📊 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 provides data indicating AI systems are already automating parts of their own development. While full recursive self-improvement is not yet achieved, the trend suggests it could happen sooner than expected, raising important questions about AI progress.
Anthropic has released new evidence suggesting that AI systems are already capable of automating significant portions of their own development processes, a step toward what is known as recursive self-improvement. This development, based on internal data and public benchmarks, indicates that AI could accelerate its own progress at a pace faster than human-led research, although full self-improvement remains unachieved and uncertain.
The Anthropic Institute’s report emphasizes that AI is increasingly handling tasks such as coding, experiment execution, and problem solving within research labs. Specifically, data shows that over 80% of code merged into Anthropic’s projects as of May 2026 was authored by AI systems like Claude, up from just a few percent in early 2025. Public benchmarks, such as METR and CORE-Bench, reveal that models are rapidly improving their ability to perform complex tasks—doubling their capabilities every four months on average, with some tasks now within hours or days of human performance.
Beyond external metrics, internal data indicates that AI models are climbing the ‘ladder’ of research tasks—from executing well-defined problems to designing experiments—though they still lag at the highest decision-making levels, such as selecting which problems to pursue. The authors clarify that while AI can now write significant amounts of code and perform research tasks independently, the critical bottleneck remains in decision-making, which is still primarily human-controlled.
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
Designing Instruction with Generative AI: 24/7 Support for Optimizing Teaching and Learning
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 suggests that AI systems are already contributing more actively to their own development, potentially setting the stage for recursive self-improvement. If the trend continues and the decision-making bottleneck is eventually overcome, AI could enter a feedback loop of rapid self-enhancement. Such a development could dramatically accelerate AI progress, impacting research, industry, and policy. However, experts emphasize that this is not an imminent certainty but a possibility that warrants careful monitoring and preparation.
Recent Advances in AI Capabilities and Benchmarks
Over the past two years, AI models like Claude have shown exponential improvements in handling complex tasks, with capabilities doubling roughly every four months according to public benchmarks like METR and SWE-bench. These benchmarks measure the ability to perform tasks such as fixing bugs, reproducing research results, and executing code, with models progressing from minimal success to near-complete proficiency within short periods. Internally, Anthropic’s data reveals that AI is increasingly involved in core research activities, from coding to experimental design.
This rapid progress has raised questions about whether AI can eventually automate the entire research cycle, including the critical decision-making stages that currently rely on human judgment. The report underscores that while progress is tangible, the leap to full autonomous self-improvement remains unconfirmed and complex.
“The data from Anthropic suggests we’re already seeing AI systems take over substantial parts of their own development, which could accelerate beyond current expectations.”
— Thorsten Meyer, AI researcher
Unconfirmed Potential for Full Recursive Self-Improvement
It remains unclear whether AI will eventually automate the entire research cycle, including the critical decision-making processes, without human intervention. The report emphasizes that full recursive self-improvement is not yet achieved and depends on overcoming significant technical and conceptual challenges. Experts caution that while the trend is promising, the timeline and feasibility of fully autonomous AI self-improvement are still uncertain.
Monitoring AI Development and Preparing for Possible Breakthroughs
Researchers and policymakers will need to closely track AI capabilities and internal progress reports from labs like Anthropic. Future developments may include more autonomous AI systems capable of designing and improving themselves, which could prompt discussions about safety, control, and regulation. Continued transparency and research into the pathways toward self-improving AI are essential to understanding and managing potential risks.
Key Questions
What is recursive self-improvement in AI?
Recursive self-improvement refers to an AI system’s ability to autonomously enhance its own capabilities, leading to an accelerating cycle of improvements without human intervention.
Is AI already self-improving at a significant scale?
Current evidence suggests AI systems are increasingly automating parts of their development, such as coding and experimental design, but full autonomous self-improvement has not yet been achieved.
What are the risks of AI self-improvement?
If AI systems can self-improve rapidly, it could lead to unpredictable behaviors or capabilities, raising concerns about safety, control, and ethical implications. Experts emphasize cautious monitoring.
How soon could full recursive self-improvement happen?
It is uncertain. The report indicates it could occur sooner than most expect if current trends continue and key bottlenecks are overcome, but no definitive timeline exists.
Source: ThorstenMeyerAI.com
