AI makes AI better.

7 picked from 113 candidates · 31 sources read · 69 loops indexed

1. 01
   Loop of the dayPaperarXiv22 Oct 2026Tools and harness

   Recursive Self-Improvement on a Budget: Scoring Code Edits Before They Run

   The loop: A coding agent proposes and scores edits to its own scaffold, which improves its solve rate. This higher solve rate then guides the selection of the next scaffold edit.

   By scoring proposed edits before they run, this agent drastically reduces the evaluation budget required for open-ended self-modification.

   loop fit 10/10Erfan Darzi, Sree Bhargavi Balija, Sheng Hevia arXiv
   
2. 02
   PaperarXiv21 Oct 2026Architecture and optimizer search

   Recursive Code Evolution for Task-Oriented Semantic Communications

   The loop: A large language model modifies its own communication program and training policy. The accepted code and textual feedback then inform the model's subsequent code proposals.

   loop fit 9/10Yan Gong, Zheng Chu et al.via arXiv
   
3. 03
   ProjectGitHub19 Oct 2026Self-modifying agents

   quqxui/RSI-Flow

   The loop: A recursive self-improvement workflow evolves target agents and its own system architecture. This improves the capability of the system to perform further evolution on itself and other agents.

   loop fit 7/10quqxuivia GitHub
   
4. 04
   Forum postLessWrong21 Oct 2026Interpretability and oversight

   Even GPT-2 passes most functional global workspace tests

   The loop: An automated interpretability tool uses the Jacobian lens to evaluate functional global workspace representations in language models. This identifies structural deficiencies that can be addressed in the design of successor architectures.

   loop fit 6/10Finn Cairnsvia LessWrong
   
5. 05
   Blog postSakana AI21 Oct 2026Tools and harness

   Sakana AI Engineer Open House 2026

   The loop: An orchestration model named Fugu dynamically switches between multiple models to manage geopolitical risks and cultural alignment. This improves the reliability of the overall agent system it operates within.

   loop fit 6/10Sakana AIvia Sakana AI
   
6. 06
   PaperarXiv20 Oct 2026Tools and harness

   LVMH: A Meta-Optimization Harness for Long-Video Agent Memory

   The loop: An automated multi-agent framework formulates memory-policy development as a multi-objective code search. It records past proposals, traces, and failures in a persistent meta-memory to guide its subsequent policy revisions.

   loop fit 9/10Liwei Che, Benlin Liu, Dai Li, Xiangyan Li, Zijian Wu, Changjiang Li, Minghao...via arXiv
   
7. 07
   PaperarXiv20 Oct 2026Self-reward and self-play

   Self-Evolving Agents through Privileged Retrospective Distillation

   The loop: A model acts as its own teacher by re-scoring its trajectories with retrospective context. The resulting teacher distribution is then distilled back into the same policy over multiple rounds.

   loop fit 9/10Yuhang Zhou, Yifan Wu et al.via arXiv
   
8. 08
   PaperarXiv20 Oct 2026Tools and harness

   Training the Whole Agent, Not Just the Model

   The loop: Teacher models evolve an agent's harness and tools during reinforcement learning based on its trajectories. The policy trains with this adapted harness, improving its coding performance and generating new trajectories that guide further adaptation.

   loop fit 10/10Lingxiao Tang, Wei Tao et al.via arXiv