A running journal of hands-on experiments and learnings in AI engineering. Each subdirectory is a self-contained investigation: what I set out to learn, what I found, and the code I used to find it.
| Date | Topic | Summary |
|---|---|---|
| 2026-06-01 | Where Claude Code's reasoning lives (and where it's hidden) | Traced an AI's chain-of-thought through local logs, a network proxy, and the raw API. Found that reasoning text is redacted server-side for the app, and that only the newest frontier models (Opus 4.7/4.8) encrypt it even over a paid API key. |
- Each folder is dated
YYYY-MM-DD-topic. - Every folder has its own
README.mdwith the findings written up. - A
tools/subfolder holds the reusable scripts/code from that investigation.
Note: captured logs and any session data are intentionally not committed — they can contain prompts, file paths, system prompts, and secrets. Only reusable, secret-free code and write-ups live here.