Subagents Conwidered Harmful

bob1029

Sub agents make the most sense at pre existing boundaries that already have a high impedance mismatch for humans. Without sub agents I would be able to test maybe two product scenarios before the main context needs compacting. With sub agents I can go 40-50 scenarios wide and easily compare their results in a single turn of the main agent loop. The loss in fidelity is a feature, not a bug in this context. I don't want the main agent to "see" literally every logical view that was experienced. I only want it to receive a testing summary with respect to the original instructions and experience log. Anything beyond this is too much, even for a human.

amoss

Are we doing typos in headlines now?

simianwords

> A fork costs more tokens, but it knows exactly what’s been asked for, what’s been corrected and which approaches have already been thrown out But a fork is prompt-cache friendly so it might end up costing lesser

gijsbos

Isn't the root cause that the summaries aren't accurate enough?

xrd

GitHub pioneered the idea of a holocracy (flat hierarchy). https://en.wikipedia.org/wiki/Holacracy It's a good idea because, as this article points out, the more layers of delegation, the more likely the original vision gets lost. However, and it's just my opinion, but this was how the Julie Horvath incident happened. No one was checking in with the people. https://www.theregister.com/on-prem/2014/04/22/github-cofoun... It feels like you could do holocracy well (keep things flat) if your subagents don't sexually harass other agents. I would avoid grok for this reason.