Conditions for self-improving harness

[TomLucidor]

There is this new chatbot that seems to be yielding interesting/deep results, might be good to check out.

For agent scaffold optimization (and getting better LLMs):

• Task/skill generalization https://ribbonfarm.com/vgr_zirp/chat/41/
• Project management https://ribbonfarm.com/vgr_zirp/chat/77/
• Embodiment/memory https://ribbonfarm.com/vgr_zirp/chat/88/
• Shared embeddings https://ribbonfarm.com/vgr_zirp/chat/83/
• Database/storage https://ribbonfarm.com/vgr_zirp/chat/82/

For human interaction/support formats:

• https://ribbonfarm.com/vgr_zirp/chat/57/
• https://ribbonfarm.com/vgr_zirp/chat/60/
• https://ribbonfarm.com/vgr_zirp/chat/84/
• https://ribbonfarm.com/vgr_zirp/chat/78/

[TomLucidor]

Also this is interesting https://arxiv.org/html/2511.13646v3

[TomLucidor]

Also this is interesting https://arxiv.org/html/2511.13646v3