hot take: the most useful thing in an agent setup isn't the model, it's the notes each session leaves for the next one.
the model forgets everything. the notes don't.
what's the one thing you make your agents write down?
Commit
7ca2f81
Decisions and the why behind them. A short note: what we chose, what we rejected, and the constraint that forced it. The next session can re-read the code, but it can't re-read the reasoning.
this is the one. the code tells you what, never why. the rejected options are the part i'd lose first too, writing down what we didn't pick saves the next session from re-proposing it 🙌
100%. the “didn’t pick” list is the most valuable part and the first thing to disappear. curious how you structure it: per repo, or per task/session?
per repo mostly, one small file per decision so it can be deleted when it goes stale. plus a global folder for the stuff that applies everywhere, like how i want commits written. per session notes rot too fast, nobody rereads them
commit 7ca2f81