Press Space to continue
Longer context windows increase compounding errors in agent runs
Press Space to continue
Trying to put my finger on how the more context you use in a context window (I called it context depth), over longer runs, the more errors can compound / the agent can drift. Aka if you want more reliablity, use less context + have shorter runs A quick sketch:
open on x · 2 viewsViews and clicks count once per visitor per day. Clicks are opens of this post on X.