ### Agent Evaluation: Why Tool Traces and Verification Matter URL: https://muhammadhuzaifa.pro/agent-evaluation-why-tool-traces-and-verification-matter Published: 2026-06-22 Long-context agent evaluation is a different problem from judging one clean final answer. The useful framing from Agent Judge is that production agents should be evaluated on the work they did across the whole trajectory, not only the last response. A normal LLM judge can miss the parts that determine whether an agent is reliable: did it search the right places, did it verify state after taking an action, and did it adjust when feedback changed the rubric? A few takeaways I would keep in mind when building or reviewing agentic systems: 1. Search quality matters. An agent that answers confidently without looking in the right systems is not production-ready. 2. Verification should be explicit. If the agent changes state in a real system, the eval should check whether it confirmed the result. 3. Rubrics need to evolve from feedback. Static evals help, but harder scenarios need criteria that reflect real failure modes. For production workflows, evidence should be part of the evaluation contract. That feels especially relevant for teams moving from demos to agents that touch real customer data, tickets, inboxes, billing, or deployments. What do you currently trust more for agent evals: final answer quality, tool trace review, or outcome verification? ### Observability for AI Agents: Why Your Traces Need to Become an API URL: https://muhammadhuzaifa.pro/observability-for-ai-agents-why-your-traces-need-to-become-an-api Published: 2026-06-20 The more agentic systems we build, the more I think observability needs a small shift in mindset. For years, metrics, logs, and traces were designed around a human workflow: something breaks, an engineer opens a dashboard, scans the logs, jumps through traces, and builds a mental model of what happened. That model still matters. But LLMs and agents are starting to consume the same operational data directly. In that world, traces may become less of a debugging artifact and more of an interface. A few takeaways I liked from this angle: 1. Traces are probably the most useful entry point for agents because they preserve flow, timing, causality, and context. 2. Trace schemas should be treated like APIs. If agents depend on them, they need stable names, clear meanings, versioning, and documentation. 3. Human-readable observability is not enough. Machine-readable observability needs predictable structure, not clever labels and inconsistent conventions. 4. This applies beyond infra teams. Laravel apps, queues, jobs, webhooks, payments, and third-party integrations all benefit when traces explain intent, state, and failure context clearly. If an agent is going to help debug production, the trace is part of the contract. ### judging at nasa space apps challenge in hyderabad URL: https://muhammadhuzaifa.pro/judging-at-nasa-space-apps-challenge-in-hyderabad Published: 2025-10-10 # Judging at NASA Space Apps Challenge in Hyderabad Hyderabad hosted its first NASA Space Apps Challenge, and I was formally invited as a local judge to evaluate projects from an engineering perspective. Over two intense days, student teams built prototypes, explored Machine Learning algorithms, and learned how to integrate LLMs' APIs. I was joined by respected academics, and we saw thoughtful work from universities as well as two standout college teams. I guided one AI-heavy team on balancing prompt use with strong computer science fundamentals, at the same time I also applauded their efforts. The event was organized, managed and run by an undergraduate core team from [Mehran University of Engineering and Technology](https://sw.muet.edu.pk/index.php), and [Hyderabad Institute for Technology and Management Sciences](https://hitms.edu.pk/), hosted by [Iqra University Hyderabad Campus](https://iqra.edu.pk/campus/hyderabad-campus/), and supported by different sponsors. ## College Teams Two memorable teams came from colleges rather than universities, which is rare in a second tier city here. Students from two different colleges studied hard and finished their challenges. One team leaned heavily on AI and solved most tasks through prompt engineering while not knowing what prompt engineering is. When I asked them to explain their solution for the given challenge, they struggled, so I guided them on why strong computer science basics should come first and how AI can be used responsibly. The room also had undergrads from several universities and a few early stage AI founders. The overall spirit was practical and focused. ## What is [NASA International Space Apps Challenge](https://spaceappschallenge.org) NASA International Space Apps Challenge is a worldwide hackathon where people use open data from NASA and partner agencies to make ideas that help life on Earth and in space. Local events happen in many cities and connect to a global community that learns, builds, and shares together. The challenge themes usually cover Earth observation, climate and weather, remote sensing, data visualization, and education tools. The learning outcomes are valuable for any student or early professional: teamwork, data literacy, reading documentation, rapid prototyping, and clear storytelling. People also learn to clean and interpret datasets and explain technical choices in simple language. ## Project highlights and tech stacks We had more than 100 participants across around 20 teams. All of the solutions were web based, frontends used HTML, CSS, JavaScript, and React, and TailwindCSS. backends were in Node.js or Python. Some teams tried Machine Learning and Deep Learning algorithms. A few integrated large language models through API integrations. The range of solutions showed careful research, fast prototyping, and sensible trade offs for a weekend sprint. ## My judging experience and the panel I was invited to judge the engineering side of projects. I enjoyed the company of [Dr Amjad Ali Syed](https://www.linkedin.com/in/dr-amjad-ali-syed-3482b032/), Associate Professor at [MUET](https://site.muet.edu.pk/departments/biomedical-engineering), along with [Dr Mehwish Jawaid](https://www.linkedin.com/in/mehwish-jaweed-0664bb206/) from [SZABIST](https://hyd.szabist.edu.pk/), and [Noorulain Mushtaq](https://www.linkedin.com/in/noorulain-mushtaq-004036185/), also from MUET with a background in Data Science. It was a pleasure to learn from their academic perspective while sharing my industry view. I felt honored to be among such talented and experienced people. ## Shoutout to the organizing team The event was planned by an undergraduate core team from [Mehran University of Engineering and Technology](https://sw.muet.edu.pk/index.php), and [Hyderabad Institute for Technology and Management Sciences](https://hitms.edu.pk/). They handled real pressure and still delivered a smooth experience for participants and judges. Special thanks to: * [Haziq ul Khair](https://www.linkedin.com/in/muhammad-haziqul-khair/) (Local Lead - MUET), * [Abdullah Ali](https://www.linkedin.com/in/abdullah-ali-2k4/) (Co Lead External Affairs - MUET), * Esha Salam (Co Lead External Affairs - MUET), * [Ibad Hussain](https://www.linkedin.com/in/ibad-hussain-5b9608356/) (Co Lead Internal Affairs- HITMS), * [Bushra Laraib](https://www.linkedin.com/in/bushra-laraib-2795982aa/) (Co Lead Internal Affairs - MUET), and * [Muhammad Faisal](https://www.linkedin.com/in/muhammad-faisal-70482427a/) (Operations Lead - MUET). Thank you for the invitation and for trusting me as a judge. ## Appreciation for sponsors and hosts I also appreciate the support from [Iqra University Hyderabad Campus](https://iqra.edu.pk/campus/hyderabad-campus/) for the venue and on ground help and all the sponsors. Your support made this first edition possible. ## A note to the local software industry One thing made me a bit sad. I did not see sponsors or attendees from the local software industry. These events are ideal places for academia and industry to meet, mentor, and build a talent pipeline. After speaking with the core team, I learned they did not formally reach out to many companies, which is understandable for a first time effort. For the next edition, I suggest early, formal invitations to leading Hyderabad tech firms. I also invite companies to show up, sponsor, and mentor. The impact will multiply when both sides participate. Hyderabad’s first NASA Space Apps showed what students can do with open data, mentorship, and a shared mission. I plan to support future editions, help connect local companies, and keep mentoring teams so our city grows its presence on the global Space Apps map. If you are a student, join next year. If you are in industry or academia, get involved. Let us build this together. ### pakistan software export board (pseb): a practical guide for pakistan’s tech community URL: https://muhammadhuzaifa.pro/pakistan-software-export-board-pseb-a-practical-guide-for-pakistans-tech-community Published: 2025-09-04 # Pakistan Software Export Board (PSEB): A Practical Guide For Pakistan’s Tech Community ## The problem we face If you build software in Pakistan, you quickly find the same roadblocks. Getting international clients is hard, credibility takes time, taxes and compliance feel confusing, and the cost of good office space and reliable connectivity can drain your budget. Many founders, freelancers, and even mid-size firms have the talent, yet struggle with visibility, certifications, and access to global events and buyers. ## The solution PSEB offers The Pakistan Software Export Board exists to reduce these barriers. In simple words, PSEB helps you look bigger than you are by backing you with government recognition, training, export facilitation, and introductions to international markets. It supports the full life cycle of an IT business in Pakistan, from the first step of registration to showcasing your work on global stages. ## How PSEB came into being and how it is governed PSEB was set up in the mid-1990s when the government saw the need for a dedicated body to grow software and IT-enabled services. It is a government-owned company that works under the Ministry of Information Technology and Telecommunication (MoITT). PSEB is guided by a board that brings the public and private sectors together, which keeps programs practical and aligned with industry needs. The goal is straightforward. Help Pakistani IT talent compete globally, increase exports, and create skilled jobs at home. ## What PSEB actually does Think of PSEB as a one-stop partner for the IT industry in Pakistan. * Builds and manages Software Technology Parks with IT-enabled office space * Runs training and internship programs to improve skills and create a hiring pipeline * Helps companies get quality certifications and compliance support * Markets Pakistani IT companies abroad and supports trade show participation * Maintains directories and does buyer matchmaking to connect you with clients * Provides facilitation on policies, registrations, and regulatory matters * Supports call centers and other IT-enabled services with required compliances ## Who should join * Software houses and product companies * IT service providers and consultancies * Startups and scaleups in software or IT-enabled services * Freelancers and small teams exporting IT services * Call centers and BPOs If you build, sell, or export IT products or services from Pakistan, membership is relevant. ## How to become a PSEB member You can apply online. The process is simple and usually quick if your documents are ready. ### Step-by-step 1. **Confirm eligibility** You are an IT or IT-enabled services company, startup, or freelancer operating in Pakistan. 2. **Prepare documents** * For companies: NTN, incorporation certificate, basic company profile, directors’ CNICs, business bank account details * For freelancers: CNIC, bank account details, basic service profile or portfolio 3. **Create an account on the PSEB portal** Choose the correct category, fill in your details, and verify your email. 4. **Complete the application** Provide accurate company or freelancer information, upload documents, and review. 5. **Pay the applicable fee** The portal explains payment methods. Keep the receipt or upload the proof as required. 6. **Approval and certificate** After verification, you receive your PSEB registration certificate. You can then access member benefits. ### Common tips * Keep your company description clear and simple * Make sure names and numbers match your official documents * If you export, maintain basic records of export proceeds for future benefits ## What you get in return Below is a quick snapshot. Details and availability can vary with programs, but this covers the core value. | Benefit | What it means in practice | Who gains most | | -------------------------- | ------------------------------------------------------------------------------ | ----------------------------------------- | | Government recognition | A formal certificate that builds trust with clients and partners | Startups, SMEs, freelancers | | Marketing and visibility | Listing in directories, support for global events and delegations | Firms seeking international clients | | Matchmaking and leads | Introductions to buyers, investors, and partners | Product and service companies | | Training and internships | Subsidized upskilling and a pipeline of interns who often convert to full-time | Growing teams that need talent | | Certifications support | Help with quality standards and audits to win enterprise clients | Companies selling to regulated industries | | Tech parks and facilities | IT-enabled office space with reliable connectivity and services | Teams that want to scale professionally | | Policy and compliance help | Guidance on registrations, exports, and special cases like call centers | New and expanding businesses | In my view, this combination of credibility, exposure, and enablement is the key reason membership pays back quickly for serious IT businesses. ## Real examples and use cases * **Startup to first export client** A three-person web agency registered with PSEB, used the member listing to showcase credibility, and secured its first export contract after being discovered by a Middle East buyer who was specifically searching for Pakistani partners. * **Product company entering trade shows** A SaaS startup qualified for subsidized participation in a regional tech expo. That single event produced pilot deals with two distributors and a reseller agreement, which would have been costly to achieve alone. * **Hiring pipeline through internships** A mid-size services firm hosted PSEB-supported interns in data engineering and QA. Three interns converted to full-time roles after six months, lowering hiring risk and onboarding time. * **Quality certifications to unlock enterprise sales** A software house needed formal quality credentials to bid for a telecom project. With PSEB’s support and guidance, the team completed the certification path and closed the deal. These are the typical outcomes I see. The pattern is consistent. Recognition brings trust, trust opens doors, and doors bring revenue. ## Frequently asked questions **Is PSEB only for big companies** No. Freelancers, small teams, and early startups can all register. Many benefits are designed for early-stage growth. **Do I need to be export-ready before joining** Joining early helps. The recognition and training can prepare you for exports and make your first deal easier. **What if I work remotely for foreign clients as a solo freelancer** Registering still helps with credibility, simple compliance, and smoother banking for export proceeds. **Is membership useful if I already have clients** Yes. International marketing, certification support, and matchmaking can help you move up the value chain. ## Final thoughts Pakistan has the talent. What we often need is structure, trust, and global reach. PSEB gives you that support. If you are building a serious IT career or business in Pakistan, I strongly recommend registering. It is a practical way to reduce friction, improve credibility, and grow faster with support that is aligned with how our industry actually works. ### dijkstra vs the new deterministic sssp algorithm: easy guide, php examples, and visuals URL: https://muhammadhuzaifa.pro/dijkstra-vs-the-new-deterministic-sssp-algorithm-easy-guide-php-examples-and-visuals Published: 2025-08-18 ## TL;DR * Dijkstra’s algorithm is the classic way to compute shortest paths with non-negative weights, running in about O(m + n log n). * A 2025 result gives a **deterministic** SSSP for **directed** graphs that runs in **O(m · (log n)^(2/3))** in the comparison-addition model. It beats the long-standing n log n “sorting barrier” on sparse graphs. * Big idea in plain words: instead of strictly sorting by distance at every step like Dijkstra, the new method processes nodes in smart **batches** using bounded multi-source relaxations and a reduced **frontier** of pivots. --- # The problem in one minute Single-Source Shortest Path (SSSP) asks: given a source node, what is the minimum distance to every other node in a weighted graph. Classic answer: **Dijkstra’s algorithm** for non-negative weights. It repeatedly picks the closest unfinished node, then relaxes its outgoing edges. That “pick the closest” step creates a sorting-like cost of about log n per operation, which is why the total time has the n log n behavior. Researchers showed a new **deterministic** way that reduces the dependency on sorting by exploring nodes in **layers** and shrinking the active **frontier** so heavy work applies only to a small subset. This changes the asymptotic behavior to **O(m · (log n)^(2/3))**. --- # Visuals to anchor the intuition ### 1) Growth trend: n·log n vs n·(log n)^(2/3) ![Asymptotic growth comparison](https://muhammadhuzaifa.pro/storage/photos/1/sssp_growth.webp) What you see: both grow with n, but n·(log n)^(2/3) grows a bit slower than n·log n. For huge graphs, that gap matters. ### 2) What changes operationally Left: Dijkstra picks exactly the closest next node each step. Right: the new mindset processes a **batch frontier** and runs a few **k-hop relaxations**, then moves the frontier forward. ![Frontier schematic](https://muhammadhuzaifa.pro/storage/photos/1/frontier_schematic.webp) ### 3) Tiny graph we will use in code ![Sample directed graph](https://muhammadhuzaifa.pro/storage/photos/1/sample_graph.webp) Edges: A→B(2), A→C(2), B→C(1), B→D(3), C→D(2), C→E(3), D→E(1) Expected distances from A: A 0, B 2, C 2, D 4, E 5 --- # Practical example 1: Dijkstra in PHP This is a clean, didactic version. It uses a simple linear scan for the minimum to keep it readable. For production, use a binary heap or Fibonacci heap. ```php // Dijkstra's Algorithm — simple, highly commented version $graph = [ 'A' => ['B' => 2, 'C' => 2], 'B' => ['C' => 1, 'D' => 3], 'C' => ['D' => 2, 'E' => 3], 'D' => ['E' => 1], 'E' => [] ]; $source = 'A'; $INF = 1_000_000; // 1) Distance init $dist = []; foreach ($graph as $node => $_) $dist[$node] = $INF; $dist[$source] = 0; // 2) Finalized set $visited = []; // 3) Repeatedly extract the closest unvisited node and relax edges while (count($visited) < count($graph)) { // extract-min $minNode = null; $minDist = $INF; foreach ($dist as $node => $d) { if (!isset($visited[$node]) && $d < $minDist) { $minDist = $d; $minNode = $node; } } if ($minNode === null) break; // no more reachable nodes // finalize $visited[$minNode] = true; // relax neighbors foreach ($graph[$minNode] as $nbr => $w) { if (!isset($visited[$nbr])) { $alt = $dist[$minNode] + $w; if ($alt < $dist[$nbr]) $dist[$nbr] = $alt; } } } // 4) Print echo "Dijkstra distances from $source:\n"; foreach ($dist as $node => $d) echo "$node: $d\n"; ``` **What to notice** * The algorithm always orders the frontier to pick the absolute smallest. * That global ordering is where the log n overhead comes from. --- # Practical example 2: Batch-style deterministic idea in PHP The full research algorithm is involved. The script below captures the **core flavor**: limited-depth relaxations from a **frontier** as a batch, then move the frontier forward. It avoids a global extract-min. It is meant for learning and mirrors the high-level design decisions. ```php // Batch-style multi-source relaxations — educational approximation $graph = [ 'A' => ['B' => 2, 'C' => 2], 'B' => ['C' => 1, 'D' => 3], 'C' => ['D' => 2, 'E' => 3], 'D' => ['E' => 1], 'E' => [] ]; $source = 'A'; $INF = 1_000_000; // Distances $dist = []; foreach ($graph as $node => $_) $dist[$node] = $INF; $dist[$source] = 0; // Limit per batch — think "k-hop" relaxations $k = 2; // Start from the source frontier $frontier = [$source]; $changed = true; while ($changed) { $changed = false; // If frontier is empty but improvements may exist, restart from all nodes if (empty($frontier)) $frontier = array_keys($graph); // Track nodes improved in this batch $touched = []; // Up to k layers from current frontier $wave = $frontier; for ($round = 1; $round <= $k; $round++) { $nextWave = []; foreach ($wave as $u) { foreach ($graph[$u] as $v => $w) { $alt = $dist[$u] + $w; if ($alt < $dist[$v]) { $dist[$v] = $alt; $touched[$v] = true; $nextWave[] = $v; $changed = true; } } } if (empty($nextWave)) break; $wave = $nextWave; } // New frontier only where progress happened $nextFrontier = []; foreach (array_keys($touched) as $u) { foreach ($graph[$u] as $v => $_w) $nextFrontier[$v] = true; } $frontier = array_keys($nextFrontier); } // Print echo "Batch-style distances from $source:\n"; foreach ($dist as $node => $d) echo "$node: $d\n"; ``` **What to notice** * No priority queue. * Work happens in short waves from a reduced frontier, not by a global strict order every step. * This matches the paper’s idea of bounded multi-source relaxations and frontier reduction, without reproducing all the complexity. --- # Why the new idea helps Dijkstra pays a log n cost in selecting the next node because it maintains a full order of candidate distances. The new method avoids that by: 1. grouping work into batches, 2. limiting relaxations to k layers per batch, 3. shrinking the frontier to a small set of pivots before recursing. This breaks the sorting bottleneck and yields **O(m · (log n)^(2/3))** time. --- # Complexity comparison | Algorithm | Works on | Idea in one line | Time complexity | | ---------------------- | ------------------------------ | ------------------------------------------------------------------- | -------------------------------------------- | | Dijkstra | Non-negative weights | Always extract the closest unfinished node, then relax | O(m + n log n) in comparison-addition model | | New deterministic SSSP | Directed, non-negative weights | Batch relaxations from a reduced frontier with recursion and pivots | O(m · (log n)^(2/3)) | **Rule of thumb** * Small or medium graphs: Dijkstra is simple and often fast enough. * Very large directed graphs: the new approach has better asymptotic behavior and is theoretically faster. --- # FAQ **Does the new algorithm replace Dijkstra in everyday code** Not necessarily. Dijkstra is simple and widely available. The new method is deeper and shines at scale. **Can the new idea handle negative weights** The result is for non-negative weights in the comparison-addition model. Negative weights are a different problem with separate lines of research. --- # Final takeaways * Dijkstra remains a practical baseline for non-negative weights. * The 2025 deterministic SSSP shows we can beat the classic n log n barrier by changing how we explore the frontier. It uses bounded multi-source relaxations, a shrunken frontier, and recursion to avoid repeated global sorting. --- If you need an engineer who can translate advanced algorithmic ideas into production-ready systems with clean PHP and JavaScript, integrate complex SaaS APIs like HubSpot or Salesforce, design scalable architectures, and lead teams with DevOps discipline, I would love to help. I have 13+ years of experience across PHP, Laravel, Node.js, MySQL, PostgreSQL, and modern frontend stacks, plus deep integration work with NetSuite, SAP, and Shopify. I build maintainable code, robust pipelines, and clear documentation. **Let us discuss your project**. You can reach me through [my Upwork profile](https://www.upwork.com/freelancers/~01dcc2f82551b47dee). --- ### hacked the laravel artisan route command to enforce route naming conventions URL: https://muhammadhuzaifa.pro/hacked-the-laravel-artisan-route-command-to-enforce-route-naming-conventions Published: 2025-08-06 Have you ever tried using a piece of functionality from the Laravel framework in a way it’s not supposed to be used? Well, I did. Actually, I tried, and I succeeded. Not long ago, I wrote an article on [architectural testing](https://muhammadhuzaifa.pro/enforcing-clean-controller-design-in-laravel-with-pestphp-architectural-testing) using [PestPHP](https://pestphp.com) and its benefits within the Laravel ecosystem. That article was mostly focused on enforcing clean controller design using architectural testing. This time, I had a different idea. I wanted to validate that every **named route** must follow the **resourceful naming convention**. --- ### What is a Laravel Resource Controller? A **Laravel resource controller** is a convenient way to group typical "CRUD" (Create, Read, Update, Delete) routes behind a single controller class. Laravel automatically assigns standard routes for actions like listing records, showing a single record, storing new entries, updating, and deleting. You don’t need to define each one manually. According to the [Laravel documentation](https://laravel.com/docs/12.x/controllers#resource-controllers), a resource controller maps the following actions: | Verb | URI | Action | Route Name | |------------|---------------------------|---------|------------------| | GET | /photos | index | photos.index | | GET | /photos/create | create | photos.create | | POST | /photos | store | photos.store | | GET | /photos/{photo} | show | photos.show | | GET | /photos/{photo}/edit | edit | photos.edit | | PUT/PATCH | /photos/{photo} | update | photos.update | | DELETE | /photos/{photo} | destroy | photos.destroy | As you can see, the route names end in either `index`, `create`, `store`, `show`, `edit`, `update`, or `destroy`. I wanted to ensure this convention was followed throughout the project. So, I started writing a custom expectation. At first, I used `Illuminate\Support\Facades\Route`, which provides the static method `getRoutes`. ```php // vendor/laravel/framework/src/Illuminate/Support/Facades/Route.php /** * @method static \Illuminate\Routing\RouteCollectionInterface getRoutes() **/ ``` This method returns an instance that implements `RouteCollectionInterface`, which in turn offers another `getRoutes` method: ```php // vendor/laravel/framework/src/Illuminate/Routing/RouteCollectionInterface.php /** * Get all of the routes in the collection. * * @return \Illuminate\Routing\Route[] */ public function getRoutes(); ``` That method returns an array of `\Illuminate\Routing\Route[]`. Once I had that, the rest was simple: traverse the routes, check their names, and if any didn't follow the convention, fail the expectation. It worked like a charm and gave the expected output. I joined all the invalid route names into a string and displayed it to the output stream. However, I wasn't happy, or you could say, I wasn't fully satisfied. I thought: would it be possible to fetch the routes through the Artisan `route:list` command instead? The Artisan command already comes with built-in functionality, which would make my work easier and cleaner. Plus, unlike my initial approach, I wouldn't need to add custom logic for excluding vendor-specific routes or formatting the display output. --- ### Exploring the Artisan Route Command Internals So I started digging into the internals of the `route:list` command and discovered some functions that could help me with my goal: ```php // vendor/laravel/framework/src/Illuminate/Foundation/Console/RouteListCommand.php /** * Determine if the route has been defined outside of the application. * * @param \Illuminate\Routing\Route $route * @return bool */ protected function isVendorRoute(Route $route) { if ($route->action['uses'] instanceof Closure) { $path = (new ReflectionFunction($route->action['uses'])) ->getFileName(); } elseif (is_string($route->action['uses']) && str_contains($route->action['uses'], 'SerializableClosure')) { return false; } elseif (is_string($route->action['uses'])) { if ($this->isFrameworkController($route)) { return false; } $path = (new ReflectionClass($route->getControllerClass())) ->getFileName(); } else { return false; } return str_starts_with($path, base_path('vendor')); } /** * Convert the given routes to regular CLI output. * * @param \Illuminate\Support\Collection $routes * @return array */ protected function forCli($routes) { $routes = $routes->map( fn ($route) => array_merge($route, [ 'action' => $this->formatActionForCli($route), 'method' => $route['method'] == 'GET|HEAD|POST|PUT|PATCH|DELETE|OPTIONS' ? 'ANY' : $route['method'], 'uri' => $route['domain'] ? ($route['domain'].'/'.ltrim($route['uri'], '/')) : $route['uri'], ]), ); $maxMethod = mb_strlen($routes->max('method')); $terminalWidth = $this->getTerminalWidth(); $routeCount = $this->determineRouteCountOutput($routes, $terminalWidth); return $routes->map(function ($route) use ($maxMethod, $terminalWidth) { [ 'action' => $action, 'domain' => $domain, 'method' => $method, 'middleware' => $middleware, 'uri' => $uri, ] = $route; $middleware = (new Stringable($middleware))->explode("\n")->filter()->whenNotEmpty( fn ($collection) => $collection->map( fn ($middleware) => sprintf(' %s⇂ %s', str_repeat(' ', $maxMethod), $middleware) ) )->implode("\n"); $spaces = str_repeat(' ', max($maxMethod + 6 - mb_strlen($method), 0)); $dots = str_repeat('.', max( $terminalWidth - mb_strlen($method.$spaces.$uri.$action) - 6 - ($action ? 1 : 0), 0 )); $dots = empty($dots) ? $dots : " $dots"; if ($action && ! $this->output->isVerbose() && mb_strlen($method.$spaces.$uri.$action.$dots) > ($terminalWidth - 6)) { $action = substr($action, 0, $terminalWidth - 7 - mb_strlen($method.$spaces.$uri.$dots)).'…'; } $method = (new Stringable($method))->explode('|')->map( fn ($method) => sprintf('%s', $this->verbColors[$method] ?? 'default', $method), )->implode('|'); return [sprintf( ' %s %s%s%s %s', $method, $spaces, preg_replace('#({[^}]+})#', '$1', $uri), $dots, str_replace(' ', ' › ', $action ?? ''), ), $this->output->isVerbose() && ! empty($middleware) ? "$middleware" : null]; }) ->flatten() ->filter() ->prepend('') ->push('')->push($routeCount)->push('') ->toArray(); } ``` * `isVendorRoute()` excludes vendor-specific routes. * `forCLI()` gives a nicely formatted output, ready for the terminal. Perfect, right? Almost. The problem was, both of these functions are marked as `protected`. --- ### What is `protected` in PHP? The `protected` access modifier in PHP means that the property or method can only be accessed within the class itself or in child classes (subclasses). It’s not accessible from outside the class directly. Here’s a quick example: ```php class Example { protected function greet() { return "Hello from inside!"; } } $ex = new Example(); // This will throw an error echo $ex->greet(); // Fatal error: Uncaught Error ``` As you can see, calling the `greet()` method directly from outside the class won't work. --- As you know, I can’t access anything marked as `protected` directly, and I can’t override the core Artisan command either. But luckily, PHP offers something powerful called `ReflectionClass`. --- ### What is `ReflectionClass` in PHP? `ReflectionClass` is a part of PHP’s Reflection API, which lets you inspect classes, methods, properties, and even invoke methods that are otherwise inaccessible (like protected or private ones). It’s commonly used in testing, debugging, and advanced use cases like this one. Here’s a basic example: ```php class Secret { protected function reveal() { return "Top secret!"; } } $refClass = new ReflectionClass(Secret::class); $method = $refClass->getMethod('reveal'); $method->setAccessible(true); $instance = new Secret(); echo $method->invoke($instance); // Outputs: Top secret! ``` Using this, I was able to tap into the protected methods of the route command and do exactly what I needed, without rewriting or duplicating logic. So by now, you probably get the idea that came to my mind. I used `ReflectionClass`, changed the access modifiers for both of the functions, and used them. And guess what? It worked exactly as I hoped. --- Here’s the final expectation: ```php expect()->extend('toFollowLaravelNamingConvention', function () { /** * @var RouteListCommand $routeListCommand */ $routeListCommandClass = app()->make(RouteListCommand::class); $reflection = new ReflectionClass($routeListCommandClass); $getRoutes = $reflection->getMethod('getRoutes'); $getRoutes->setAccessible(true); $isVendorRoute = $reflection->getMethod('isVendorRoute'); $isVendorRoute->setAccessible(true); $forCLI = $reflection->getMethod('forCLI'); $forCLI->setAccessible(true); $reflection->getProperty('laravel')->setValue($routeListCommandClass, app()); $output = new ConsoleOutput(OutputInterface::VERBOSITY_NORMAL, true); $reflection->getProperty('output')->setValue($routeListCommandClass, $output); $routes = collect(Route::getRoutes()->getRoutes()) // Exclude nameless routes ->filter(fn($route) => $route->getName()) // Exclude vendor specific routes ->filter( fn($route) => !$isVendorRoute ->invoke( $routeListCommandClass, $route ) ) // ensure name ends with allowed suffixes ->filter( function ($route) { $parts = explode('.', $route->getName()); $suffix = end($parts); return !in_array( $suffix, [ 'index', 'create', 'store', 'show', 'edit', 'update', 'destroy', ] ); } ) // transform route object to RouteListCommand compatible array ->map(fn($route) => [ 'domain' => $route->domain(), 'method' => implode('|', $route->methods()), 'uri' => $route->uri(), 'name' => $route->getName(), 'action' => ltrim($route->getActionName(), '\\'), 'middleware' => "", 'vendor' => false, ]); /** * If routes array is not empty, there're some routes not following convention, * fail expectation with developer friendly message. */ if (!$routes->isEmpty()) { test()->fail(implode("\n", array_merge( ["Following routes are not following Laravel resourceful naming conventions"], $forCLI->invoke($routeListCommandClass, $routes) ))); } return true; }); ``` And here’s the assertion output from the expectation. Isn't it marvelous? ![image](https://muhammadhuzaifa.pro/storage/photos/1/hacked-the-laravel-artisan-route-command-to-enforce-route-naming-conventions-assertion-output.webp) --- This approach not only saved me from writing redundant logic but also allowed me to take full advantage of Laravel's built-in features. Sure, it's a bit of a hacky trick, but sometimes bending the rules a little gives you cleaner, more maintainable code. If you found this approach insightful or learned something new about bending Laravel to your will, I’d love to hear your thoughts. This is just one example of how curiosity and a bit of unconventional thinking can help you get the most out of your tools without bloating your codebase. I'm always experimenting with better ways to write clean, maintainable software that fits real-world needs. If you're looking for someone who can bring that kind of mindset to your projects, feel free to check out my [Upwork profile](https://www.upwork.com/freelancers/~01dcc2f82551b47dee) and let’s work together. ### laravel utm tracking: built-in simplicity beats third-party complexity URL: https://muhammadhuzaifa.pro/laravel-utm-tracking-built-in-simplicity-beats-third-party-complexity Published: 2025-08-02 In the open-source world, it's not just a common practice anymore. It's more like a built-in habit. Whenever we need to implement some functionality, we first look around to see if there’s already a tool available. If there is, we review it, and if it meets our needs, we go ahead and use it. This is how things work — whether it’s WordPress with plugins, Drupal with modules, or Laravel with packages. But every now and then, there are tasks that are small and easy to handle on your own. They also give you a chance to stay hands-on. A good example is something I recently did on my personal website: [https://muhammadhuzaifa.pro](https://muhammadhuzaifa.pro). I wanted to forward UTM query string parameters. When a visitor lands on the site with UTM tags, I wanted to track which links they visit during their session, maybe even do more with that data later. Like always, I first looked for an existing package. I found one by Spatie that seemed related. But then I thought, this is such a small task. Why bring in a whole package for it? The package wasn't under active development and didn’t seem fully stable either. That means I’d need to test it thoroughly before using it. Since it was just for my own site, not for a client, I decided it wasn’t worth the effort. I didn’t want to overcomplicate a simple need. I had a few options in mind: 1. I could pass the UTM parameters manually by using the second argument of Laravel’s `route()` helper and fetch them from the request. But doing that for every route? Not practical. 2. I could `bind` a `custom router class` in a `service provider` and override the `route` generation using a `macro`. Definitely possible, but not something I wanted to dive deep into just for this. 3. The simplest solution was to write a `laravel middleware`. It checks for UTM parameters in the query string. If found, it `stores them in the session`. If the URL doesn’t have them but the session does, the middleware rewrites the query and continues. That’s all. No bloat. Works perfectly. The analytics data is clean and useful. ![Laravel Utm Tracking: Built-in Simplicity Beats Third-party Complexity](https://muhammadhuzaifa.pro/storage/photos/1/laravel-utm-tracking-built-in-simplicity-beats-third-party-complexity.webp) Before I wrap up, I want to share a quick note for computer science students. Using ChatGPT is fine. It’s a great help. But if you depend on it for every little thing, you're limiting your growth. Just like how child psychologists say kids should face challenges to learn problem-solving, students need to struggle a bit too. That’s how you build real skill. Try to understand the problem first and attempt a solution before asking GPT. Once you gain some experience, you’ll know when it makes sense to get help and when it’s better to go solo. If you found this post helpful, you might also enjoy two of my write-ups where I dive deeper into Laravel and SEO: [Technical SEO Configurator for Laravel by Honeystone](https://muhammadhuzaifa.pro/technical-seo-configurator-for-laravel-by-honeystone) and [Improve Your Laravel Ecommerce Store Technical SEO Using Butschster\LaravelMetaTags](https://muhammadhuzaifa.pro/improve-your-laravel-ecommerce-store-technical-seo-using-butschsterlaravelmetatags). I share hands-on insights and practical examples in both. And if you're looking for an experienced Laravel developer with a proven Upwork track record, feel free to [check out my Upwork profile](https://www.upwork.com/freelancers/~01dcc2f82551b47dee) and get in touch. Thanks for sticking around till the end. Stay happy, stay blessed. ### my first open source ai contribution: fixing laravel prismphp’s ollama stream bug URL: https://muhammadhuzaifa.pro/my-first-open-source-ai-contribution-fixing-laravel-prismphps-ollama-stream-bug Published: 2025-07-25 Another day, another achievement. I won’t hype it too much, but let’s be real—there’s something incredibly satisfying about seeing your small code contribution go live in a tool used by thousands, if not millions. I don’t usually rely on paid AI services. Instead, I prefer running local LLMs like DeepSeek and LLaMA using [Ollama](https://ollama.com/)—they’re free, fast, and my Dell G-series machine with 6GB GPU + 12-core CPU handles them just fine. Not long ago, I built a Laravel tool for myself using Laravel’s native HTTP client that talks directly with Ollama. While testing, I stumbled across a bug in the [prismphp.com](https://prismphp.com) package—specifically in `Stream.php`. The `system` param was being passed to the `/api/chat` endpoint, which the Ollama spec clearly doesn’t support. What next? My developer instinct kicked in—I jumped into the package source, fixed the issue, updated the tests, forked the repo, submitted the PR... and today I got the confirmation that **my PR has been merged into the official repo**. 🎉 Here's the diff in case if you want to review: ```diff From d2789d809490380590fc586bed0569a986456b78 Mon Sep 17 00:00:00 2001 From: Muhammad Huzaifa Date: Wed, 9 Jul 2025 18:28:13 +0500 Subject: [PATCH] fix(ollama-provider): remove unsupported 'system' param from `api/chat` endpoint and allow multiple system prompts Replace the incorrect use of the `system` body param with the correct `{ role, content }` message structure, as per Ollama's API spec. Now system prompts are prepended to the messages array to maintain compatibility with the chat format expected by `api/chat`. `system` body param is only available for `api/generate` endpoint. Since we are now prepending the system prompts, multiple system prompts are allowed by default, remove the restriction on multiple system prompts and remove the relevant tests as well. --- src/Providers/Ollama/Handlers/Stream.php | 10 +++---- src/Providers/Ollama/Handlers/Structured.php | 17 +++++------ src/Providers/Ollama/Handlers/Text.php | 10 +++---- tests/Providers/Ollama/StructuredTest.php | 30 -------------------- tests/Providers/Ollama/TextTest.php | 14 --------- 5 files changed, 15 insertions(+), 66 deletions(-) diff --git a/src/Providers/Ollama/Handlers/Stream.php b/src/Providers/Ollama/Handlers/Stream.php index bb8ae38..5b27815 100644 --- a/src/Providers/Ollama/Handlers/Stream.php +++ b/src/Providers/Ollama/Handlers/Stream.php @@ -174,17 +174,15 @@ class Stream protected function sendRequest(Request $request): Response { - if (count($request->systemPrompts()) > 1) { - throw new PrismException('Ollama does not support multiple system prompts using withSystemPrompt / withSystemPrompts. However, you can provide additional system prompts by including SystemMessages in with withMessages.'); - } - return $this ->client ->withOptions(['stream' => true]) ->post('api/chat', [ 'model' => $request->model(), - 'system' => data_get($request->systemPrompts(), '0.content', ''), - 'messages' => (new MessageMap($request->messages()))->map(), + 'messages' => (new MessageMap(array_merge( + $request->systemPrompts(), + $request->messages() + )))->map(), 'tools' => ToolMap::map($request->tools()), 'stream' => true, 'options' => Arr::whereNotNull(array_merge([ diff --git a/src/Providers/Ollama/Handlers/Structured.php b/src/Providers/Ollama/Handlers/Structured.php index 0b95cf7..e7bae61 100644 --- a/src/Providers/Ollama/Handlers/Structured.php +++ b/src/Providers/Ollama/Handlers/Structured.php @@ -6,7 +6,6 @@ namespace Prism\Prism\Providers\Ollama\Handlers; use Illuminate\Http\Client\PendingRequest; use Illuminate\Support\Arr; -use Prism\Prism\Exceptions\PrismException; use Prism\Prism\Providers\Ollama\Concerns\MapsFinishReason; use Prism\Prism\Providers\Ollama\Concerns\ValidatesResponse; use Prism\Prism\Providers\Ollama\Maps\MessageMap; @@ -76,20 +75,18 @@ class Structured */ protected function sendRequest(Request $request): array { - if (count($request->systemPrompts()) > 1) { - throw new PrismException('Ollama does not support multiple system prompts using withSystemPrompt / withSystemPrompts. However, you can provide additional system prompts by including SystemMessages in with withMessages.'); - } - $response = $this->client->post('api/chat', [ 'model' => $request->model(), - 'system' => data_get($request->systemPrompts(), '0.content', ''), - 'messages' => (new MessageMap($request->messages()))->map(), + 'messages' => (new MessageMap(array_merge( + $request->systemPrompts(), + $request->messages() + )))->map(), 'format' => $request->schema()->toArray(), 'stream' => false, 'options' => Arr::whereNotNull(array_merge([ - 'temperature' => $request->temperature(), - 'num_predict' => $request->maxTokens() ?? 2048, - 'top_p' => $request->topP(), + 'temperature' => $request->temperature(), + 'num_predict' => $request->maxTokens() ?? 2048, + 'top_p' => $request->topP(), ], $request->providerOptions())), ]); diff --git a/src/Providers/Ollama/Handlers/Text.php b/src/Providers/Ollama/Handlers/Text.php index f4777a2..b167da6 100644 --- a/src/Providers/Ollama/Handlers/Text.php +++ b/src/Providers/Ollama/Handlers/Text.php @@ -68,16 +68,14 @@ class Text */ protected function sendRequest(Request $request): array { - if (count($request->systemPrompts()) > 1) { - throw new PrismException('Ollama does not support multiple system prompts using withSystemPrompt / withSystemPrompts. However, you can provide additional system prompts by including SystemMessages in with withMessages.'); - } - $response = $this ->client ->post('api/chat', [ 'model' => $request->model(), - 'system' => data_get($request->systemPrompts(), '0.content', ''), - 'messages' => (new MessageMap($request->messages()))->map(), + 'messages' => (new MessageMap(array_merge( + $request->systemPrompts(), + $request->messages() + )))->map(), 'tools' => ToolMap::map($request->tools()), 'stream' => false, 'options' => Arr::whereNotNull(array_merge([ diff --git a/tests/Providers/Ollama/StructuredTest.php b/tests/Providers/Ollama/StructuredTest.php index 2630a8f..afb0ced 100644 --- a/tests/Providers/Ollama/StructuredTest.php +++ b/tests/Providers/Ollama/StructuredTest.php @@ -53,33 +53,3 @@ it('returns structured output', function (): void { expect($response->structured['open_source'])->toBeArray(); }); - -it('throws an exception with multiple system prompts', function (): void { - Http::preventStrayRequests(); - - $schema = new ObjectSchema( - 'output', - 'the output object', - [ - new StringSchema('name', 'The users name'), - new ArraySchema('hobbies', 'a list of the users hobbies', - new StringSchema('name', 'the name of the hobby'), - ), - new ArraySchema('open_source', 'The users open source contributions', - new StringSchema('name', 'the name of the project'), - ), - ], - ['name', 'hobbies', 'open_source'] - ); - - $response = Prism::structured() - ->using('ollama', 'qwen2.5:14b') - ->withSchema($schema) - ->withSystemPrompts([ - new SystemMessage('MODEL ADOPTS ROLE of [PERSONA: Nyx the Cthulhu]!'), - new SystemMessage('But my friends call my Nyx.'), - ]) - ->withPrompt('Who are you?') - ->asStructured(); - -})->throws(PrismException::class, 'Ollama does not support multiple system prompts using withSystemPrompt / withSystemPrompts. However, you can provide additional system prompts by including SystemMessages in with withMessages.'); diff --git a/tests/Providers/Ollama/TextTest.php b/tests/Providers/Ollama/TextTest.php index 505bf39..2a4d4f4 100644 --- a/tests/Providers/Ollama/TextTest.php +++ b/tests/Providers/Ollama/TextTest.php @@ -167,17 +167,3 @@ describe('Image support', function (): void { }); }); }); - -it('throws an exception with multiple system prompts', function (): void { - Http::preventStrayRequests(); - - $response = Prism::text() - ->using('ollama', 'qwen2.5:14b') - ->withSystemPrompts([ - new SystemMessage('MODEL ADOPTS ROLE of [PERSONA: Nyx the Cthulhu]!'), - new SystemMessage('But my friends call my Nyx.'), - ]) - ->withPrompt('Who are you?') - ->asText(); - -})->throws(PrismException::class, 'Ollama does not support multiple system prompts using withSystemPrompt / withSystemPrompts. However, you can provide additional system prompts by including SystemMessages in with withMessages.'); -- 2.34.1 ``` ![image](https://muhammadhuzaifa.pro/storage/photos/1/fix-prismphp-ollama-stream-bug.webp) ## 🚀 Why This Matters This wasn’t just a bugfix. It was a reminder that real-world contributions come from curiosity and the will to improve what you use. Whether it’s Laravel, Ollama, or a lesser-known AI package, fixing one small piece can help the whole community. ## 🔗 Let’s Build Something Together If you’re looking for someone who not only builds robust systems but also dives deep into debugging, fixing, and contributing back to the community—I’m your guy. 👉 **[Visit my Upwork Profile](https://www.upwork.com/freelancers/~01dcc2f82551b47dee)** to see what I’ve done, what I can do, and how we can work together. Let's build tools that don’t just work—they *improve others’ work too*. ### why you should touch the code that’s “working just fine” URL: https://muhammadhuzaifa.pro/why-you-should-touch-the-code-thats-working-just-fine Published: 2025-07-25 You’ve probably heard the saying (or seen the meme): > *“If it works, don’t touch it.”* Sounds funny, and sure, it gets a laugh. But if you’ve been in software development for a while—or even just care about writing code that others can read and build on—you know that mindset won’t get you far. Refactoring is *not* about fixing bugs. It’s about keeping your codebase clean, scalable, and up to standard. Think of it like cleaning your house—not because it’s falling apart, but because a tidy space helps you live better. #### My Recent Example The image below shows a cumulative Git commit. It’s a rebased commit made by squashing several refactor commits into one, then rebasing that into the `main` branch. ![Why You Should Touch the Code That’s “Working Just Fine”](https://muhammadhuzaifa.pro/storage/photos/1/refactor-laravel-code.webp) What’s special? It shows that I refactored the `Modules/Job` module by 73.5%. This didn’t just happen randomly. It started when I introduced **architectural tests**—and they failed. That failure gave me exactly what I needed: signals that the current structure had drifted. As the Chief Architect, I knew if I didn’t refactor now, the upcoming bi-weekly sprints would only stack on top of a shaky foundation. So I dove in. ### What Did the Refactor Involve? * ✅ Restructuring controllers to follow **Laravel’s resourceful conventions**, including single-action controllers * ✅ Aligning the codebase with **PSR standards** * ✅ Applying **SOLID principles** wherever possible * ✅ Creating a temporary `route.php` file with permanent redirects from old to new route structure (yes, even redirects deserve structure!) ### Why Refactor Anyway? Because every time you avoid it, tech debt grows. You don’t have to refactor daily. But weekly or bi-weekly, especially when code smells start showing? That’s the time. And if you’re interested in learning more about *how* to refactor well, I’d recommend following [Martin Fowler](https://martinfowler.com/), whose books and posts are a goldmine for understanding refactoring patterns. ### Want Your Code Refactored? If you’ve got a Laravel codebase that could use a cleanup—or even just want a second pair of eyes on it—I’m available for freelance consulting. You can reach out to me on [Upwork](https://www.upwork.com/freelancers/~01dcc2f82551b47dee), and I’d be happy to take a look. ### enforcing clean controller design in laravel with pestphp architectural testing URL: https://muhammadhuzaifa.pro/enforcing-clean-controller-design-in-laravel-with-pestphp-architectural-testing Published: 2025-07-20 As someone who’s always trying to bring structure and discipline to Laravel applications, I’ve been diving into **architectural testing using [PestPHP](https://pestphp.com)** lately. Pest has some excellent presets, especially when it comes to enforcing clean and maintainable code through expectations. One area I was particularly interested in was **controller method restrictions**. We already had a basic architectural test to ensure that Laravel controllers do not expose any unnecessary public methods—just the conventional ones like `index`, `show`, `store`, etc. But I wanted to take it a step further. > If a controller is invokable (i.e., has the `__invoke` method), then it should not have any other public methods except `__construct` and `__invoke`. Simple rule, but enforcing it across your growing codebase is a different story. ### PestPHP Limitation & Custom Test I initially tried to achieve this using PestPHP’s native `expect(...)->toHavePublicMethodsOnly([...])`, but there wasn’t a straightforward way to **conditionally skip** classes that aren't invokable. So I ended up writing a **custom architectural test** that walks through each controller file, determines if the class defines `__invoke`, and applies the rule accordingly. Here’s the snippet I came up with: ```php test('invokable controllers only have __construct and __invoke as public methods', function () { foreach (glob(module_path('User', 'app/Http/Controllers/**/*.php')) as $file) { $class = Str::of($file) ->after('/User') ->replace( ['/app', '.php'], '' )->toString(); $classNamespace = app() ->getProvider(UserServiceProvider::class) ->module_namespace('User', $class); if (method_exists($classNamespace, '__invoke')) { expect($classNamespace) ->not ->toHavePublicMethodsBesides(['__construct', '__invoke']); } } }); ``` ![pestphp-architectural-testing-output](https://muhammadhuzaifa.pro/storage/photos/1/pestphp-architectural-testing-output.webp) ### Why This Matters In a large codebase, small things like this matter more than we think. Enforcing clean, intention-driven controller design means fewer surprises, better maintainability, and ultimately a more robust system. Architectural testing is one of those underrated practices that can significantly improve your team’s consistency without becoming a bottleneck. And PestPHP makes it surprisingly elegant. If you're using **Laravel + Pest**, I highly recommend investing time in writing these small, meaningful assertions. It pays off. 💬 Have ideas on how this can be improved or made even cleaner? I’d love to hear your thoughts. If you're exploring similar architectural patterns in your Laravel apps, or if you have a project where you'd like robust architectural or feature tests written, feel free to reach, check out my work and connect with me on [Upwork](https://www.upwork.com/freelancers/~01dcc2f82551b47dee). ### 14 years in tech – a journey of curiosity, craft, and commitment URL: https://muhammadhuzaifa.pro/14-years-in-tech-a-journey-of-curiosity-craft-and-commitment Published: 2025-07-12 ## 14 Years In Tech – A Journey of Curiosity, Craft, and Commitment This July marks **14 years** since I wrote my first line of production code professionally — for a WordPress project. But the spark started long before that. Somewhere between 2002 and 2004, curiosity took root. It became something deeper in 2011, when I wrote my first “Hello World” in C. That simple line opened the doors to a journey that’s been nothing short of transformative — full of challenges, constant learning, and the chance to build, lead, and contribute meaningfully to the tech community. Over the years, I’ve worn many hats — from an ordinary student to a Chief Architect. I’ve worked with global teams, solved real business problems, mentored developers, and explored technologies that weren’t even on the radar when I first started. Whether it was integrating SAP with HubSpot, architecting multi-tenant SaaS platforms, contributing to open-source tools like **Laravel Telescope Guzzle Watcher**, or experimenting with **LLMs like Ollama**, my approach has remained simple: **build with purpose, stay grounded, and grow with people**. But this milestone isn’t just about the number of years. It’s about the journey — the systems I’ve helped shape, the incredible people I’ve worked with, the late nights chasing obscure bugs, and the quiet moments of pride when something just *clicks* and works. If there’s one piece of advice I could give to someone starting out: **learn to learn**. Stay humble, and focus on understanding problems deeply before rushing into solutions. One thing I’ve noticed lately — and I say this with some concern — is how easily we get caught up in tech trends. “This language is dying,” “That tool is obsolete,” “AI is replacing X.” When I chose PHP as my primary domain, I saw those headlines too. Then came Node.js, and the same culture followed. Here’s what I’ve learned: **tools change, but fundamentals don’t**. Strengthen your foundation in computer science. Learn how programs are built and run. Writing a C program teaches you things that apply to everything else. Master the basics, read the docs, build things yourself, and practice until it becomes muscle memory. I see many students today proudly build things using AI, but miss the essence of doing it themselves. Cognitive intelligence comes from your own knowledge and experience. AI — while powerful — is like a calculator. It should support your thinking, not replace it. I use LLMs myself, mostly for language and exploring ideas, but I still **implement and build things manually** — and that’s where real growth happens. To everyone who’s been part of this journey — mentors, teammates, clients, friends, and even the tough critics — thank you. You've shaped who I am today. ### enhancing laravel's apache solr integration with php-fig psr standards compliance URL: https://muhammadhuzaifa.pro/enhancing-laravels-apache-solr-integration-with-php-fig-psr-standards-compliance Published: 2024-11-08 # Enhancing Laravel's Apache Solr Integration with PHP-FIG PSR Standards Compliance Open source projects are the foundation of countless innovations, enabling developers to share knowledge and improve each other’s work. Recently, I had the chance to contribute to a valuable Laravel package designed for integrating Apache Solr—a powerful search engine known for scalability and speed—into Laravel projects. This package simplifies Solr core management and provides a fluent interface for managing Solr operations, making it a useful tool for developers who rely on Solr’s capabilities. ## Why Contribution Standards Matter: Ensuring Compliance with PHP PSR-0 While reviewing this package, I noticed a fundamental misalignment with the PHP PSR-0 standards, specifically regarding namespace conventions. PSR-0 specifies that namespaces should be in StudlyCaps, a convention that makes code more readable and maintains consistency across PHP projects. However, this package had used lowercase namespaces throughout its codebase, which directly violated the PSR-0 standard. Adhering to such standards isn’t just about aesthetics; it’s about setting up code for scalability and maintainability. Consistent use of standards makes it easier for developers to navigate and contribute to projects, as the code structure remains predictable. ## Contribution Highlights: Refactoring for Standards Compliance To align the package with PSR-0, I refactored all namespaces to StudlyCaps and updated references accordingly. This simple but essential change brought the package into compliance, making it easier for other engineers to understand and contribute without encountering namespace issues or the need for custom workarounds. ```diff -php artisan vendor:publish --provider="haiderjabbar\laravelsolr\LaravelSolrServiceProvider" +php artisan vendor:publish --provider="HaiderJabbar\LaravelSolr\LaravelSolrServiceProvider" ``` ### The Impact of Standards on Collaboration By aligning the package with PHP-FIG PSR standards, the code now follows widely accepted conventions, ensuring that future developers working on this project will face fewer barriers to entry. This change reflects a best practice that’s integral to quality software—following standards that make collaboration easier and codebases more maintainable. ## Have a Project You’d Like Reviewed? If you have an open-source project and would like a fresh set of eyes on it, feel free to reach out. I’m always open to collaborating and helping projects meet industry standards. ## Useful Links * [haiderjabbar/laravelsolr](https://github.com/haiderjabbar/laravelsolr) * [Pull Request](https://github.com/haiderjabbar/laravelsolr/pull/2) * [Apache Solr](https://solr.apache.org/) * [PHP FIG](https://www.php-fig.org/) ### fixing a hash mismatch bug in zen php sdk: ensuring accurate payment data validation URL: https://muhammadhuzaifa.pro/fixing-a-hash-mismatch-bug-in-zen-php-sdk-ensuring-accurate-payment-data-validation Published: 2024-11-06 # Fixing a Hash Mismatch Bug in Zen PHP SDK: Ensuring Accurate Payment Data Validation PR Link : [https://github.com/obanach/zen-php-sdk/pull/2](https://github.com/obanach/zen-php-sdk/pull/2) Recently, I encountered a bug in the Zen PHP SDK, a package designed for seamless integration with Zen.com, a prominent payment gateway service provider. The issue stemmed from a discrepancy in the Instant Payment Notification (IPN) data, specifically with how the package handled numerical values in the hash generation process. The IPN data I received included an amount of 2.70, but during hash generation, this value was interpreted as 2.7, leading to a mismatch. This seemingly minor difference was enough to disrupt the accuracy of the hash, which is critical for validating secure transactions. After a detailed investigation, I traced the problem to the `HashGenerator::generate` function, where the `$amount` parameter was defined as a `float`. PHP's internal type juggling may interpret float values in unpredictable ways, resulting in 2.7 instead of 2.70. This small discrepancy in decimal precision can cause significant issues in payment validation systems, where even minor differences can trigger security protocols or prevent successful verification. To resolve this, I modified the generate function by changing the `$amount` parameter type from `float` to `string`. This adjustment ensures that the exact data, including decimal precision, is passed to the SDK. By treating the amount as a string, we eliminate the risk of PHP adjusting the decimal precision, enabling reliable hash generation and validation. Although a minor contribution, I am proud to have improved the accuracy of the Zen PHP SDK. This fix underscores the importance of precision in open-source software and highlights the value of community-driven contributions to enhancing functionality and security in payment gateway integrations. Long live open-source systems! ### refactoring shopify's laravel web app template: improving best practices and code efficiency URL: https://muhammadhuzaifa.pro/refactoring-shopify-laravel-web-app-template-improving-best-practices-and-code-efficiency Published: 2024-10-31 Hello everyone, I hope you're all having a great day! While building a Shopify web application using Laravel for the first time, I encountered several areas where the code deviated from Laravel's best practices. The implementation included common pitfalls, such as: - **Fat Web Routes File**: Business logic and route handling cluttering up `web.php`. - **Direct Business Logic in Routes**: Essential logic embedded directly in routes, leading to an unstructured and hard-to-maintain codebase. - **ENV Variables Accessed Without Configuration Resolver**: Sensitive configuration data accessed improperly, lacking the necessary security and abstraction layers. - **Service Logic in App Service Provider**: Core service logic written directly into the `AppServiceProvider`, which is designed to handle application-wide settings, not heavy business logic. These issues sparked my motivation to contribute to improving the Shopify app's Laravel template. I made several optimizations and submitted a pull request (PR) to Shopify. The PR introduces enhancements aligned with Laravel best practices, making the template more maintainable, secure, and in line with standard development practices. Whether or not Shopify decides to merge the PR, I believe these updates are essential for anyone embarking on a new Shopify app project in Laravel. If you're planning to develop a Shopify app with Laravel, I'd recommend checking out the PR. Here's the link: [PR #524: Shopify Laravel Web App Improvements](https://github.com/Shopify/shopify-app-template-php/pull/524) You can also apply these changes directly by cherry-picking the diff: ```sh curl https://patch-diff.githubusercontent.com/raw/Shopify/shopify-app-template-php/pull/524.patch | git apply ``` Feel free to explore these updates, as they can streamline development and significantly improve the maintainability of your Shopify-Laravel integration. ```diff From 36ef3ff908b23d77fc1171b90702eae08f8a6579 Mon Sep 17 00:00:00 2001 From: Muhammad Huzaifa Date: Tue, 8 Oct 2024 22:49:07 +0500 Subject: [PATCH 1/2] refactor: update the code as the laravel best practices --- web/app/Providers/AppServiceProvider.php | 53 +------------- web/app/Providers/ShopifyServiceProvider.php | 74 ++++++++++++++++++++ web/config/app.php | 1 + web/config/shopify.php | 71 +++++++++++++++++-- 4 files changed, 141 insertions(+), 58 deletions(-) create mode 100644 web/app/Providers/ShopifyServiceProvider.php diff --git a/web/app/Providers/AppServiceProvider.php b/web/app/Providers/AppServiceProvider.php index 5f5288fa4..857898dcc 100644 --- a/web/app/Providers/AppServiceProvider.php +++ b/web/app/Providers/AppServiceProvider.php @@ -2,17 +2,7 @@ namespace App\Providers; -use App\Lib\DbSessionStorage; -use App\Lib\Handlers\AppUninstalled; -use App\Lib\Handlers\Privacy\CustomersDataRequest; -use App\Lib\Handlers\Privacy\CustomersRedact; -use App\Lib\Handlers\Privacy\ShopRedact; use Illuminate\Support\ServiceProvider; -use Illuminate\Support\Facades\URL; -use Shopify\Context; -use Shopify\ApiVersion; -use Shopify\Webhooks\Registry; -use Shopify\Webhooks\Topics; class AppServiceProvider extends ServiceProvider { @@ -26,49 +16,8 @@ public function register() // } - /** - * Bootstrap any application services. - * - * @return void - * @throws \Shopify\Exception\MissingArgumentException - */ public function boot() { - $host = str_replace('https://', '', env('HOST', 'not_defined')); - - $customDomain = env('SHOP_CUSTOM_DOMAIN', null); - Context::initialize( - env('SHOPIFY_API_KEY', 'not_defined'), - env('SHOPIFY_API_SECRET', 'not_defined'), - env('SCOPES', 'not_defined'), - $host, - new DbSessionStorage(), - ApiVersion::LATEST, - true, - false, - null, - '', - null, - (array)$customDomain, - ); - - URL::forceRootUrl("https://$host"); - URL::forceScheme('https'); - - Registry::addHandler(Topics::APP_UNINSTALLED, new AppUninstalled()); - - /* - * This sets up the mandatory privacy webhooks. You’ll need to fill in the endpoint to be used by your app in - * the “Privacy webhooks” section in the “App setup” tab, and customize the code when you store customer data - * in the handlers being registered below. - * - * More details can be found on shopify.dev: - * https://shopify.dev/docs/apps/webhooks/configuration/mandatory-webhooks - * - * Note that you'll only receive these webhooks if your app has the relevant scopes as detailed in the docs. - */ - Registry::addHandler('CUSTOMERS_DATA_REQUEST', new CustomersDataRequest()); - Registry::addHandler('CUSTOMERS_REDACT', new CustomersRedact()); - Registry::addHandler('SHOP_REDACT', new ShopRedact()); + // } } diff --git a/web/app/Providers/ShopifyServiceProvider.php b/web/app/Providers/ShopifyServiceProvider.php new file mode 100644 index 000000000..4a35d82a4 --- /dev/null +++ b/web/app/Providers/ShopifyServiceProvider.php @@ -0,0 +1,74 @@ + env('HOST'), + + /* + |-------------------------------------------------------------------------- + | Shopify custom domain + |-------------------------------------------------------------------------- + | + | One or more regexps to use when validating domains. + | + */ + 'shop_custom_domain' => env('SHOP_CUSTOM_DOMAIN'), + + /* + |-------------------------------------------------------------------------- + | Shopify API Key + |-------------------------------------------------------------------------- + | + | The client ID of the app, retrieved using Shopify CLI. + | + | Learn more about in documentation: https://shopify.dev/docs/apps/launch/deployment/deploy-web-app/deploy-to-hosting-service#step-4-set-up-environment-variables + | + */ + 'api_key' => env('SHOPIFY_API_KEY'), + + /* + |-------------------------------------------------------------------------- + | Shopify API Secret + |-------------------------------------------------------------------------- + | + | The client secret of the app, retrieved using Shopify CLI. + | + | Learn more about in documentation: https://shopify.dev/docs/apps/launch/deployment/deploy-web-app/deploy-to-hosting-service#step-4-set-up-environment-variables + | + */ + 'api_secret' => env('SHOPIFY_API_SECRET'), + + /* + |-------------------------------------------------------------------------- + | Shopify Scopes + |-------------------------------------------------------------------------- + | + | The app's access scopes, retrieved using Shopify CLI. This is optional if you're using Shopify-managed installation. + | + | Learn more about in documentation: https://shopify.dev/docs/apps/launch/deployment/deploy-web-app/deploy-to-hosting-service#step-4-set-up-environment-variables + | + */ + 'scopes' => env('SCOPES'), + /* |-------------------------------------------------------------------------- | Shopify billing @@ -17,14 +76,14 @@ | Learn more about billing in our documentation: https://shopify.dev/docs/apps/billing | */ - "billing" => [ - "required" => false, + 'billing' => [ + 'required' => false, // Example set of values to create a charge for $5 one time - "chargeName" => "My Shopify App One-Time Billing", - "amount" => 5.0, - "currencyCode" => "USD", // Currently only supports USD - "interval" => EnsureBilling::INTERVAL_ONE_TIME, + 'chargeName' => 'My Shopify App One-Time Billing', + 'amount' => 5.0, + 'currencyCode' => 'USD', // Currently only supports USD + 'interval' => EnsureBilling::INTERVAL_ONE_TIME, ], ]; From 84bd33f0999fb88e800349d099878ab558b5eff6 Mon Sep 17 00:00:00 2001 From: Muhammad Huzaifa Date: Sun, 27 Oct 2024 23:01:14 +0500 Subject: [PATCH 2/2] refactor: organize web.php file by extracting route business logics to their own controllers. --- .../Controllers/Shopify/AuthController.php | 64 +++++++++ .../Shopify/FallbackController.php | 27 ++++ .../Controllers/Shopify/ProductController.php | 57 ++++++++ .../Controllers/Shopify/WebhookController.php | 38 +++++ web/app/Providers/ShopifyServiceProvider.php | 3 + web/routes/web.php | 132 ++---------------- 6 files changed, 199 insertions(+), 122 deletions(-) create mode 100644 web/app/Http/Controllers/Shopify/AuthController.php create mode 100644 web/app/Http/Controllers/Shopify/FallbackController.php create mode 100644 web/app/Http/Controllers/Shopify/ProductController.php create mode 100644 web/app/Http/Controllers/Shopify/WebhookController.php diff --git a/web/app/Http/Controllers/Shopify/AuthController.php b/web/app/Http/Controllers/Shopify/AuthController.php new file mode 100644 index 000000000..4a9c6b103 --- /dev/null +++ b/web/app/Http/Controllers/Shopify/AuthController.php @@ -0,0 +1,64 @@ + + */ +class AuthController extends Controller +{ + public function index(Request $request) + { + $shop = Utils::sanitizeShopDomain($request->query('shop')); + + // Delete any previously created OAuth sessions that were not completed (don't have an access token) + Session::where('shop', $shop)->where('access_token', null)->delete(); + + return AuthRedirection::redirect($request); + } + + public function callback(Request $request) + { + $session = OAuth::callback( + $request->cookie(), + $request->query(), + ['App\Lib\CookieHandler', 'saveShopifyCookie'], + ); + + $host = $request->query('host'); + $shop = Utils::sanitizeShopDomain($request->query('shop')); + + $response = Registry::register('/api/webhooks', Topics::APP_UNINSTALLED, $shop, $session->getAccessToken()); + if ($response->isSuccess()) { + Log::debug("Registered APP_UNINSTALLED webhook for shop $shop"); + } else { + Log::error( + "Failed to register APP_UNINSTALLED webhook for shop $shop with response body: ". + print_r($response->getBody(), true) + ); + } + + $redirectUrl = Utils::getEmbeddedAppUrl($host); + if (Config::get('shopify.billing.required')) { + [$hasPayment, $confirmationUrl] = EnsureBilling::check($session, Config::get('shopify.billing')); + + if (! $hasPayment) { + $redirectUrl = $confirmationUrl; + } + } + + return redirect($redirectUrl); + } +} diff --git a/web/app/Http/Controllers/Shopify/FallbackController.php b/web/app/Http/Controllers/Shopify/FallbackController.php new file mode 100644 index 000000000..b2b6193c0 --- /dev/null +++ b/web/app/Http/Controllers/Shopify/FallbackController.php @@ -0,0 +1,27 @@ + + */ +class FallbackController extends Controller +{ + public function __invoke(Request $request) + { + if (Context::$IS_EMBEDDED_APP && $request->query('embedded', false) === '1') { + if (env('APP_ENV') === 'production') { + return file_get_contents(public_path('index.html')); + } else { + return file_get_contents(base_path('frontend/index.html')); + } + } + + return redirect(Utils::getEmbeddedAppUrl($request->query('host', null)).'/'.$request->path()); + } +} diff --git a/web/app/Http/Controllers/Shopify/ProductController.php b/web/app/Http/Controllers/Shopify/ProductController.php new file mode 100644 index 000000000..77146696c --- /dev/null +++ b/web/app/Http/Controllers/Shopify/ProductController.php @@ -0,0 +1,57 @@ + + */ +class ProductController extends Controller +{ + public function count(Request $request) + { + /** @var \Shopify\Auth\Session */ + $session = $request->get('shopifySession'); // Provided by the shopify.auth middleware, guaranteed to be active + + $client = new Rest($session->getShop(), $session->getAccessToken()); + $result = $client->get('products/count'); + + return response($result->getDecodedBody()); + } + + public function store(Request $request) + { + /** @var \Shopify\Auth\Session */ + $session = $request->get('shopifySession'); // Provided by the shopify.auth middleware, guaranteed to be active + + $success = $code = $error = null; + try { + ProductCreator::call($session, 5); + $success = true; + $code = 200; + $error = null; + } catch (\Exception $e) { + $success = false; + + if ($e instanceof ShopifyProductCreatorException) { + $code = $e->response->getStatusCode(); + $error = $e->response->getDecodedBody(); + if (array_key_exists('errors', $error)) { + $error = $error['errors']; + } + } else { + $code = 500; + $error = $e->getMessage(); + } + + Log::error("Failed to create products: $error"); + } finally { + return response()->json(['success' => $success, 'error' => $error], $code); + } + } +} diff --git a/web/app/Http/Controllers/Shopify/WebhookController.php b/web/app/Http/Controllers/Shopify/WebhookController.php new file mode 100644 index 000000000..95a65ab1d --- /dev/null +++ b/web/app/Http/Controllers/Shopify/WebhookController.php @@ -0,0 +1,38 @@ + + */ +class WebhookController extends Controller +{ + public function __invoke(Request $request) + { + try { + $topic = $request->header(HttpHeaders::X_SHOPIFY_TOPIC, ''); + + $response = Registry::process($request->header(), $request->getContent()); + if (! $response->isSuccess()) { + Log::error("Failed to process '$topic' webhook: {$response->getErrorMessage()}"); + + return response()->json(['message' => "Failed to process '$topic' webhook"], 500); + } + } catch (InvalidWebhookException $e) { + Log::error("Got invalid webhook request for topic '$topic': {$e->getMessage()}"); + + return response()->json(['message' => "Got invalid webhook request for topic '$topic'"], 401); + } catch (\Exception $e) { + Log::error("Got an exception when handling '$topic' webhook: {$e->getMessage()}"); + + return response()->json(['message' => "Got an exception when handling '$topic' webhook"], 500); + } + } +} diff --git a/web/app/Providers/ShopifyServiceProvider.php b/web/app/Providers/ShopifyServiceProvider.php index 4a35d82a4..5054d131c 100644 --- a/web/app/Providers/ShopifyServiceProvider.php +++ b/web/app/Providers/ShopifyServiceProvider.php @@ -14,6 +14,9 @@ use Shopify\Webhooks\Registry; use Shopify\Webhooks\Topics; +/** + * @author Muhammad Huzaifa + */ class ShopifyServiceProvider extends ServiceProvider { /** diff --git a/web/routes/web.php b/web/routes/web.php index f8e6c01b6..019137036 100644 --- a/web/routes/web.php +++ b/web/routes/web.php @@ -1,23 +1,10 @@ query("embedded", false) === "1") { - if (env('APP_ENV') === 'production') { - return file_get_contents(public_path('index.html')); - } else { - return file_get_contents(base_path('frontend/index.html')); - } - } else { - return redirect(Utils::getEmbeddedAppUrl($request->query("host", null)) . "/" . $request->path()); - } -})->middleware('shopify.installed'); +Route::fallback(FallbackController::class)->middleware('shopify.installed'); -Route::get('/api/auth', function (Request $request) { - $shop = Utils::sanitizeShopDomain($request->query('shop')); +Route::get('/api/auth', [AuthController::class, 'index']); - // Delete any previously created OAuth sessions that were not completed (don't have an access token) - Session::where('shop', $shop)->where('access_token', null)->delete(); +Route::get('/api/auth/callback', [AuthController::class, 'callback']); - return AuthRedirection::redirect($request); -}); +Route::get('/api/products/count', [ProductController::class, 'count'])->middleware('shopify.auth'); -Route::get('/api/auth/callback', function (Request $request) { - $session = OAuth::callback( - $request->cookie(), - $request->query(), - ['App\Lib\CookieHandler', 'saveShopifyCookie'], - ); +Route::post('/api/products', [ProductController::class, 'store'])->middleware('shopify.auth'); - $host = $request->query('host'); - $shop = Utils::sanitizeShopDomain($request->query('shop')); - - $response = Registry::register('/api/webhooks', Topics::APP_UNINSTALLED, $shop, $session->getAccessToken()); - if ($response->isSuccess()) { - Log::debug("Registered APP_UNINSTALLED webhook for shop $shop"); - } else { - Log::error( - "Failed to register APP_UNINSTALLED webhook for shop $shop with response body: " . - print_r($response->getBody(), true) - ); - } - - $redirectUrl = Utils::getEmbeddedAppUrl($host); - if (Config::get('shopify.billing.required')) { - list($hasPayment, $confirmationUrl) = EnsureBilling::check($session, Config::get('shopify.billing')); - - if (!$hasPayment) { - $redirectUrl = $confirmationUrl; - } - } - - return redirect($redirectUrl); -}); - -Route::get('/api/products/count', function (Request $request) { - /** @var AuthSession */ - $session = $request->get('shopifySession'); // Provided by the shopify.auth middleware, guaranteed to be active - - $client = new Rest($session->getShop(), $session->getAccessToken()); - $result = $client->get('products/count'); - - return response($result->getDecodedBody()); -})->middleware('shopify.auth'); - -Route::post('/api/products', function (Request $request) { - /** @var AuthSession */ - $session = $request->get('shopifySession'); // Provided by the shopify.auth middleware, guaranteed to be active - - $success = $code = $error = null; - try { - ProductCreator::call($session, 5); - $success = true; - $code = 200; - $error = null; - } catch (\Exception $e) { - $success = false; - - if ($e instanceof ShopifyProductCreatorException) { - $code = $e->response->getStatusCode(); - $error = $e->response->getDecodedBody(); - if (array_key_exists("errors", $error)) { - $error = $error["errors"]; - } - } else { - $code = 500; - $error = $e->getMessage(); - } - - Log::error("Failed to create products: $error"); - } finally { - return response()->json(["success" => $success, "error" => $error], $code); - } -})->middleware('shopify.auth'); - -Route::post('/api/webhooks', function (Request $request) { - try { - $topic = $request->header(HttpHeaders::X_SHOPIFY_TOPIC, ''); - - $response = Registry::process($request->header(), $request->getContent()); - if (!$response->isSuccess()) { - Log::error("Failed to process '$topic' webhook: {$response->getErrorMessage()}"); - return response()->json(['message' => "Failed to process '$topic' webhook"], 500); - } - } catch (InvalidWebhookException $e) { - Log::error("Got invalid webhook request for topic '$topic': {$e->getMessage()}"); - return response()->json(['message' => "Got invalid webhook request for topic '$topic'"], 401); - } catch (\Exception $e) { - Log::error("Got an exception when handling '$topic' webhook: {$e->getMessage()}"); - return response()->json(['message' => "Got an exception when handling '$topic' webhook"], 500); - } -}); +Route::post('/api/webhooks', WebhookController::class); ``` ### enhanced laravel telescope guzzle watcher with multipart request support URL: https://muhammadhuzaifa.pro/enhanced-laravel-telescope-guzzle-watcher-with-multipart-request-support Published: 2024-10-26 # Enhanced Laravel Telescope Guzzle Watcher With Multipart Request Support A key advancement in this project is the addition of multipart request support. Recognizing the need for robust tracking capabilities, especially when handling multipart payloads, I implemented functionality that now allows Laravel Telescope Guzzle Watcher to record multipart requests effectively. In instances where a payload includes files, the solution smartly excludes file content while preserving essential metadata and remaining data for streamlined monitoring and enhanced performance. ```php use GuzzleHttp\Psr7; $client->request('POST', '/post', [ 'multipart' => [ [ 'name' => 'foo', 'contents' => 'data', 'headers' => ['X-Baz' => 'bar'] ], [ 'name' => 'baz', 'contents' => Psr7\Utils::tryFopen(base_path('composer.json'), 'r') ], ] ]); ``` The above request will be displayed like the one below after being recorded in the request. ![image](https://muhammadhuzaifa.pro/storage/photos/1/laravel-telescope-guzzle-watcher-3.2.0-1.png) The package is a perfect fit for applications that rely on complex integrations requiring nuanced monitoring, and it provides developers with deep insights into API interactions through Guzzle requests within Laravel applications. This package has now become a trusted component in the Laravel ecosystem for developers looking to elevate their project’s monitoring and debugging capabilities. ### why should you learn design patterns? URL: https://muhammadhuzaifa.pro/why-should-you-learn-design-patterns Published: 2024-10-23 # Why Should You Learn Design Patterns? Design patterns are essential in software development because they provide proven solutions to common problems. By understanding and applying these patterns, you can write more maintainable, scalable, and flexible code. One great example of this is how the **Adapter Pattern** can simplify and streamline code when working with external libraries or APIs. Take a look at this article by [Stefan Zweifel](https://stefanzweifel.dev/posts/2022/05/29/use-laravels-http-client-with-facebooks-business-sdk/), where he uses the **Adapter Pattern** to integrate Laravel’s HTTP client with Facebook’s Business SDK. This design pattern made it easy for him to extend and implement custom business logic to handle HTTP requests more efficiently. By employing the **Adapter Pattern**, he was able to: - Leverage Laravel's **HTTP facade**, including the fake feature for testing without modifying the SDK itself. - Monitor HTTP requests easily using **Laravel's Pulse** and **[Telescope](https://muhammadhuzaifa.pro/laravel-telescope-first-party-laravel-package-for-debugging)** tools, giving full visibility into each request and response during development. But what exactly does the Adapter Pattern do? Simply put, it allows two incompatible interfaces to work together. In this case, Stefan wanted to use Laravel’s clean and robust HTTP client while still interacting with Facebook’s SDK, which had its own implementation for handling HTTP requests. The Adapter Pattern provided a seamless way to bridge these two worlds, encapsulating the differences and allowing him to use Laravel's native tools without changing the SDK’s core functionality. ## Benefits of the Adapter Pattern: 1. **Flexibility**: It allows you to integrate third-party libraries or legacy systems with your codebase without modifying the original source. 2. **Extensibility**: You can implement custom business logic as needed without affecting the core systems. 3. **Maintainability**: By keeping the implementation separate from the client code, any future changes in the third-party API can be handled within the adapter, leaving the rest of your application unaffected. This is just one example of why learning design patterns like the Adapter Pattern can elevate your coding skills. They are key to writing clean, reusable, and scalable code. ### technical seo configurator for laravel by honeystone URL: https://muhammadhuzaifa.pro/technical-seo-configurator-for-laravel-by-honeystone Published: 2024-09-17 Last updated: 2024-09-19 # Technical SEO Configurator for Laravel by honeystone > [Improve Your Laravel Ecommerce Store Technical Seo Using butschster/laravelmetatags](https://muhammadhuzaifa.pro/improve-your-laravel-ecommerce-store-technical-seo-using-butschsterlaravelmetatags) could be a good alternative for managing Technical SEO in your Laravel project. The Honeystone SEO package makes configuring SEO metadata from anywhere within your Laravel application a breeze. Included are metadata generators for general metadata, X (Formally Twitter) Cards, Open Graph, JSON-LD Schema, and Favicons (generated using [RealFaviconGenerator](https://realfavicongenerator.net)). This package was designed with extensibility in mind, so your own custom metadata generators can also be added with ease. ## Installation ```shell composer require honeystone/laravel-seo ``` Publish the configuration file with: ```shell php artisan vendor:publish --tag=honeystone-seo-config ``` ## Usage The package provides a helper function, `seo()`, and some Blade directives, `@metadata` and `@openGraphPrefix`. You can also typehint the `Honeystone\Seo\MetadataDirector` if you prefer to use dependency injection. Setting metadata is a simple as chaining methods: ```php seo() ->title('A fantastic blog post', 'My Awesome Website!') ->description('Theres really a lot of great stuff in here...') ->images( 'https://mywebsite.com/images/blog-1/cover-image.webp', 'https://mywebsite.com/images/blog-1/another-image.webp', ); ``` Once you've set your metadata, you can render it using: ```php seo()->generate(); ``` Alternatively, you can also use the `@metadata` Blade directive. The rendered result will look something like this: ```html A fantastic blog post - My Awesome Website! ``` ### Default methods Values provided to default methods will automatically propagate to all configured metadata generators. The following default methods are available: ```php seo() ->locale('en_GB') ->title('A fantastic blog post', template: '🔥🔥 {title} 🔥🔥') ->description('Theres really a lot of great stuff in here...') ->keywords('foo', 'bar', 'baz') ->url('https://mywebsite.com/blog/a-fantastic-blog-post') //defaults to the current url ->canonical('https://mywebsite.com/blog/a-fantastic-blog-post') //by default url and canonical are in sync, see config ->canonicalEnabled(true) //enabled by default, see config ->images( 'https://mywebsite.com/images/blog-1/cover-image.webp', 'https://mywebsite.com/images/blog-1/another-image.webp', ) ->robots('🤖', '🤖', '🤖'); ``` The full baseline looks like this: ```html 🔥🔥 A fantastic blog post 🔥🔥 ``` For your homepage you'll probably want to disable the title template: ```php seo()->title('My Awesome Website!', template: false); ``` ### Meta methods The meta methods are provided by the `Honeystone\Seo\Generators\MetaGenerator` class. Here's the full list: ```php seo() ->metaTitle('A fantastic blog post') ->metaTitleTemplate('🔥🔥 {title} 🔥🔥') ->metaDescription('Theres really a lot of great stuff in here...') ->metaKeywords('foo', 'bar', 'baz') ->metaCanonical('https://mywebsite.com/blog/a-fantastic-blog-post') ->metaCanonicalEnabled(true) ->metaRobots('🤖', '🤖', '🤖'); ``` All of these are provided by the default methods and propagate through to the meta generator. If you only want to render the meta generator, use `seo()->generate('meta')` or `@metadata('meta')` ### Twitter methods The meta methods are provided by the `Honeystone\Seo\Generators\TwitterGenerator` class. Here's the full list: ```php seo() ->twitterEnabled(true) //enabled by default, see config ->twitterSite('@MyWebsite') ->twitterCreator('@MyTwitter') ->twitterTitle('A fantastic blog post') //defaults to title() ->twitterDescription('Theres really a lot of great stuff in here...') //defaults to description() ->twitterImage('https://mywebsite.com/images/blog-1/cover-image.webp'); //defaults to the first in images() ``` You can learn more about this package, get full installation instructions, and view the [source code](https://github.com/Honeystone/laravel-seo) on GitHub ### laravel telescope: first party laravel package for debugging URL: https://muhammadhuzaifa.pro/laravel-telescope-first-party-laravel-package-for-debugging Published: 2024-09-10 # Laravel Telescope: First Party Laravel Package for Debugging > Laravel Telescope makes a wonderful companion to your local Laravel development environment. Telescope provides insight into the requests coming into your application, exceptions, log entries, database queries, queued jobs, mail, notifications, cache operations, scheduled tasks, variable dumps, and more. > > -- [Laravel Telescope Documentation](https://laravel.com/docs/11.x/telescope#introduction) Laravel Telescope is a first party package, means it is developed by the same team who develop Laravel core and other first party packages. Laravel Telescope intercepts all of your interactions with its core API such as exceptions, log entries, database queries and much more. Laravel Telescope provides the following features or watchers, lets briefly discussed each one of them. ## Client Watcher (Incoming HTTP Requests) `Laravel\Telescope\Watchers\ClientRequestWatcher` is responsible for watching HTTP requests and it listens for the `Illuminate\Http\Client\Events\ConnectionFailed` or `Illuminate\Http\Client\Events\ResponseReceived` events and record the request under the Requests section. ## Command Watcher `Laravel\Telescope\Watchers\CommandWatcher` is responsible for watching commands and it listens for the `Illuminate\Console\Events\CommandFinished` events and record the command under the Commands section. Following commands are ignored by default and won't be recorded by the watcher. ```php /** * Determine if the event should be ignored. * * @param mixed $event * @return bool */ private function shouldIgnore($event) { return in_array($event->command, array_merge($this->options['ignore'] ?? [], [ 'schedule:run', 'schedule:finish', 'package:discover', ])); } ``` ## Schedule Watcher `Laravel\Telescope\Watchers\ScheduleWatcher` is responsible for watching scheduler and it listens for the `Illuminate\Console\Events\CommandStarting` at first and then each event from the `Schedule::class` and record them under the Schedule section. ```php /** * Record a scheduled command was executed. * * @param \Illuminate\Console\Events\CommandStarting $event * @return void */ public function recordCommand(CommandStarting $event) { if (! Telescope::isRecording() || $event->command !== 'schedule:run' && $event->command !== 'schedule:finish') { return; } collect(app(Schedule::class)->events())->each(function ($event) { $event->then(function () use ($event) { Telescope::recordScheduledCommand(IncomingEntry::make([ 'command' => $event instanceof CallbackEvent ? 'Closure' : $event->command, 'description' => $event->description, 'expression' => $event->expression, 'timezone' => $event->timezone, 'user' => $event->user, 'output' => $this->getEventOutput($event), ])); }); }); } ``` ## Batch Watcher `Laravel\Telescope\Watchers\BatchWatcher` is responsible for watching dispatched batches and it listens for the `Illuminate\Bus\Events\BatchDispatched` event and record it under the Batch section. ## Cache Watcher `Laravel\Telescope\Watchers\CacheWatcher` is responsible for watching cache data and it listens for the following events - `Illuminate\Cache\Events\CacheHit` - `Illuminate\Cache\Events\CacheMissed` - `Illuminate\Cache\Events\KeyWritten` - `Illuminate\Cache\Events\KeyForgotten` ```php /** * Register the watcher. * * @param \Illuminate\Contracts\Foundation\Application $app * @return void */ public function register($app) { $app['events']->listen(CacheHit::class, [$this, 'recordCacheHit']); $app['events']->listen(CacheMissed::class, [$this, 'recordCacheMissed']); $app['events']->listen(KeyWritten::class, [$this, 'recordKeyWritten']); $app['events']->listen(KeyForgotten::class, [$this, 'recordKeyForgotten']); } ``` ## Dump Watcher `Laravel\Telescope\Watchers\DumpWatcher` is responsible for recording any dumps you might be using. It does listens for any events, the watcher uses the following `Symfony` components and just record the dump as it is. - `Symfony\Component\VarDumper\Cloner\VarCloner` - `Symfony\Component\VarDumper\Dumper\HtmlDumper` - `Symfony\Component\VarDumper\VarDumper` ## Event Watcher `Laravel\Telescope\Watchers\EventWatcher` is a global watcher and it listens for almost every event and record it under the Events section, however there are couple of events which are ignored by the Laravel Telescope and are fired internally by the Laravel framework. ```php /** * Determine if the event was fired internally by Laravel. * * @param string $eventName * @return bool */ protected function eventIsFiredByTheFramework($eventName) { return Str::is( [ 'Illuminate\*', 'Laravel\Octane\*', 'Laravel\Scout\Events\ModelsImported', 'eloquent*', 'bootstrapped*', 'bootstrapping*', 'creating*', 'composing*', ], $eventName ); } ``` We will look into the further watchers in the next article. ### birthday wishes from team computan URL: https://muhammadhuzaifa.pro/birthday-wishes-from-team-computan Published: 2024-09-10 > Your dedication and enthusiasm make a real difference, and it’s great to have you on board. Here’s to celebrating you today and wishing you a year full of success and happiness. Thank you team [Computan](https://computan.com) for the well wishes. ### improve your laravel ecommerce store technical seo using butschster/laravelmetatags URL: https://muhammadhuzaifa.pro/improve-your-laravel-ecommerce-store-technical-seo-using-butschsterlaravelmetatags Published: 2024-09-09 Last updated: 2024-09-17 # The most powerful and extendable tools for managing SEO Meta Tags in your Laravel project > [Technical Seo Configurator For Laravel By Honeystone](https://muhammadhuzaifa.pro/technical-seo-configurator-for-laravel-by-honeystone) could be a good alternative for managing Technical SEO in your Laravel project. Laravel SEO Meta Tags offers a sophisticated solution for Laravel applications, allowing developers to seamlessly manage header meta tags, CSS, JavaScript, and other relevant tags. Its primary objective is to simplify the process of managing search engine optimization (SEO) tags within your application. **Moreover, its versatile API makes it compatible with frameworks like [Inertiajs](https://inertiajs.com/), [VueJS](https://nuxtjs.org/docs/2.x/features/meta-tags-seo) and other JavaScript frameworks.** ## Features - **Meta Management:** Effortlessly set titles, charset, pagination links, and more. - **Styles & Scripts:** Organize and place styles and scripts anywhere in your HTML. - **Custom Tags:** Make your own tags to suit specific needs. - **Rich Media Integration:** Supports both Open Graph & Twitter Cards. - **Analytics Ready:** Comes with Google Analytics and Yandex Metrika tracking code support, including a code builder for the latter. - **Site Verification:** Supports webmaster tools site verifier tags. - **Package System:** Group tags, styles, and scripts into named packages for easy inclusion anywhere. - **Robust Documentation:** Clear instructions and guidelines for a seamless setup. - **Thoroughly Tested:** Built to ensure reliability and stability. ### Requirements - Laravel version: 9.x to 10.x - PHP version: 8.0 or higher ## Installation and Configuration 1. **Install the package:** Use the following command to install the Meta Tags package in your awesome application. ```shell composer require butschster/meta-tags ``` 2. **Register the Service Provider:** After installing the package, you must register its service provider. You can do it using the following artisan command: ```shell php artisan meta-tags:install ``` This command will activate the `App\Providers\MetaTagsServiceProvider` and publish the configuration at `config/meta_tags.php`. Within this configuration file, you can set default titles, keywords, descriptions, and other meta tags which will be automatically inserted into your HTML. 3. **Verification:** Ensure that `App\Providers\MetaTagsServiceProvider` has been added to the providers array in your `config/app.php` configuration file. If it hasn't, you'll need to add it manually. Remember, if your application isn't using the `App` namespace, update the provider class name accordingly. And that's all! Your Laravel project is now equipped to handle SEO meta tags with ease. ## Usage ### Controller You can use either Facade `\Butschster\Head\Facades\Meta` or `\Butschster\Head\Contracts\MetaTags\MetaInterface` in your controller ```php use Butschster\Head\MetaTags\MetaInterface; class HomeController extends Controller { public function __contruct( protected MetaInterface $meta ) { } public function index() { $news = News::paginate(); // Prepend title part to the default title $this->meta // Will render "Home page - Default Title" ->prependTitle('Home page') // Will include next, prev, canonical links ->setPaginationLinks($news) // Will change default favicon ->setFavicon('/favicon-index.ico') } } // Or you can use the facade use Butschster\Head\Facades\Meta; class HomeController extends Controller { public function index() { $news = News::paginate(); Meta::prependTitle('Home page') ->setPaginationLinks($news) ->setFavicon('favicon-index.ico'); } } ``` #### View To integrate meta tags into your HTML, simply insert `{!! Meta::toHtml() !!}` wherever required. > **Note** > You have two options to insert meta tags: `{!! Meta::toHtml() !!}` or the Blade directive `@meta_tags`. Here's an example of how you can use it in your view: ```html {!! Meta::toHtml() !!} ... ``` ### Placements The package provides flexibility to insert meta tags beyond the header. You can define specific placements in your templates. - To place meta tags in the footer, use: `Meta::footer()->toHtml()` - For custom placements, use: `Meta::placement('placement_name')->toHtml()` - Alternatively, the Blade directive can also be used: `@meta_tags('placement_name')` ```html ... {!! Meta::placement('middle_of_the_page')->toHtml() !!} ... {!! Meta::footer()->toHtml() !!} ``` You can learn more about this package, get full installation instructions, and view the [source code](https://github.com/butschster/LaravelMetaTags) on GitHub ### laravel cloud trailor by taylor URL: https://muhammadhuzaifa.pro/laravel-cloud-trailor-by-taylor Published: 2024-09-06 Laravel Cloud is your new fully managed infrastructure platform for Laravel. Go from Hello World to Hello Web in one minute or less. ### laravel raises a $57 million series a from accel URL: https://muhammadhuzaifa.pro/laravel-raises-a-57-million-series-a-from-accel Published: 2024-09-06

Taylor tweeted:

"I'm excited to announce that Laravel has raised a $57M Series A in partnership with Accel."

"I believe that Laravel is the most productive way to build full-stack web applications, and Laravel Cloud will be the platform for shipping those applications that this community deserves."

Partnering with Accel has allowed Laravel to start building a world-class engineering and leadership team that has resulted in Laravel Cloud and also another product that will be announced at Laracon AU in November.

He also follows up with some important takeaways from the announcement.

First, I’m not going anywhere. I’m still leading Laravel as CEO, and I’m working closely with all of our teams to ensure we’re building the best products possible for our community.

Second, we remain committed to open source. Since partnering with Accel, we’ve hired additional engineering support dedicated to open source development. I also remain the primary curator of all features in the Laravel framework. Inertia 2.0 and our first-party VS Code extension are a direct result of our increased open source engineering capacity. Accel deeply understands open source and developer focused tooling, with previous investments in Sentry, Vercel, and Linear.

### just earned laravel architect badge on laracasts URL: https://muhammadhuzaifa.pro/just-earned-laravel-architect-badge-on-laracasts Published: 2024-09-04 Last updated: 2024-09-06

Learn Laravel Architect Path

 

Just completed the recommended pathway through Laracasts and earned a new Path-specific achievement badge named Laravel Architect. The pathway conists of multiple levels, each level having multiple series, and then each series have its own episodes. It's an honor to learn from Jeffrey Way, James Brooks, Stephen Rees-Carter, Jonathan Reinink, Kevin McKee, Mohamed Said, and Jeremy McPeak.

### unveiling laravel telescope guzzle watcher & laravel pulse guzzle recorder: a journey into efficient api monitoring at laravel live pakistan 2024 URL: https://muhammadhuzaifa.pro/unveiling-laravel-telescope-guzzle-watcher-laravel-pulse-guzzle-recorder-a-journey-into-efficient-api-monitoring-at-laravel-live-pakistan-2024 Published: 2024-02-26 Last updated: 2024-08-20

As the vibrant tech community of Pakistan gears up for yet another exhilarating event, Laravel Live Pakistan 2024 promises to be a melting pot of innovation, collaboration, and groundbreaking revelations. Among the distinguished speakers, I am thrilled to announce my participation in this esteemed gathering, where I will be unveiling two groundbreaking packages: Laravel Telescope Guzzle Watcher and Laravel Pulse Guzzle Recorder.

For developers immersed in the Laravel ecosystem, the significance of efficient API monitoring cannot be overstated. With the increasing complexity of modern web applications, seamless integration with external services and APIs has become the norm. However, ensuring the reliability and performance of these integrations poses a considerable challenge. This is precisely where Laravel Telescope Guzzle Watcher and Laravel Pulse Guzzle Recorder step in to revolutionize the landscape of API monitoring.

Laravel Telescope Guzzle Watcher: Elevating Debugging to New Heights

Imagine having a comprehensive tool at your disposal that offers unparalleled insights into the intricacies of HTTP requests within your Laravel application. Laravel Telescope Guzzle Watcher does exactly that and more. By seamlessly integrating with Laravel Telescope, this package provides developers with real-time visibility into Guzzle HTTP client requests, empowering them to diagnose and troubleshoot issues with unprecedented precision.

With Laravel Telescope Guzzle Watcher, gone are the days of tedious manual debugging. Say hello to a streamlined workflow where every HTTP request is meticulously monitored, allowing developers to identify bottlenecks, track performance metrics, and optimize their applications for peak efficiency. Whether you're building a RESTful API or consuming third-party services, this indispensable tool will revolutionize your development experience.

Laravel Pulse Guzzle Recorder: Capturing the Essence of API Interactions

In the dynamic realm of web development, capturing and analyzing API interactions is crucial for ensuring the seamless functioning of your applications. Laravel Pulse Guzzle Recorder emerges as the ultimate solution for recording and replaying Guzzle HTTP client requests with unparalleled ease and precision. By seamlessly integrating with Laravel Pulse, this package empowers developers to capture the essence of API interactions, facilitating thorough testing, debugging, and performance optimization.

With Laravel Pulse Guzzle Recorder, developers gain the ability to record HTTP requests and responses, simulate various scenarios, and analyze the behavior of their applications under different conditions. Whether you're testing edge cases, replicating production issues, or fine-tuning your API integrations, this versatile tool will prove indispensable in your arsenal.

Join Me at Laravel Live Pakistan 2024

As I prepare to take the stage at Laravel Live Pakistan 2024, I am filled with excitement and anticipation to share these groundbreaking packages with fellow developers, enthusiasts, and industry leaders. Together, let us embark on a journey into the realm of efficient API monitoring, where innovation knows no bounds and collaboration fuels progress.

I invite you to join me at this momentous event as we explore the limitless possibilities offered by Laravel Telescope Guzzle Watcher and Laravel Pulse Guzzle Recorder. Together, let us empower the Laravel community to elevate their development workflows, unlock new insights, and embrace the future of web application monitoring.

See you at Laravel Live Pakistan 2024, where the journey begins anew!

### 🎉 exciting announcement: laravel live pakistan is coming to lahore on saturday, 2nd march 2024! URL: https://muhammadhuzaifa.pro/exciting-announcement-laravel-live-pakistan-is-coming-to-lahore-on-saturday-2nd-march-2024 Published: 2024-02-15 Last updated: 2024-08-20

🎉 Exciting Announcement: Laravel Live Pakistan is coming to Lahore on Saturday, 2nd March 2024! 🎉

Join us for an immersive experience that goes beyond code, into the very heart of the Laravel ecosystem. This event represents the pinnacle of innovation and communal learning in the tech world.

📍 Venue: Lahore, Pakistan - a city that blends rich heritage with a vibrant tech scene.

📅 Date: 2nd March 2024

🌟 Experience: Engaging talks on Laravel's latest trends, delicious lunch, and unparalleled networking opportunities.

This is not just another tech event; it's a gateway to mastering the art of Laravel, shared by enthusiasts and experts alike. And we're making it as inclusive as possible:

✅ Free registration

🔍 Registration requires approval for a curated experience

🍲 Complimentary lunch provided

🔗 Register now to secure your spot: https://lnkd.in/dPuyjTkQ

Whether you're looking to brush up on your skills, learn about the newest Laravel features, or just meet others who share your passion for development, #LaravelLivePakistan2024 is where you need to be.

### 🚀 laravel telescope guzzle watcher: a top 9 package of 2023 with 80k+ downloads! URL: https://muhammadhuzaifa.pro/laravel-telescope-guzzle-watcher-a-top-9-package-of-2023-with-80k-downloads Published: 2024-01-08 Last updated: 2024-08-20

In the dynamic realm of Laravel packages, one gem has soared to the 9th spot, capturing the spotlight with an impressive 80,000+ downloads. 🌟 Introducing the Laravel Telescope Guzzle Watcher, a powerful open-source tool that has left its mark on the Laravel community.

🔍 What is Laravel Telescope Guzzle Watcher? Designed to seamlessly integrate with Laravel Telescope, this package adds a new layer of visibility to your HTTP requests. Now, you can effortlessly monitor Guzzle HTTP client requests right from the Telescope dashboard. 🕵️‍♂️

🚨 Why it Stands Out:

  • Simplicity at its Core: With a user-friendly setup, the Guzzle Watcher ensures a hassle-free experience for developers.
  • Real-time Monitoring: Dive deep into your application's HTTP interactions in real-time, aiding in swift issue identification and resolution.
  • Telescope Integration: Harness the power of Laravel Telescope to visualize Guzzle HTTP client requests with ease.

📈 Impressive Stats:

  • Top 9 Ranking: Securing the 9th position in the list of top packages released in 2023 is no small feat.
  • 80,000+ Downloads: The community has spoken, and the Guzzle Watcher has been embraced by thousands of developers globally.

🛠️ Installation and Setup: Getting started is a breeze! Follow the straightforward installation guide provided in the GitHub repository, and you'll be up and running in no time.

👏 Community Love: The Laravel Telescope Guzzle Watcher owes its success to the vibrant Laravel community. Your feedback, contributions, and support have propelled this package to new heights.

🌐 Where to Find It: Explore the package on GitHub: Telescope Guzzle Watcher

In Conclusion: Laravel Telescope Guzzle Watcher has proven itself as an invaluable addition to Laravel developers' toolkits, earning its place among the top releases of 2023. With its growing popularity and commitment to simplicity, it's a must-have for anyone looking to enhance their HTTP request monitoring capabilities.

🎉 Join the Guzzle Watcher community and elevate your Laravel development experience today!

### embracing laravel and vue.js with inertia.js: my journey from wordpress to a modern monolith architecture URL: https://muhammadhuzaifa.pro/embracing-laravel-and-vuejs-with-inertiajs-my-journey-from-wordpress-to-a-modern-monolith-architecture Published: 2024-01-05 Last updated: 2024-08-20

In the pursuit of a seamless and efficient blogging experience, I embarked on a journey to create my own blog. Initially opting for the convenience of WordPress, I found myself entangled in the limitations and the time-consuming aspects of customization. Fueled by the desire to streamline my blogging process amidst a hectic schedule of work and academics, I decided to explore a more modern approach. This article marks a significant milestone as the first post on my blog, now built using Laravel and Vue.js with Inertia.js, embracing the power of a modern monolith architecture.

The WordPress Challenge:

While WordPress undoubtedly offers a user-friendly environment for blogging, I soon realized that the constraints of pre-designed themes and plugins were hindering my ability to create a truly unique and personalized platform. The constant need for updates and security checks added an extra layer of complexity, diverting my focus from content creation to maintenance.

The Laravel and Vue.js Paradigm Shift:


Seeking a more robust and customizable solution, I turned to Laravel, a PHP web application framework known for its elegant syntax and developer-friendly features. Coupled with Vue.js, a progressive JavaScript framework for building user interfaces, I found the perfect combination to bring my vision to life.

Inertia.js: The Glue Between Laravel and Vue.js:


Inertia.js emerged as a game-changer in my development journey, providing a modern monolith architecture that seamlessly connects Laravel and Vue.js. With Inertia.js, I could build dynamic, single-page applications without sacrificing the simplicity and elegance of server-driven page loads. This eliminated the need for constant API calls and improved the overall performance of my blog.

Advantages of the Modern Monolith Architecture:

  1.  Efficiency: Inertia.js reduced the complexity of managing separate frontend and backend codebases, enabling me to write less code and achieve more functionality.
  2. Seamless Integration: The tight integration between Laravel and Vue.js allowed for a smooth development process, promoting code reusability and maintainability.
  3. Improved Performance: With Inertia.js handling the communication between the server and client, my blog now boasts faster page loads and a more responsive user experience.
  4. Enhanced Customization: Laravel's Blade templating engine and Vue.js components empowered me to design and customize every aspect of my blog to align with my personal style.

Conclusion:


Transitioning from WordPress to a modern monolith architecture built on Laravel, Vue.js, and Inertia.js has been a transformative experience. This migration has not only freed up my time from routine maintenance but also allowed me to fully express my creativity in crafting a unique and personalized blogging platform. As I embark on this new chapter, I look forward to sharing more insights and content through my blog, now powered by the dynamic synergy of Laravel and Vue.js.