📊 Full opportunity report: When AI Builds Itself: Inside Anthropic’s Evidence on Recursive Self-Improvement on ThorstenMeyerAI.com — validation score, market gap, and execution plan.
TL;DR
Anthropic’s internal data shows AI systems are increasingly automating AI development tasks, with potential for self-improvement loops. This could accelerate progress if the final human decision-making step is automated.
Anthropic has released new internal data indicating that AI systems are increasingly capable of automating key tasks involved in their own development, suggesting the potential for recursive self-improvement if the final human decision step is also automated. This development could significantly accelerate AI progress, although experts emphasize that it is not yet happening at the scale or certainty some claims suggest.
The report from The Anthropic Institute presents measurable evidence that AI models, particularly those from Anthropic, are rapidly advancing in automating tasks related to coding, experimentation, and problem-solving. For example, over 80% of code merged into Anthropic’s projects by May 2026 was authored by AI models like Claude, up from single digits in early 2025. Public benchmarks such as METR show AI capabilities doubling every four months, with models handling increasingly complex tasks, from fixing bugs to reproducing research results.
Inside labs, data indicates that AI systems are improving in their ability to perform engineering tasks, such as coding, and research tasks, like designing experiments. However, the report emphasizes that the critical bottleneck remains the human decision-making about which problems to pursue and which results to trust. The authors suggest that if this last step also becomes automatable, it could trigger a loop of recursive self-improvement, vastly speeding up AI development.
While the evidence shows rapid progress in AI capabilities, the authors caution that this does not mean self-improvement is imminent or inevitable. The internal data reveals strong performance in lower-level tasks but persistent gaps at higher-level decision-making, which remain human-controlled for now.
When AI builds itself
Anthropic is delegating a growing share of AI development to AI. Taken far enough, that points to a system that designs its own successor — recursive self-improvement. Not here yet, not inevitable. But the case isn’t speculation: it’s data on what AI is doing to AI development right now.
The curve that hasn’t bent
METR tracks the length of tasks AI can reliably complete on its own. That horizon is doubling roughly every four months — up from every seven. Anyone can check this in public data.
Task horizon — how long a job AI can handle solo
Each model handles dramatically longer tasks than the one a year before. The line keeps going up.
AI coding assistant
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Two kinds of work, one persistent gap
Building a frontier model splits into engineering and research. Across both, the pattern is the same — and so is the one thing AI still can’t do well.
Code, infrastructure, training
Claude can take an underspecified problem and find a method. Humans supply the goal; they no longer need to supply the method.
Which experiments, what they mean
Claude can match or outperform skilled humans at executing a well-specified experiment. But choosing which experiment still needs a human.
The same ladder Anthropic employees climb with experience
The No-BS Guide to AI for Trading & Market Research: How to Use ChatGPT, Claude & AI Tools for Market Analysis, Stock Research & Data-Driven Trading … … Required (The No-BS AI Playbooks Book 3)
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Watch the human share shrink, rung by rung
Walk up the four stages of AI development. At each, the human/AI split shifts — and the real internal numbers show exactly where AI has reached parity, gone superhuman, or still trails. Tap a rung.
The human role across the development loop
The doing now costs almost nothing in human time. What’s left is the deciding.
Spec-Driven Software Development with AI: A Practical Handbook for Turning Requirements into Designs, Tests, Tasks, and Production-Ready Code with AI Coding Agents
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Agents ran an open research project end to end
April 2026: the first demonstration of Claude running an open-ended research project from hypotheses to findings — on a real AI-safety problem.
Can a weaker model reliably supervise a stronger one?
Agents were left to solve it: proposing hypotheses, testing them, sharing findings across parallel agents, iterating. Measured against the gap between a “floor” (weak supervisor alone) and “ceiling” (strong model trained on correct answers).
(humans: ~23% in a week)
· ~$18,000 compute
the agents themselves
machine learning experiment platform
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Picking a better next step than the human
Real research sessions where a human took a wrong turn. Models saw only the work before the detour and proposed a next step; a judge that knew the outcome scored them. The day-to-day of research is this chain of next-step calls.
“Can the model pick a better next step than the human?”
Share of moments where the model’s next step was judged better. The amber line is the practical ceiling (an ideal answer that could see the whole session).
It depends on whether the trend continues — and what we do
The piece refuses a single prediction. It lays out three scenarios, and is clear about which it finds most likely.
The exponentials turn out to be S-curves
Maybe taste can’t be scaled into existence; maybe the constraint is the supply chain — chips, grid, interconnect — not intelligence. Even so, the world still changes: Glasswing’s Mythos found 10,000+ critical vulnerabilities in weeks, and a 100-person firm does the work of 1,000.
included for completeness · they doubt itDevelopment automates; humans still steer
100-person companies doing the work of tens of thousands — revolutionary, but turnable to harm (population-scale surveillance, tailored manipulation). Bound by Amdahl’s law: speeding one part shifts the bottleneck — which is exactly why human code review became Anthropic’s new chokepoint.
★ they think we’re likely heading hereAI designs and refines its own successors
Progress paced only by compute. Humans move to oversight of an expanding “virtual lab.” The future they understand least — especially whether alignment holds, or whether rare misalignments compound as models build successors, until control slips.
the one they’re most uncertain aboutBuild the option to slow down — verifiably
The piece ends on policy, not product. A unilateral pause just changes who leads; what’s missing is the ability to verify others have actually slowed.
Why a credible pause is hard — and worth building toward
A slowdown that only lets the least cautious catch up leaves everyone less safe. So the goal is the option: systems that let frontier labs verify others have genuinely stopped. Anthropic says if such systems existed and peers paused verifiably, it expects it would too.
Detection beats verification — and even that’s tough
Training runs are easier to conceal than missile silos, inputs are general-purpose, and whoever continues while others pause inherits the lead.
We’ve done it before — slowly
Regimes like the INF Treaty built verification and trust over decades. The authors’ blunt line: “We don’t have that long.”
Reading it in proportion
- This is one lab’s account of its own internal data — much previously unreported, not independently audited.
- The soft spots are stated in the original: lines-of-code overstates productivity; the self-reported 4× is probably high; the headline research result didn’t transfer to production scale; the next-step test used cherry-picked moments.
- “More autonomous” is not “fully autonomous” — every standout result still had a human framing the problem and defining success.
- That the authors surface these caveats themselves — against their own incentive — is part of what makes the document serious.
Implications of Accelerating AI Self-Development
This evidence raises important questions about the future of AI development, particularly the possibility of machines autonomously designing and improving themselves. If AI can fully automate the research and engineering processes, progress could accelerate exponentially, potentially outpacing human oversight. This has implications for AI safety, regulation, and the pace of technological change, making it a critical area for ongoing monitoring and policy development.
Background on AI Self-Improvement and Recent Developments
Recent years have seen rapid advancements in AI capabilities, with models increasingly able to perform complex tasks with minimal human input. Public benchmarks like METR and SWE-bench have documented steady improvements, but the internal data from Anthropic offers a new perspective by revealing how AI is contributing to its own development within labs. The concept of recursive self-improvement has long been theorized but lacked concrete evidence until now. The current data suggests that we are approaching a threshold where AI could begin to automate more of its own research and development processes, although full automation remains a work in progress.
“The data from Anthropic indicates that AI systems are not just performing tasks but are increasingly capable of automating the process of their own development, which could lead to rapid self-improvement.”
— Thorsten Meyer, AI researcher
Unresolved Questions About AI Self-Improvement Pace
It is not yet clear whether AI will soon automate the final human decision step necessary for true recursive self-improvement, or if technical, safety, and ethical barriers will slow or prevent this. The internal data shows progress but does not confirm imminent self-improvement loops at scale, and the timeline remains uncertain.
Future Monitoring of AI Development and Policy Implications
Researchers and policymakers will need to closely monitor ongoing AI capabilities and internal lab data to assess whether self-improvement loops are approaching feasibility. Further transparency from labs about internal metrics and experiments will be critical. Additionally, discussions around AI safety, regulation, and control strategies are likely to intensify as the potential for rapid self-enhancement becomes more tangible.
Key Questions
What is recursive self-improvement in AI?
Recursive self-improvement refers to the process where an AI system improves its own capabilities, potentially leading to exponential growth in intelligence and performance, if it can autonomously modify or enhance itself without human intervention.
Is AI currently self-improving without human input?
While AI systems are increasingly automating tasks involved in their development, the internal data suggests that the final decision-making step—choosing which problems to pursue—remains human-controlled. Full autonomous self-improvement has not yet been observed.
Why does this potential matter for AI safety?
If AI can self-improve rapidly and autonomously, it could outpace human oversight, raising concerns about control, safety, and alignment. Understanding the current progress helps inform regulation and safety measures.
How reliable are the internal data and benchmarks?
Public benchmarks like METR and SWE-bench provide measurable progress, but internal lab data is less accessible and may not fully represent future capabilities. The report emphasizes cautious interpretation and ongoing monitoring.
What are the next steps for researchers and regulators?
Researchers will need to continue measuring AI progress and transparency, while regulators should prepare for the possibility of rapid advances by establishing safety frameworks and oversight mechanisms.
Source: ThorstenMeyerAI.com
