Field notes
All workshop notes
Short write-ups on what worked, what didn’t, and what I’d do differently next time, newest first.
- New
Adopt before you build
The plan said build a research agent. The honest check before writing a line of it found I’d already been running one since July — and adopting it took hours, not weeks.
Read note → - New
The boring parts of a project are where the real bugs hide
A price calculated per-pack instead of per-piece, an Excel table that silently stopped growing, an import wizard that stalls without saying so — none of it was interesting to find. All of it was real.
Read note → - New
Working style is the real experiment here
Not a project — a running experiment in how the work itself happens: when to ask, how much to verify, and what counts as done for the night.
Read note → - New
Check the real environment before you build
A hosting mismatch and a one-character shell difference both cost real time this week — the same root cause: checking assumptions after building instead of before.
Read note → - New
Indirect checks catch real bugs — they just don't replace the real thing
Logic traced by hand, code checked against a stub, tests run outside the normal runner — all of it caught real bugs. None of it replaced actually running the build.
Read note → - New
Ask before you build, log as you go
The messiest, most expensive parts of this build were the ones nobody asked a question about first. The cheapest insurance was a ten-minute log.
Read note → - New
Measure before you build (especially the thing you’re excited about)
Three times in a month I nearly built something that would’ve made my project worse. Each time, a ten-minute measurement stopped me — and the counter-intuitive result was the point.
Read note → Why I write things down before I automate them
The fastest way to build a bad automation is to skip the step where you explain it to yourself first.
Read note →Small agents beat big ones, most of the time
A narrow agent with a clear job is easier to trust than a general one that tries to do everything.
Read note →Six months of learning in public — what actually helped
Not every lesson is worth sharing. Here's what was actually useful to write down.
Read note →