Tutorials, deep-dives and field notes on React, TypeScript, Sitecore and
the platform features that quietly change how I build. Updated when
something earns the time.
A hands-on evaluation of PostHog's AI setup wizard on a real Astro site with View Transitions. One command produced a working integration, eight custom events across six files, a pre-built dashboard, and — unexpectedly — an agent skill for future AI-assisted work. And the snippet it installed is the on-ramp to the rest of the platform: feature flags, experiments, session replay, and surveys on the same event stream.
A hands-on evaluation of OpenAI's Codex Security against a traditional SAST baseline. On the same codebases, an LLM-driven scanner surfaced whole classes of vulnerabilities our rule-based tooling was structurally blind to — committed credentials, stored XSS, and a CI/CD takeover path. Here's what it found, where it falls short, and how to operationalize it.
An evaluation of Claude Code, Anthropic's terminal-first agentic coding tool. Beyond autocomplete: it plans, edits across files, runs tests, and handles git — with the enterprise governance hooks (permissions, OpenTelemetry, admin analytics) that decide whether you can roll it out to a team.
A headless analytics component that tracks page views for Sitecore Search. It captures user context, generates canonical page identifiers, and publishes entity view events to the Sitecore Discovery API without rendering any visible UI.
A deep dive into how the ResourceListWidget manages filter strategies from load and leverages Sitecore Search SDK for faceted filtering and dynamic search results. This document explains the architecture and data flow of the filtering mechanism.
An evaluation of Jules, Google's experimental AI coding agent. It runs asynchronously in a cloud VM, takes GitHub issues, plans, writes code, and opens PRs — powered by Gemini. Strong on routine work and unusually transparent, limited on large codebases and tight beta quotas.
An evaluation of running Vercel's v0 model directly inside Cursor via its OpenAI-compatible endpoint. The setup is trivial and the React/Next output is clean — but at roughly $0.80 per prompt, the economics, not the quality, decide whether it's worth it.
An evaluation of Figma Sites — publish responsive websites straight from Figma, no handoff. A dream for designers and static content, but bloated non-semantic code, no Git/CI, and missing CMS features keep it out of engineering-led production work.
A hands-on evaluation of Astro — the islands-architecture framework that ships zero JavaScript by default. Where it wins almost for free, the boundary that defines when it's the wrong tool, and what to validate before trusting it on enterprise work.
An evaluation of Supabase — the open-source backend-as-a-service built on PostgreSQL. Auth, real-time, storage and edge functions out of the box, with self-hosting for data ownership. Where it fits, where the free tier bites, and why it quietly became the backend behind a wave of AI app builders.
An evaluation of Lovable, the AI tool that builds full web apps from natural-language prompts. A first-prompt result that genuinely impressed, a slick Supabase integration, and the message-limit and code-ownership realities that decide whether it fits complex, integration-heavy work.
An evaluation of the browser Speculation Rules API for prefetching and prerendering. How near-instant page loads work, the eagerness levels that govern the cost, the bandwidth-CPU-server tax of guessing wrong, and why the right call was to hold.
An evaluation of Vercel's v0.dev — a chat interface that turns prompts and design screenshots into React/Next.js components. How feeding it images changes everything, why it's a Next.js story, and the pairing that turns 'generated' into 'production-ready.'
An evaluation of Cursor, the AI-first code editor built on VS Code. How its codebase-aware chat and model choice compare to GitHub Copilot, the privacy trade-off hiding in Privacy Mode, and whether it's worth replacing your entire editor.
Understanding concurrent features, automatic batching, and how to use Suspense for better user experiences. A technical deep dive into the internal mechanics of the new React rendering engine.
An evaluation of the Vercel AI SDK — an open-source library for building streaming, conversational AI UIs in JavaScript/TypeScript across React, Svelte, and Vue. Provider-agnostic model adapters, streaming-first UX, and generative UI that lets the frontend own AI integration.
An evaluation of Lighthouse CI — running Lighthouse audits automatically on every change to guard performance, accessibility, SEO, and best practices. Shareable HTML reports, headless-Chrome runs in the pipeline, and the one manual step that resisted automation.
An evaluation of Chromatic, the visual testing addon for Storybook. Snapshot-based regression detection with a one-click, no-CI-required workflow, faster than Percy in testing — bounded by its tight coupling to Storybook.
An evaluation of Radix UI — low-level, unstyled, WAI-ARIA-compliant React primitives that hand you accessibility and behavior while leaving design entirely to you. Why it fits a React/Next stack, the per-component install model, and the styling work you sign up for.
An evaluation of Durable, the AI website builder that generates a site from three questions in about 30 seconds. Genuinely fast — but a customization ceiling that makes it a mismatch for an agency built on bespoke work. The one exploration that ended in a clear 'stop.'
An evaluation of Galileo AI, a prompt-based UI generation tool for fast design ideation across web and mobile. Strong for inspiration and rapid prototyping, but unpredictability and over-generation make it a design-phase accelerator, not a delivery tool.
Server Actions allow you to build efficient data mutation without the need for API routes. Here's a comprehensive look at how I integrated them into my latest e-commerce project to reduce client-side JavaScript.
Accessibility shouldn't be an afterthought. In this guide, I break down key WCAG guidelines and provide practical implementation tips for building inclusive semantic HTML components that work for everyone.
The journey of refactoring a large legacy codebase from Styled Components to Tailwind CSS. I discuss the performance benefits observed, the development velocity improvements, and how utility-first CSS changed my workflow.