The Governance Gauntlet: A Dual-Rubric Extension of Karpathy's Auto-Research Loop — Detecting Silent Metric-Gaming in Recursive Self-Improvement Systems | Zenodo Skip to main Communities My dashboard Log in Sign up Published April 22, 2026 | Version 1.0 Working paper Open The Governance Gauntlet: A Dual-Rubric Extension of Karpathy's Auto-Research Loop — Detecting Silent Metric-Gaming in Recursive Self-Improvement Systems Authors/Creators Accornero, Paul Ferrando (Researcher) Description Karpathy's auto-research loop (March 2026) and its rapid derivatives (Gu 2026; Lütke 2026; SkyPilot 2026) establish a minimal, powerful architecture for recursive self-improvement: one editable surface, one scalar metric, one time budget per trial, keep-or-revert on scalar. The design is an elegant concession to the bitter lesson — less structure, more search. It is also structurally vulnerable to Goodhart's Law. We identify one class of failure mode that the vanilla loop cannot detect: silent metric-gaming, in which the primary meta-agent accumulates edits that increase the scalar metric through mechanisms the scalar was not designed to reward. We formalise the vulnerability using Manheim & Garrabrant's (2018) four-variant Goodhart taxonomy and propose the Governance Gauntlet, a minimal dual-rubric extension in which a second, same-family LLM meta-agent runs an adversarial integrity rubric in parallel with the primary loop. Keep-or-revert now requires BOTH primary metric non-degraded AND adversarial auditor verdict non-GAMING. We pre-register a six-subject empirical evaluation (Subject α, Subject β, four gaming archetypes, three arms) on the Open Science Framework and release this version as the priority-date pre-registration; empirical fills follow in v2 within the publication window. We argue the Gauntlet is a concrete operationalisation of EU AI Act Articles 14 (human oversight) and 15 (accuracy, robustness and cybersecurity) for any Karpathy-style deployment in a regulated domain, and sketch extensions to the Four Ds Framework for algorithmic readiness in agentic commerce. Notes (English) Version 1 (pre-registration release). Priority-date deposit on 22 April 2026. Companion pre-registration at OSF (osf.io/skpgn). SSRN deposit at abstract 6625918. Empirical fills follow in v2 within the publication window. Full reproducibility bundle (code, anonymised subject substrates, analysis notebook) will be released as a separate Zenodo record within 14 days. Files EXPERIMENT_DESIGN_EXPLAINER_standalone.pdf Files (582.1 kB) Name Size Download all EXPERIMENT_DESIGN_EXPLAINER_standalone.pdf md5:c3b6b69f6b8e94ee82b5aff4ff64aca2 250.9 kB Preview Download Karpathy_Governance_Gauntlet_v1.pdf md5:aa064bc7a14b95c951fe0dc949b96b21 331.2 kB Preview Download Additional details Identifiers URL https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6625918 URL https://osf.io/skpgn URL https://ssrn.com/abstract=6625918 Related works Is supplement to Working paper: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6625918 (URL) Other: https://osf.io/skpgn (URL) Dates Issued 2026-04-22 References Karpathy, A. (2026). autoresearch: AI agents running research on single-GPU nanochat training automatically. GitHub repository. First commit 6 March 2026. https://github.com/karpathy/autoresearch Kahn, J. (2026). "The Karpathy Loop": 700 Experiments, 2 Days, and a Glimpse of Where AI Is Heading. Fortune, 17 March 2026. https://fortune.com/2026/03/17/andrej-karpathy-loop-autonomous-ai-agents-future/ Goodhart, C. A. E. (1975). Problems of monetary management: the UK experience. Papers in Monetary Economics, Reserve Bank of Australia, vol. 1. Manheim, D., & Garrabrant, S. (2018). Categorizing variants of Goodhart's Law. arXiv:1803.04585. https://arxiv.org/abs/1803.04585 Hubinger, E., van Merwijk, C., Mikulik, V., Skalse, J., & Garrabrant, S. (2019). Risks from learned optimization in advanced machine learning systems. arXiv:1906.01820. https://arxiv.org/abs/1906.01820 Krakovna, V., Uesato, J., Mikulik, V., Rahtz, M., Everitt, T., Kumar, R., Kenton, Z., Leike, J., & Legg, S. (ongoing). Specification gaming: the flip side of AI ingenuity. Google DeepMind. https://deepmind.google/blog/specification-gaming-the-flip-side-of-ai-ingenuity/ Irving, G., Christiano, P., & Amodei, D. (2018). AI safety via debate. arXiv:1805.00899. https://arxiv.org/abs/1805.00899 Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act). Official Journal of the European Union, L 2024/1689, 12 July 2024. https://eur-lex.europa.eu/eli/reg/2024/1689/oj 208 Views 119 Downloads Show more details All versions This version Views Total views 208 208 Downloads Total downloads 119 119 Data volume Total data volume 34.2 MB 34.2 MB More info on how stats are collected.... Versions External resources Indexed in OpenAIRE Communities Keywords and subjects Keywords Karpathy Loop auto-research agentic AI AI governance Goodhart's Law specification gaming dual-rubric recursive self-improvement metric gaming EU AI Act Governance Gauntlet pre-registration Details DOI DOI Badge DOI 10.5281/zenodo.19689504 Markdown [](https://doi.org/10.5281/zenodo.19689504) reStructuredText .. image:: https://zenodo.org/badge/DOI/10.5281/zenodo.19689504.svg :target: https://doi.org/10.5281/zenodo.19689504 HTML <a href="https://doi.org/10.5281/zenodo.19689504"><img src="https://zenodo.org/badge/DOI/10.5281/zenodo.19689504.svg" alt="DOI"></a> Image URL https://zenodo.org/badge/DOI/10.5281/zenodo.19689504.svg Target URL https://doi.org/10.5281/zenodo.19689504 Resource type Working paper Publisher The AI Praxis Languages English Rights License Creative Commons Attribution 4.0 International The Creative Commons Attribution license allows re-distribution and re-use of a licensed work on the condition that the creator is appropriately credited. Read more Copyright Copyright (c) 2026 Paul F. Accornero Citation Export Technical metadata Created April 23, 2026 Modified April 23, 2026 Jump up About About Policies Infrastructure Principles Projects Roadmap Contact Blog Blog Support Help FAQ Developers REST API OAI-PMH Contribute GitHub Donate Funded by Powered by CERN Data Centre & InvenioRDM Status Privacy policy Cookie policy Terms of Use This site uses cookies. Find out more on how we use cookies Accept all cookies Accept only essential cookies