🔍 Read the full analysis: Understanding AI Self-Improvement And The Debate Over Bans on ThorstenMeyerAI.com
Get the latest gadgets delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
A November 13 essay by podcaster Dwarkesh Patel renewed debate over whether AI is beginning to improve AI development. Available evidence points to AI assisting researchers, while fully autonomous successor-building has not been documented; two congressional bills proposing restrictions remain in committee.
A viral essay published November 13 by podcaster Dwarkesh Patel has renewed debate over whether artificial intelligence is beginning to accelerate the development of more capable AI. The evidence described in the essay and in assessments by Anthropic and the evaluator METR points to AI helping human researchers—not independently designing and building its own successor—while two congressional bills proposing restrictions on AI development remain in committee.
Patel’s essay, titled “The avalanche that hit math is coming for AI development,” draws on published company figures and a METR assessment. Its claims about what may happen next are Patel’s forecast, not a new demonstration that AI systems can independently improve themselves. The distinction matters because “recursive self-improvement” can describe a broad range of human-supervised assistance or a much stronger scenario in which a system builds its successor without human direction.
Anthropic’s figures for August 2026 say that more than 90% of its AI research and development work involves AI at or above a collaboration level, and that Claude leads 26% of the work. The company says no measured subset of its AI research and development is fully autonomous. Anthropic defines the more demanding endpoint as an AI system that can fully autonomously design and develop its own successor, and says that has not been reached.
METR, a nonprofit evaluator, examined internal Anthropic data from August 12 to October 29. In a November 5 assessment, it said Anthropic was likely improving model capabilities on well-specified tasks at about twice the rate seen between March 2024 and October 2025. METR attached high uncertainty to that estimate, saying the true acceleration could be closer to 1.3 times or three times. It attributed most of the likely increase to AI assistance, while saying AI was not yet making highly significant, fully autonomous contributions to the most important parts of model training.
Has it started? · explainer
AI already does much of the work of building AI. Not alone.
How far AI research and development is automated at one lab, on a scale from AL0 (no AI) to AL5 (no human in the loop), and where the proposed ban would draw its line.
Anthropic’s AI R&D work, August 2026
Automation levels from Epoch AI’s scale, as measured by Anthropic
Anthropic: “We are not there yet.”
- AI collaborates on most of the workAbove 90% of the work is at or above “AI collaborates”, where AI does large chunks under close human direction.
- AI leads about a quarterIn 26%, Claude completes most of a task from a high-level prompt while a human supervises.
- No closed loopNo measured part runs fully autonomously. That is the step “recursive self-improvement” usually means.
- Where would a ban start?The bill names “greatly accelerate” as a warning sign, so its reach depends on a measurement nobody yet makes in a standard way.
Sources: Anthropic, “Measurements for understanding the pace of AI development inside frontier labs” and “When AI builds itself” (figures as of August 2026, Anthropic’s own measurements); Ban Artificial Superintelligence Act of 2026, S.5493 and H.R.10538, sec. 3(6)(A). Bars show shares of Anthropic’s AI R&D work.
Graphic: Powered by Thorsten Meyer AI
The Difference Between Assistance and Autonomy
The policy debate turns on what counts as self-improvement. AI systems that help researchers write code, analyze results or complete technical tasks may speed up development even when people choose research goals, evaluate findings and decide whether to train or deploy a model. That is a real change in how research is done, but it is not the same as a system independently creating its successor.
This distinction has practical consequences for proposed restrictions. A law aimed at fully autonomous systems would have a different reach from one covering tools that can automate or greatly accelerate AI research, a broader category that may include human-supervised work. The latter could affect research and development well before an AI system operates without people in the loop. The bills’ scope therefore matters to researchers, companies and users, even though they have not advanced and nothing changes for businesses or users today.
Evidence about the pace of progress also affects public expectations. METR’s estimate suggests faster work on a defined class of tasks, but its uncertainty and the limits it identifies do not establish a timeline for autonomous self-improvement. Treating a forecast as a confirmed milestone risks blurring what has been measured with what remains hypothetical.
Products Worth Considering

Amazon Kindle Paperwhite Signature Edition (2024), 32 GB, Metallic Black

Bose QuietComfort Wireless Noise-Cancelling Headphones, Cypress Green

Bose SoundLink Max Portable Bluetooth Speaker, IP67 Waterproof and Dustproof, Up to 20 Hours of Battery Life, Black
As an affiliate, we earn on qualifying purchases.
What Congress Has Proposed
The Ban Artificial Superintelligence Act of 2026 was introduced by Sen. Bernie Sanders as S.5493 on September 23 and by Rep. Greg Casar as H.R.10538 on September 24. The bill text would pause training, modification and fine-tuning of large AI systems, including through recursive self-improvement, until a new Department of Artificial Intelligence established rules. It would also prohibit systems with the capacity to automate or greatly accelerate AI research and development.
According to the latest status records described in the source material, both bills were referred to committee and had not advanced. The House measure had 12 Democratic cosponsors. A separate proposal from Rep. Ro Khanna, called the Human Control Over AI Act, was reported by CNBC in September as a summary; the source says it found no bill text to review.
The proposals returned to attention after the Associated Press called both chambers of Congress for Democrats on November 3. Supporters of a ban see the next Congress as a possible opportunity to revisit restrictions. That political development does not change the bills’ current status or establish that either proposal will pass.
““We are not there yet, and recursive self-improvement is not inevitable. But it could come sooner than most institutions are prepared for.””
— Anthropic, in its essay “When AI builds itself”
Limits of the Available Evidence
The evidence does not settle how quickly AI research assistance may improve or whether today’s measured tasks predict broader capabilities. METR said its estimate carried high uncertainty, and reported that Anthropic redacted much of the evidence it considered most informative. Its assessment used internal company data, so readers cannot independently inspect all the underlying material from the information provided here.
OpenAI also reported progress toward an “automated research intern” in September, according to The Decoder’s account of an OpenAI blog post. The source material says OpenAI reported that tasks under 15 minutes succeeded 86% of the time without intervention, while more than half of successful tasks lasting four to eight human hours still required at least one human step. The Decoder noted that the figures were internal and that OpenAI described no independent review. These company-reported measurements are not proof of fully autonomous research.
It remains unclear how the proposed laws would define “greatly accelerate” research in practice, what rules a new department would set, or whether the bills will receive further action. The available material also does not establish that Patel’s forecast will occur on a particular timeline.
Committee Action and New Evidence
The immediate legislative milestone is whether the House and Senate bills receive committee consideration. Their referral is the latest status reported here; no committee vote or enactment is described. Any future action would clarify whether Congress is pursuing a broad pause on AI development or a narrower restriction tied to more autonomous systems.
On the research side, further independent evaluations and clearer disclosure of methods would help readers judge claims about faster AI development. Company figures, external assessments and forecasts should remain distinct as new data appears. For now, the documented picture is AI assisting human researchers, with the extent and pace of future automation still uncertain.
Key Questions
Has recursive self-improvement started?
That depends on the definition. AI is assisting and, in some measured work, leading parts of research under human supervision. The source material does not document an AI system autonomously designing and building its own successor.
What did METR find about Anthropic’s progress?
METR estimated that Anthropic was likely improving capabilities on well-specified tasks at about twice its rate from March 2024 to October 2025. It described the estimate as highly uncertain and said AI assistance likely accounted for most of the increase.
Would the proposed ban stop all AI research?
The bill text described in the source would pause training, modification and fine-tuning of large AI systems pending rules, and would prohibit systems able to automate or greatly accelerate AI research and development. How those terms would apply in practice is not yet settled, and the bills have not advanced beyond committee referral.
Are the bills law now?
No. The Sanders and Casar bills were referred to committee, and the source reports no enactment. The proposals do not change current requirements for businesses or users.
Source: ThorstenMeyerAI.com
