Posts tagged “ai”

Watch your coding agents the way you watch your application logs: look for repeated work, hidden friction, and patterns you can turn into better tools.

LazyPi annoyed minimalists because it packages Pi with opinions, but that same convenience may help newcomers see the point faster.

Claude Code’s `/insights` command is not magic; it is a pipeline of session filtering, metadata extraction, LLM analysis, and HTML reporting.

Anthropic’s subscription rules are confusing enough that developers need a map of where Claude can and cannot be used.

Fizzy’s 8,152 commits read like a product documentary, from Splats on a windshield to a finished 37signals issue tracker.

A tiny Claude Code skill makes Mermaid diagrams prove they render before the agent calls the docs done.

Sonic felt like Sonnet at triple speed, with a huge context window and enough Rails competence to demand attention.

A day on Opus and a day on Sonnet suggested the cheaper model may be good enough for most hands-on Claude Code work.

Claude Code’s in-session survey interruption is a reminder that paid developer tools should ask for feedback without hijacking flow.

A $200 coding assistant only has to save two developer hours a month before the economics get awkward for skeptics.

Context7 gives coding agents fresh, version-specific docs before they hallucinate yesterday’s API into your app.

Let the AI crows eat the CRUD work so humans can chase the harder, stranger, more valuable problems.

A custom Claude Code `/quiz` command turns each coding session into a quick check that you actually learned something.

A clean codebase turns AI agents into multipliers; a tangled one turns them into very fast guessers.

The danger is not using AI; it is using AI with so little friction that your brain quietly leaves the room.

AI workflow experiments deserve curiosity before dunking, because today’s weird loop might be tomorrow’s normal development practice.