@sd141414: Spent the afternoon studying why my AI agent performs so well on one specific repo. Three things, none of them the model. A doc that explains why, not just what. An issue tracker that gets read, not just written to. Nothing trusted until it's checked against a clean baseline. Curious where your own setup stands - Systems Thinking Quiz, link in bio.
and logs everywhere, raw and abstracted version (grafana). much easier for an agent when it has: codebase, adrs, and actual logs of what’s happening in production
2026-09-14 23:12:33
2
XEQUALS :
harness > model
2026-09-14 19:23:39
4
kuhlsnu :
So what happened when you added these features to your global claude file?
2026-09-15 17:05:14
1
ChroMikey :
Week 3: Governance is where I found the real challenges. The only answer I found was to put supervisor and foreman roles directly on my desktop so they are on top of everything else
2026-09-15 02:45:10
2
Titi998 :
I created topology.py and systematic adverse agents for validation
2026-09-15 17:34:56
1
To see more videos from user @sd141414, please go to the Tikwm
homepage.