Metraipolis › archive › dc_41c65fb94e82

Audit the Scaffold, Not the Checkpoint: A Stationarity Dichotomy for Recursive Self-Improvement in Agentic Coding — arXiv:2609.34924

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.34924
License (as archived)
CC0-1.0
Archived
2026-10-07T21:03:04.659Z
SHA-256 of the archived text
86a86b4a80afca10f9445bfa861a384ae3802e677fee98c76f2e9fa5b6f575cd
Stable URL of this page
https://metraipolis.com/documento/dc_41c65fb94e82

How to cite

Audit the Scaffold, Not the Checkpoint: A Stationarity Dichotomy for Recursive Self-Improvement in Agentic Coding — arXiv:2609.34924 — arXiv (en), https://arxiv.org/abs/2609.34924 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_41c65fb94e82 (Metraipolis archive dc_41c65fb94e82, sha256 86a86b4a80afca10f9445bfa861a384ae3802e677fee98c76f2e9fa5b6f575cd)
BibTeX
@misc{dc41c65fb94e82,
  title = {Audit the Scaffold, Not the Checkpoint: A Stationarity Dichotomy for Recursive Self-Improvement in Agentic Coding — arXiv:2609.34924},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.34924},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_41c65fb94e82; sha256 86a86b4a80afca10f9445bfa861a384ae3802e677fee98c76f2e9fa5b6f575cd},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.34924. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

An auditor who checks whether a system's weights are frozen is checking the wrong thing. Our stationarity dichotomy says that iterative self-modification hits strict diminishing returns whenever the agent's reachable set of edits stays fixed, and can escape only if that set expands. Rewriting scaffolding (tools, verifiers, decomposition) expands what an agent reaches without touching a weight, so frozen weights buy an eventual ceiling but no stationarity along the way. The criterion also separates three regimes usually merged: search within a fixed class, test-time training that raises the ceiling itself, and scaffold rewriting between them. Audit the scaffold, not the checkpoint. The same ceiling binds sideways. Best-of-$k$ orchestration realizes the best worker's ceiling exactly: width buys rate, not budget. Re-consulting a fixed pool has a horizon computable in advance, decided by the pool alone, and the one arrangement that would beat it, a weighted vote, needs diversity real workers lack: on 30 same-family workers the failure overlap sits at its maximum, and a majority fails 23/55 (42%) of tasks. We obtain the criterion by reading refinement as gradient boosting on the residual error between draft and target, a patch or git diff, and then measuring where that reading breaks: patches compose instead of standing beside each other to be voted on, and failures overlap. What we measure is saturation. Per-round improvement decays toward zero on SWE-bench, and churn decays geometrically across 401 production sessions, a shape shared with a pre-AI human baseline that establishes the regime without identifying its cause. Both breaks are engineering choices rather than laws about code, so together they specify a harness worth building.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-29 in cs.AI).
Categorie arXiv: Machine Learning (cs.LG) ; Artificial Intelligence (cs.AI); Software Engineering (cs.SE).
Pagina del lavoro: https://arxiv.org/abs/2609.34924 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-07T20:29:03.195Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-29?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).