Discussion about this post

User's avatar
Neural Foundry's avatar

Brilliant take on keeping agent workflows clean. The tmp/ approach captures something critical about how LLMs work best when they can externalize theirreasoning into persistent artifacts. I've found that having these intermediate plans visible actually changes how I prompt since I can reference specific sections rather than rebuilding context each time. One thing worth mentioning is that tmp/ also becomes useful for debugging when agents go off track.

Expand full comment

No posts

Ready for more?