Split JSX transpiler into libyggdrasil + thin CLI with C API for embedding
Takeaway: Next time I'd flag up front that any CLI flags removed during a refactor need a compatibility shim, since removing --recover silently broke one test fixture.
12 sessions on Bash with a coding agent: the agent failed at least once in 9, and 3 finished without a human stepping in.
Takeaway: Next time I'd flag up front that any CLI flags removed during a refactor need a compatibility shim, since removing --recover silently broke one test fixture.
Takeaway: Next time I'd put the full bokeh-fix-plus-breathing-room task in one prompt from the start instead of splitting it into scout/backup/implement stages — the orchestrator ended up doing that anyway.
Takeaway: Next time I'd start by dumping the raw config files myself instead of pasting agent transcripts and letting Claude re-diagnose from scratch — it wasted a lot of turns re-reading settings.json across live and repo copies.
Takeaway: Next time I'd tell it upfront to just use agy for background workers since pi -p silently hangs without doing real work — that cost a good chunk of the session polling dead processes.
Takeaway: I'd start by asking it to grep the installed package for HEADROOM_ env vars directly instead of trusting the hosted docs, since those turned out to be too vague to use.
Takeaway: Next time I'd specify upfront which CLI flags must stay wired through even if the C API doesn't support them yet, since dropping --recover during the refactor silently broke a fixture test and cost a round of debugging.
Takeaway: I'd start the next session by pointing straight at the handoff file and the specific list of flagged clips instead of re-deriving triage state from scratch — that reconstruction step ate a lot of time.
Takeaway: Next time I'd start with 'trace the actual config precedence and provider fallback logic, don't just read the top-level JSON' — that's what caught the real bug (searchProvider silently overriding provider).
Takeaway: Next time I'd start the audit prompt with 'compare live config to the dotfiles repo and flag drift' up front, since half the useful findings came from that diff rather than the visual review.
Takeaway: Next time I'd tell it upfront to sanity-check that pi workers are actually producing output within a minute or two, since three of them silently hung for over five minutes doing nothing while I waited.
Takeaway: Next time I'd put the cover-photo disposition check and the taste-notes review earlier in the same prompt, since the disposition bug (attached_pic=0) only got caught because I forced an ffprobe verification step right after muxing — that's the correction that actually mattered here.
Takeaway: I'd put the exact three taste notes and the agy chunking method in the first prompt so the orchestrator doesn't have to rediscover conventions from the handoff doc every round.