📊 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 new report offers data indicating AI is already automating parts of its own development, hinting at the potential for recursive self-improvement if human decision-making steps are automated. The evidence is based on internal metrics and public benchmarks, but key gaps remain.
Anthropic has released new evidence indicating that AI systems are now capable of automating significant aspects of their own development, raising the possibility that recursive self-improvement could occur if human oversight diminishes. This development is based on internal data and public benchmarks, and it suggests that AI could accelerate its progress at a pace faster than human researchers can match, though key gaps remain.
The Anthropic Institute’s report presents data showing AI models like Claude are increasingly capable of performing tasks that previously required human intervention, including writing code and conducting experiments. For example, the proportion of code authored by AI in Anthropic’s projects rose from single digits in early 2025 to over 80% by May 2026. Public benchmarks such as METR, SWE-bench, and CORE-Bench demonstrate rapid improvements in AI’s ability to handle complex tasks, with models progressing from hours-long tasks to potentially days-long ones within a year.
Inside labs, data indicates that AI models are now capable of designing solutions and executing experiments that once required human expertise. The report distinguishes between engineering tasks—such as coding and infrastructure—and research tasks, like goal-setting and experiment interpretation. While models like Claude have shown strong performance at lower levels, they still lag in making strategic decisions about what problems to pursue, a gap that could be critical if AI begins to autonomously select and pursue research directions.
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.
Coding with AI For Dummies (For Dummies: Learning Made Easy)
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
CLAUDE AI UNLEASHED From First Prompts to Pro: The Complete Guide to Claude AI for Writing, Research, Coding, and Business (The Claude AI Mastery Series)
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.
Architecting Data and Machine Learning Platforms: Enable Analytics and AI-Driven Innovation in the Cloud
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
Code: The Hidden Language of Computer Hardware and Software
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 AI Automating Its Own Development
This evidence suggests that AI systems could soon reach a stage where they not only perform research tasks but also design and improve themselves without human input. If such recursive self-improvement becomes possible, it could accelerate AI progress exponentially, impacting fields ranging from scientific research to technological innovation. However, the report emphasizes that human oversight remains crucial, especially in decision-making stages where AI still lags behind.
Background on AI Self-Improvement and Recent Developments
The concept of recursive self-improvement has long been discussed in AI research, often viewed as a potential path to rapid technological advancement. Until now, most claims have been speculative, based on future projections. Anthropic’s recent report, however, bases its conclusions on concrete internal data and public benchmarks, marking a shift toward empirical evidence. The trend of AI models taking on more complex tasks has been accelerating since 2024, with models like Claude demonstrating notable progress in coding, bug fixing, and research simulation tasks.
Despite these advances, experts caution that full autonomous self-improvement—where AI independently designs its successor—remains a distant goal. The current evidence indicates increasing automation but not complete independence in research decision-making, which is critical for true recursive self-improvement.
“The data clearly shows that AI is automating more of its own development tasks, but the leap to autonomous self-design is still a significant gap.”
— Thorsten Meyer, AI researcher
Unanswered Questions About Autonomous AI Self-Improvement
It remains unclear whether AI systems will eventually reach a point where they can autonomously select research goals and design their own successors without human oversight. The current data shows progress in executing tasks but not in strategic decision-making, which is crucial for true recursive self-improvement. Experts warn that significant technical and safety challenges could slow or prevent this development, and it is not yet known when or if AI will cross this threshold.
Future Steps and Monitoring AI Self-Development Progress
Researchers and industry leaders will closely monitor internal metrics and benchmark performance to assess whether AI models continue to improve in automation capabilities. Further transparency from labs about internal development processes and decision-making will be critical. Additionally, discussions around safety, control, and ethical implications are expected to intensify as AI approaches higher levels of autonomy in research tasks.
Key Questions
What is recursive self-improvement in AI?
Recursive self-improvement refers to AI systems’ ability to autonomously improve or redesign themselves, potentially leading to rapid technological advancement without human intervention.
How is Anthropic measuring AI progress in self-improvement?
Anthropic uses internal data on code contributions, benchmark tests like METR, SWE-bench, and CORE-Bench, and metrics on task completion times to evaluate AI’s automation capabilities.
Does this mean AI can now fully develop itself without humans?
No, current evidence shows AI can automate certain research and coding tasks, but strategic decision-making and goal-setting still require human input. Full autonomous self-improvement remains a future possibility, not an immediate reality.
What are the risks if AI begins self-improving at a rapid pace?
Potential risks include loss of human oversight, unintended behaviors, and safety concerns. Experts emphasize the importance of safety measures and regulatory oversight as AI capabilities grow.
Source: ThorstenMeyerAI.com
