About

Ank's notes on working with AI. No news roundups, no hype — only things I've actually run myself: reproducible, with the potholes marked.

How I write here

Only what I've used. Every post is backed by either a specific experiment or a workflow I use daily. If I haven't run it, I don't write it.
Conclusion first. Result up front, reasoning after — so you can decide in two seconds whether to keep reading.
Fewer posts over filler. One usable post a week beats a daily copy-pasted “AI digest”. If one line says it, I don't stretch it into a paragraph.

What I write about

AI app development - Claude / GPT APIs, agent frameworks, RAG in practice
Prompt & context engineering - treating the model as a colleague with boundaries
Model selection & cost - what Opus, Sonnet, and Haiku are each good for
AI x workflows - automating the everyday with Claude Code, Cursor, and scripts
Reading & notes - occasionally looking up from the tools to ask where all this is heading

Finally

I use every tool mentioned here, and I write up the potholes as I hit them. Take what's useful; if you see it differently, leave a message or just reach out.