📊 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 presents data indicating AI systems are increasingly capable of automating AI research tasks. While human decision-making remains a bottleneck, evidence suggests rapid progress toward autonomous self-improvement, though this is not yet achieved.
Anthropic’s latest report reveals that AI systems are already significantly automating parts of their own development, with evidence showing rapid progress in capabilities that could, if bottlenecks are removed, lead to recursive self-improvement.
The report, from The Anthropic Institute, states that AI models like Claude are increasingly able to perform research and coding tasks independently. Public benchmarks show a doubling of AI capabilities every four months, notably in tasks like code fixing and reproducing research results. Internal data from Anthropic indicates that over 80% of code merged in May 2026 was authored by Claude, up from low single digits in early 2025.
Despite these advances, the report emphasizes that the critical remaining challenge is human decision-making—choosing which problems to pursue and trusting AI-generated results. The authors argue that progress suggests AI is climbing the ‘ladder’ from executing tasks to designing experiments and setting research goals, but full autonomous self-improvement remains unachieved.
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.
AI development environment
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
AI self-improving systems
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 development matters because it indicates AI systems are rapidly advancing in their ability to perform research and development tasks, potentially reducing the time and human effort needed to create new AI models. If the bottleneck of human decision-making is eventually automated, it could lead to a self-reinforcing loop of AI improvement, raising questions about control, safety, and the timeline of AI capabilities.
Current Evidence of AI Progress and Limitations
Anthropic’s findings are based on public benchmarks like METR, SWE-bench, and CORE-Bench, which show AI models improving from handling simple tasks to complex, multi-hour activities within months. Internal data reveals AI models are increasingly responsible for code development and experimental execution, but the decision-making aspect—choosing what to research—is still human-driven. The report underscores that while capabilities are advancing rapidly, full automation of the research cycle remains a future possibility, not an immediate reality.
“Our data shows AI is climbing the research ladder, from executing tasks to designing experiments, but the final step—autonomous goal setting—remains elusive.”
— Thorsten Meyer, author of the report
Unresolved Questions About AI Self-Improvement
It is not yet clear when or if AI will fully automate the research cycle, including goal setting and strategic decision-making. The possibility of a rapid, uncontrollable self-improvement loop remains speculative, with the authors emphasizing that it is not inevitable and depends on future breakthroughs.
Next Steps in Monitoring AI Development
Researchers and institutions will likely focus on tracking internal metrics of AI-driven research activities, as well as developing safety protocols for increasingly autonomous AI systems. Further transparency from labs about internal progress will be crucial to understanding whether the trend toward self-improvement accelerates or stalls.
Key Questions
What is recursive self-improvement in AI?
Recursive self-improvement refers to AI systems improving their own capabilities without human intervention, potentially leading to rapid, exponential growth in intelligence.
How close are we to AI autonomously designing new models?
Current evidence suggests AI can already perform many research tasks, but fully autonomous model design and strategic decision-making are still beyond reach, with significant gaps remaining.
What are the risks of AI self-improvement?
If AI systems begin to improve themselves rapidly and autonomously, it could lead to unpredictable behaviors and challenges in maintaining control, raising safety and ethical concerns.
Will this development accelerate AI safety debates?
Yes, as AI systems become more capable of self-improvement, discussions about safety, oversight, and regulation are likely to intensify to prevent unintended consequences.
Source: ThorstenMeyerAI.com
