Timely reading for anyone building agent harnesses in domain specific / OOD problem spaces
from @DrScottClark
from @DrScottClark
Aug 4, 2026 · 9:13 PM UTC
2
2
4
190
@itsyourcode @DrScottClark yeah, not my experience. Models know how to handle and keep the context in one session. so the stuff out of distribution that gets created by your imagination and the stuff that through enhancement inside the knowledge cutoff accumulates fast.
1
0
1
21
@tryingET @DrScottClark that's the harness doing that though, not the model :)
1
0
2
21


