We're investigating whether running more coding agents is actually better for engineering outcomes.
Our area of interest is not so much about how much code agents generate, more about the human attention required in steering, handling interruptions, reviewing and fixing PRs etc.
How many agents in parallel is your optimum before you start to get merge issues?
Full disclosure: I'm building Ascenda Flow which measures these patterns locally. We're also gathering a benchmark study.
YC companies and AI-first are particularly interesting as I've seen teams of 2-4 engineers out of YC building faster than much larger teams.
marisolchallen•16m ago
Our area of interest is not so much about how much code agents generate, more about the human attention required in steering, handling interruptions, reviewing and fixing PRs etc.
How many agents in parallel is your optimum before you start to get merge issues?
Full disclosure: I'm building Ascenda Flow which measures these patterns locally. We're also gathering a benchmark study.
YC companies and AI-first are particularly interesting as I've seen teams of 2-4 engineers out of YC building faster than much larger teams.