📊 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 significant parts of their own development. While full recursive self-improvement is not yet achieved, the evidence suggests it could happen sooner than expected, raising important questions about AI progress and control.
Anthropic’s new report presents evidence that AI systems are increasingly capable of automating core tasks involved in their own development, suggesting the possibility of recursive self-improvement. This development is significant because it indicates AI could accelerate its own progress at a pace faster than human-led efforts, though experts emphasize it is not yet happening at that scale.
The report from The Anthropic Institute draws on publicly available benchmarks and internal data to show that AI models like Claude are rapidly improving in their ability to perform tasks traditionally done by humans, such as coding, experiment design, and problem-solving. For example, Anthropic engineers now ship eight times more code per quarter than they did in 2021–2025, a clear sign of accelerating development.
Public benchmarks like METR, SWE-bench, and CORE-Bench demonstrate that AI’s capabilities in handling complex tasks—such as fixing bugs, reproducing research results, and tackling multi-hour projects—are growing exponentially. These trends suggest that AI could soon handle tasks requiring days or even weeks of human effort, with some models already capable of working for 16 hours continuously.
Inside labs, data indicates that AI systems are increasingly automating the lower and middle rungs of the research ladder—coding and experiment execution—while the highest-level decisions about research goals and problem selection remain human-controlled. The authors note that AI’s ability to generate code and execute experiments has improved dramatically, but the decision-making about which problems to pursue still depends on human judgment.
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 automation tools
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 experiment design software
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 automation hardware
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 research automation 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 Development
This evidence suggests that AI systems are already automating substantial portions of their own development process, which could lead to a rapid acceleration in AI capabilities if the remaining human-controlled steps are automated. Such a shift could dramatically shorten the timeline for achieving recursive self-improvement, raising questions about control, safety, and the future pace of AI progress. While full self-improvement is not yet happening, the trend indicates it could occur sooner than many institutions are prepared for, prompting urgent discussions about oversight and regulation.
Current State of AI Self-Development Evidence
The idea of AI improving itself has long been debated, but until now, most discussions relied on speculation about future systems. Anthropic’s report is notable because it relies on concrete, internal data and public benchmarks to show that AI is already automating some aspects of its own research and development. The trend of increasing capability is consistent across multiple benchmarks and internal metrics, with models rapidly climbing the ladder of research tasks from coding to experiment execution.
This acceleration aligns with broader trends in AI research, where models like Claude have gone from handling simple tasks to managing complex, multi-hour projects within a year. However, the gap remains in higher-level decision-making, which continues to depend on human judgment. The report underscores that while progress is real, the leap to fully autonomous self-improvement is not yet achieved, but the potential is there.
“The evidence shows AI is already automating significant parts of its own development, and if the last human-controlled bottleneck falls, we could see recursive self-improvement happening sooner than most expect.”
— Thorsten Meyer, author of the report
Uncertainties Surrounding AI Self-Improvement Timeline
It remains unclear when, or if, AI will fully automate the decision-making processes necessary for recursive self-improvement. The evidence shows rapid progress in lower-level tasks, but the leap to autonomous goal-setting and system design is still unconfirmed. Experts caution that while trends are promising, the timeline for achieving full self-improvement is uncertain and dependent on future breakthroughs and safety considerations.
Next Steps in Monitoring AI Development Progress
Researchers and policymakers will likely focus on tracking further internal data from AI labs, refining benchmarks to measure higher-level capabilities, and debating safety protocols for increasingly autonomous systems. The industry may also see increased investment in understanding how to control or slow down AI’s self-improvement potential, ensuring that progress remains aligned with safety standards.
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 growth in intelligence and performance without human intervention.
Are we already seeing AI self-improve?
Current evidence suggests AI is automating many research and development tasks, but full recursive self-improvement—where AI designs and improves itself without human input—has not yet been achieved.
Why is this development important?
If AI begins to self-improve at an accelerating rate, it could dramatically shorten timelines for achieving advanced AI, raising questions about safety, control, and the future of AI governance.
What are the main limitations right now?
The primary limitation is that AI systems still depend on humans for setting research goals and making high-level decisions. Automating these aspects remains a significant challenge.
What should we expect next from AI research?
Expect ongoing monitoring of internal lab data, development of new benchmarks to measure higher-level capabilities, and discussions on how to manage AI’s self-improvement potential responsibly.
Source: ThorstenMeyerAI.com
