📊 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 new data indicating AI systems are increasingly capable of automating parts of their own development. The company suggests that, if certain human decision points are automated, AI could enter a loop of self-improvement at the speed of compute, though this is not yet happening and remains uncertain.
Anthropic has released new internal data indicating that AI systems are increasingly capable of automating their own development processes, suggesting the potential for recursive self-improvement if human oversight is eliminated. The report emphasizes that while this capability is not yet fully realized, the trend is moving rapidly, and the possibility could arrive sooner than most institutions anticipate.
The report from The Anthropic Institute highlights that AI models like Claude are now capable of handling a significant portion of code writing and experimental tasks traditionally performed by humans. For example, over 80% of code merged into Anthropic’s codebase as of May 2026 was authored by Claude, up from single digits in early 2025. Public benchmarks such as METR show that AI can now complete increasingly complex tasks, with the horizon of autonomous task completion doubling approximately every four months.
Inside labs, data suggests that AI models are improving in their ability to generate code, interpret experiments, and even select research goals, although key decision-making remains human-controlled. The authors differentiate between engineering tasks, which AI is automating effectively, and research-level decision-making, where gaps still exist. They warn that if these gaps narrow further, AI could begin self-improving at a rapid pace, driven by compute power rather than human effort.
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.
Your First AI Implementation & Automation Agency: A Beginner’s Playbook to Build, Launch, and Scale a Profitable AI Automation Agency Using No-Code Tools, … Workflows (AI-POWERED AGENCY 9)
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
AI Code Generation's Supply Chain Exposure: How AI-Assisted Development Creates Hidden Vulnerabilities in Dependencies and Build Pipelines
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 research automation platforms
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 system hardware
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 Autonomous Development Capabilities
This development signals a potential shift toward AI systems capable of automating significant parts of their own creation and improvement processes. If fully realized, recursive self-improvement could accelerate AI progress beyond current human control, raising important questions about safety, regulation, and the future of AI research. While not imminent, the trend suggests that the pace of AI development could reach a point where human oversight becomes less central, with profound implications for technology and society.
Current Evidence of AI Accelerating Its Own Development
Anthropic’s report builds on public benchmarks like METR, SWE-bench, and CORE-Bench, which show AI models rapidly advancing in capabilities such as coding, bug fixing, and reproducing research results. These benchmarks indicate a near-exponential growth in AI’s ability to perform tasks that were once considered exclusively human. Internally, data from Anthropic reveals that AI now writes most of the code in its projects and is increasingly involved in experimental design and goal selection, marking a significant step toward autonomous research processes.
Historically, AI progress has been measured by external benchmarks, but the internal data suggests that the pace of development within labs is accelerating faster than publicly visible. This internal evidence underpins the report’s core claim that AI is already, to some extent, building itself, with the potential for a self-improving loop if certain human decision points are automated.
“The data indicates AI models are climbing the ladder of research and engineering tasks, and if the last human-held decision point is automated, recursive self-improvement could be within reach.”
— Thorsten Meyer, author of the report
Uncertainties Surrounding AI Self-Improvement Pace
It remains unclear when or if AI will fully automate the decision-making processes necessary for recursive self-improvement. The evidence is primarily based on internal data and public benchmarks, which cannot directly measure internal development speed or the feasibility of automating high-level research decisions. Experts warn that significant technical, safety, and regulatory hurdles could delay or prevent this scenario, and it is not yet certain whether the current trends will continue unabated.
Next Steps in Monitoring AI Self-Development Trends
Researchers and industry stakeholders will likely focus on tracking internal development metrics and benchmark progress to assess how close AI systems are to autonomous self-improvement. Companies may also begin experimenting with automating higher-level research decisions, testing the limits of AI’s capabilities. Policymakers and safety organizations will need to consider the implications of accelerating AI autonomy and prepare appropriate regulatory frameworks.
Key Questions
What is recursive self-improvement in AI?
Recursive self-improvement refers to AI systems improving their own capabilities autonomously, potentially leading to rapid, exponential progress without human intervention.
Is AI already self-improving at scale?
Current evidence suggests AI is automating many tasks involved in its development, but full autonomous self-improvement, where AI designs and improves itself without human input, has not yet been achieved.
What are the risks of AI self-improvement?
Potential risks include loss of human control, unexpected behaviors, and rapid technological changes that could outpace safety measures and regulations.
When might AI achieve full self-improvement capability?
It is uncertain; current trends suggest it could happen within the next few years if the pace of AI capability growth continues and key decision points are automated.
Source: ThorstenMeyerAI.com
