📊 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 research and development tasks. While current evidence shows significant progress, full autonomous self-improvement remains unconfirmed and uncertain.
Anthropic’s new report provides the first concrete evidence that AI systems are increasingly capable of automating parts of their own development, raising the possibility that, if certain bottlenecks are overcome, AI could begin improving itself at the speed of computation rather than human effort.
The report from The Anthropic Institute states that AI models, notably Claude, are now performing tasks that were previously reliant on human input, such as code writing and experiment execution, at an increasing rate. For example, Anthropic engineers now ship eight times more code per quarter than they did between 2021 and 2025, and public benchmarks show improvements in AI capabilities, with models handling increasingly complex tasks.
Internal data from Anthropic indicates that AI systems like Claude are already automating significant portions of the research and engineering process. Claude now writes over 80% of the code merged into Anthropic’s projects, a significant increase from less than 10% before 2025. These developments suggest a trajectory where AI could independently identify problems, design solutions, and even select research goals, although the report emphasizes that the decision-making process—what to improve—is still primarily human-driven.
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.
Vibe Coding: Building Production-Grade Software With GenAI, Chat, Agents, and Beyond
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
Mastering AI in Clinical Research: : Tools, Techniques, and Real-World Applications (AI in Clinical Research Series Book 5)
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.
Agentic Spec-Driven Development: A Practical Method for Using AI to Build Complete Specifications for Software, Products, and Knowledge Work
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
Non-Deterministic Autonomous Coding Agents: Building Self-Improving Systems That Ship While You Sleep
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.
Potential for Rapid Autonomous AI Development
This evidence suggests that AI could potentially accelerate its own development, leading to a feedback loop of improvement that might surpass human oversight. While not yet autonomous in decision-making, the trend indicates that the reliance on human intervention in research and development could decrease, raising questions about future AI capabilities and safety considerations.
Current Evidence of AI Progress and Limitations
Anthropic’s findings are based on public benchmarks like METR, SWE-bench, and CORE-Bench, which assess AI’s ability to perform tasks such as coding, bug fixing, and reproducing research results. These benchmarks show consistent growth in AI performance over the past two years. However, internal data indicates that while AI is excelling at execution tasks, it still faces challenges in higher-level decision-making, such as goal setting and strategic planning, which are essential for autonomous self-improvement.
Historically, AI development has progressed incrementally, but recent data suggests a possible shift toward faster progress, although many technical and safety challenges remain to be addressed.
“AI is already, measurably, accelerating the development of AI itself, and if a remaining bottleneck falls, it could enter a loop of self-improvement at the speed of compute.”
— Thorsten Meyer, author of the report
Unconfirmed Aspects of Full Self-Improvement
While evidence shows AI systems are automating many tasks, it remains unclear whether they can autonomously set research goals or design their own successors without human input. The critical decision-making capacity—what to improve—has not yet been demonstrated to be fully autonomous, and the idea of a self-improving loop remains speculative.
Next Steps in Monitoring AI Self-Development
Further data collection from internal experiments and external benchmarks will help clarify whether AI can progress toward autonomous self-improvement. Researchers are likely to focus on developing AI systems capable of higher-level decision-making and goal setting, with safety and control measures remaining important considerations as this area advances.
Key Questions
What is recursive self-improvement in AI?
It refers to AI systems improving their own capabilities autonomously, potentially leading to rapid development without human intervention.
Is AI currently capable of fully self-improving?
No, current AI can automate many research tasks but still relies on humans for goal setting and strategic decisions. Full autonomous self-improvement has not been demonstrated.
Why does this potential matter for the future of AI?
If AI can self-improve rapidly, it could lead to significant advancements or risks that are difficult to predict or control, impacting safety, ethics, and global stability.
What are the main challenges in achieving AI self-improvement?
The primary challenge is enabling AI systems to autonomously set meaningful research goals and design their own successors, which involves complex decision-making and safety considerations.
When might autonomous AI self-improvement become a reality?
It remains uncertain; current trends suggest it could occur within the next few years if technical and safety challenges are addressed, but no specific timeline has been established.
Source: ThorstenMeyerAI.com
