Self-improving AI / Harness²
Harness²
“we introduce Harness 2, a plug-and-play framework for recursive agent harnessing in an open world.”
Harness² is one of the projects on the self-improving AI map. Below is what it changes about itself, whether the change lasts, and whether it improves the improver, each with the line from its own source.
Where it sits
Columns: what it changes, from the surface to the deepest change. Rows: whether the change outlasts the run. A number is how many projects sit in that square. See it on the 3D map
- What it changes
- Tools, harness, its own code“optimizing the runtime scaffolding around a frozen base model”
- Does the change last
- Saved for later runs“a library transferred from JobBench improves performance on WorkBuddy-Bench”
- Does it improve the improver
- improves the task agent; no evidence the improver got better“Model weights, stage instructions, and the decision rule remain fixed, while the library evolves.”
- Who sets the goals
- people set tasks and scoring“Grading is the benchmark's composite verifier”
- How it was tested
- tested on held-out tasks“Harness 2 and the other baselines never use them: they process the test tasks as a stream, starting from an empty library or memory.”
Date is arXiv v1. Eight of the nine authors are at Google Cloud AI Research; the first author did the work as an intern there. The code is in the google-research GitHub organization. The paper calls its second loop harnessing-level recursion, but what it updates is a library of past harnesses and lessons that a fixed improver reads; the editor and its instructions are not changed. Model weights stay fixed. In the main runs the harness editor is Claude Opus 4.8, a different model from the executors. WorkBuddy-Bench is graded by hidden tests plus a Gemini 3.5 Flash judge; JobBench and LAB are judged by Claude Opus 4.8 and Claude Sonnet 4.6 against rubrics, and the JobBench judge is the same model as the editor.
Does it count as RSI?
It depends on the definition. Under the four used on What counts as RSI, by what its own source says:
- Loose (Eliezer Yudkowsky, Intelligence Explosion Microeconomics (2013)): counts
- Durable (Ren et al., self-improvement survey (2026)): counts
- Strict (Roman Yampolskiy, From Seed AI to Technological Singularity (2015)): does not count
- Autonomous (Anthropic, When AI builds itself): does not count
How it got here
Earlier work in this catalog that Harness² cites, oldest first. Under each: where that work sits on the map, then where Harness² sits (changes in bold), and what Harness²'s own paper says about it. Its whole family on the citation map
2023 Self-Refine builds on it
Prompts, memory → Tools, own code · gone when the run ends → saved for later runs · does not improve the improver → improves the task agent
“Sequential recursion applies iterative self-refinement (Madaan et al., 2023) to the execution harness.”
2026 Recursive Harness Self-Improvement sets itself apart
Prompts, memory → Tools, own code · saved for later runs · improves the task agent
“Rather than only revising an execution harness under a fixed improvement procedure (Lee et al., 2026a,b), we also update the harness guiding future revisions:”
2026 Harness²
Tools, own code · saved for later runs · improves the task agent
Also cites Reflexion 2023, STOP 2023, Self-Rewarding Language Models 2024, ADAS 2024, Darwin Gödel Machine 2025, Hyperagents 2026
Cited by Self-HPO 2026
In the same square
AutoTavern, RRSI (Regularized Recursive Self-Improvement), RSIV-Harness, Voyager