Every change my AI team landed in our new codebase waited through 8.5 minutes of tests, for two straight weeks. Our older project, one folder over, was already running the same kind of tests in parallel. Solved, shipped, working.
0 Boosts
0 Likes
Now, before my AI builds anything unfamiliar (a file format, a protocol, an integration we haven't done here), it looks around:
It searches our own stuff first, meaning the shared toolkit our projects pull from and then every project's history.
Nothing there? It compares two or three maintained open source projects that solved it, and borrows their approach.
It copies code only if the license allows.
My lead AI puts these steps in every brief to another agent, so the helpers look too.
Check the drawer.
Nobody looked over the fence, and my briefs never told anyone to. Starting fresh feels productive, and asking whether it exists feels like homework. It's working out a sauce from scratch while the family recipe sits in the kitchen drawer. (The briefs are mine, so that's mine too.)