Articles

Desk comparison of Cursor and GitHub Copilot on autocomplete, chat, codebase indexing, pricing, and when Copilot is still the better fit.

Desk comparison of ChatGPT, Claude, and Gemini for code generation, debugging, context windows, docs, and API cost — not an IDE bake-off.

Desk comparison of Cursor Composer, Claude Code, Copilot Workspace, Aider, and Amazon Q for coordinated multi-file edits.

Desk notes on ten genuinely usable free-tier coding tools — and what each one is bad at.

Desk notes on Ollama, Continue, Tabby, and open coding models you can run on your own GPU — privacy, VRAM, and when cloud is easier.

Desk notes on Roo Code custom modes: tool-group ACLs, when not to enable execute, and how the listing differs from Cursor Composer or Cline defaults.

Desk workflow for AI-assisted legacy refactoring: understand first, add tests, then incremental multi-file edits — not a wholesale rewrite.

Desk notes on Claude, Cursor, Copilot, Mintlify, and Swimm for API docs, READMEs, and documentation that goes stale.

Desk notes on Claude, Copilot Chat, Cursor, Amazon Q, and Gemini for stack traces, multi-file bugs, and cloud-platform debugging.

Desk notes on what actually changed for developers in 2026: editors vs models vs agents vs hosts, and what to check before you switch stacks.

Install Ollama or llama.cpp, size VRAM for 8B–70B models, and wire a local coding stack into VS Code or Cursor.

Desk notes on Cursor plus Ollama: HTTPS tunnel, what still stays in Cursor cloud, and when to skip a GPU buy.

Run DeepSeek-R1 distilled models with Ollama, size VRAM, and point Cursor or VS Code at the local API — not a localhost Cursor hop.

VRAM tables and desk buying notes for local LLM inference: 8B–70B sizing, NVIDIA vs AMD vs Apple Silicon, and current-card price searches.

Desk notes on MacBook Pro, ThinkPad X1, Framework 13, Zephyrus G14, and MacBook Air — memory, Linux, and local-AI fit, not listing copy.

Desk notes on client-tool agent loops: allowlists, iteration caps, and when a framework invoice is the wrong buy.

Step-by-step guide to getting started with the Anthropic Claude API. From API key to your first AI-powered app in 15 minutes.

Move beyond basic prompting. These 10 research-backed techniques improve AI output quality for coding, writing, and analysis.

Desk notes on how AI changed team workflows: review load, junior learning, and what to check before you mandate a seat.

Desk notes on AI pair-programming setups: which seat to buy for navigator vs driver vs ping-pong, and when stacking Cursor plus a chat plan is wasted.

Desk notes on when AI-generated tests lock behavior versus when they just snapshot the current implementation.

Desk notes on AI pull-request review in CI: severity contracts, and when a review SaaS is the wrong buy.

Desk notes on AI-generated code risks: injection, secrets, and what to check before you ship — or buy another scanner.

AI coding tools sometimes invent APIs that do not exist, reference wrong library versions, or produce subtly incorrect logic. Here is how to catch and prevent it.

Move past any and basic types. Advanced TypeScript patterns for type safety, maintainability, and developer experience.

Trunk-based development, feature branches, conventional commits, and PR templates. A practical git workflow for teams of 2-10.

Learn how to calculate t-calculated (tcale), compare t-calculated vs t-critical, and build a fully functional T-Test Calculator app using AI coding assistants.

Desk notes on pgvector, Chroma, Weaviate, Qdrant, and Pinecone — and when a new SaaS is the wrong buy.

Protect your API from abuse and ensure fair usage. Token bucket, sliding window, and distributed rate limiting patterns.

Desk comparison of Vercel, Cloudflare Pages, Netlify, DigitalOcean, Railway, and Hetzner for git-push sites versus real servers.

Desk notes on GitHub Actions pipelines: fail-cheap order, host previews, and when a second CI product is the wrong buy.