Hugging Face Daily PapersYi Duan, Ying Liu, Zirui TangSep 10, 20261 min readpaperintermediate

The Last AI Built by Humans: Toward Genuine Recursive Self-Improvement

Summary

The paper defines recursive self‑improvement (RSI) for AI, introduces the Headroom‑Closed Index to expose current LLM limits, and proposes a staged roadmap toward full RSI. It surveys application scenarios and outlines practical challenges.

RSI is framed as four autonomy levels ending in recursive meta‑improvement.
Headroom‑Closed Index (HCI) quantifies the gap between model capacity and usable performance.
Different domains (science, embodied AI, software engineering) need distinct RSI requirements and timelines.
Key challenges include safe experience acquisition, strategy generation, and environment adaptation.

AI researchers and safety engineers should care because achieving genuine RSI changes how future AI systems evolve and begets new safety considerations.

5/10

llm machine learning ml infra

Related reading

RRSI: Regularized Recursive Self-Improvement of Agent Harnesses arxiv.org
RRSI adds budget‑annealed edit limits, trajectory exploration incentives, a benchmark‑aware critic, and a pruning step to recursive self‑improvement of LLM agent harnesses, yielding up to +14.1 points on in‑distribution tasks and +4.7 on OOD benchmarks while cutting policy token usage by 30%.
Hugging Face Daily Papersarxiv.orgPeng Xia, Rujun Han, Zifeng Wang1d1 minpaper
Generalized Agent Iteration: One Formal Framework for Iterative Policy Improvement and Recursive Self-Improvement arxiv.org
The paper introduces Generalized Agent Iteration (GAI), a formal framework that unifies classical iterative policy improvement (GPI) and recursive self‑improvement (RSI). GAI treats an agent as a set of modifiable components and models learning as a loop of evaluation and improvement. Two binary “dials”—whether the improvement mechanism is internal to the agent and whether the evaluation standard…
Hugging Face Daily Papersarxiv.orgHongyao Tang, Yi Ma, Pengyi Li11d1 minpaper
ModularRSI: Modular and Generalizable Recursive Harness Self-Improvement arxiv.org
ModularRSI proposes a contrastive, modular approach to evolve agent harnesses by independently updating five functional modules and aggregating evidence across 2,000 benchmark‑disjoint tasks. Experiments on TB2.0 and SWE‑Bench Verified show consistent gains on unseen tasks and cross‑model transfer.
Hugging Face Daily Papersarxiv.orgSiwei Wu, Jincheng Ren, Yizhi Li8d1 minpaper
Designer-RSI: Evolving Procedural Memory from User Traffic for Agentic Graphic Design arxiv.org
The paper introduces Designer‑RSI, a continual‑adaptation system that couples a frozen design‑software‑controlling LLM with an external procedural memory of natural‑language design skills. Over five adaptation rounds on real user briefs, the memory grows from 76 to 139 procedures and lifts execution success from 72.7% to 99.3%, showing that skill accumulation and selective replay can dramatically…
Hugging Face Daily Papersarxiv.orgHongyang Du, Lan Yan, Christian Flores4d1 minpaper
ScienceBuddy: Recursive-in-Recursive Self-Improvement for Interactive Scientific Agents arxiv.org
ScienceBuddy is an interactive workspace that converts researcher prompts, feedback, and execution traces into continual‑learning tasks for AI agents. It introduces a "recursive‑in‑recursive" self‑improvement loop that alternates harness refinement and model training, and showcases case studies across four scientific task families.
Hugging Face Daily Papersarxiv.orgShuhan Xue, Jianyuan Zhong, Ziyuan Nan7d1 minpaperHN2
Agentic Skill Decay addyosmani.com
Addy Osmani warns that AI agents can short‑circuit the hands‑on practice (“reps”) that builds deep expertise and judgment. He recommends deliberately inserting hypothesis‑forming, “why” questioning, diff inspection, failure prediction, and occasional manual coding into the workflow, especially for junior engineers. A 2026 Anthropic study showed junior developers using AI scored 17 % lower on a fo…
Addy Osmaniaddyosmani.com22d16 min

The Last AI Built by Humans: Toward Genuine Recursive Self-Improvement

Summary

Related reading

RRSI: Regularized Recursive Self-Improvement of Agent Harnessesarxiv.org

Generalized Agent Iteration: One Formal Framework for Iterative Policy Improvement and Recursive Self-Improvementarxiv.org

ModularRSI: Modular and Generalizable Recursive Harness Self-Improvementarxiv.org

Designer-RSI: Evolving Procedural Memory from User Traffic for Agentic Graphic Designarxiv.org

ScienceBuddy: Recursive-in-Recursive Self-Improvement for Interactive Scientific Agentsarxiv.org

Agentic Skill Decayaddyosmani.com

RRSI: Regularized Recursive Self-Improvement of Agent Harnesses arxiv.org

Generalized Agent Iteration: One Formal Framework for Iterative Policy Improvement and Recursive Self-Improvement arxiv.org

ModularRSI: Modular and Generalizable Recursive Harness Self-Improvement arxiv.org

Designer-RSI: Evolving Procedural Memory from User Traffic for Agentic Graphic Design arxiv.org

ScienceBuddy: Recursive-in-Recursive Self-Improvement for Interactive Scientific Agents arxiv.org

Agentic Skill Decay addyosmani.com