📊 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 showing AI models are already automating parts of their own development, with potential for full self-improvement if human oversight diminishes. The evidence is based on internal metrics and public benchmarks, but key gaps remain.
Anthropic has released new evidence suggesting that AI systems are increasingly capable of automating substantial parts of their own development, including coding and experimental processes. This development raises the possibility of recursive self-improvement, where AI could potentially enhance itself at speeds surpassing human intervention, although such a scenario is not yet inevitable.
The report from Anthropic’s Institute presents data showing that AI models like Claude now perform a significant share of coding tasks, with over 80% of code merged into the company’s base authored by AI as of May 2026. Public benchmarks, such as METR and SWE-bench, demonstrate rapid improvements in AI capabilities, with models handling tasks that once required days now achievable within hours or minutes. These trends suggest that AI is already speeding up its own development process. However, the report emphasizes that the critical bottleneck remains human judgment—particularly in deciding which problems to pursue—an area where AI still lags behind human experts. The evidence is rooted in internal metrics and public data, but the authors caution that the internal pace of AI-driven research and development is less transparent and harder to measure precisely. While the data indicates a clear acceleration, whether this will lead to fully autonomous AI self-improvement remains uncertain, hinging on future breakthroughs and the potential automation of research decision-making.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.
OpenClaw Automation Blueprint: The Complete Practical Guide for Solo Entrepreneurs and Small Businesses to Build Autonomous AI Agents Without Coding (OpenClaw Automation Mastery Series Book 1)
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
The Agentic AI Bible: The Complete and Up-to-Date Guide to Design, Develop, and Scale Goal-Driven, LLM-Powered Agents that Think, Execute and Evolve
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.
Quick Start Guide to Large Language Models: Strategies and Best Practices for ChatGPT, Embeddings, Fine-Tuning, and Multimodal AI (Addison-Wesley Data & Analytics Series)
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
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.
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 could reach a point where it significantly automates its own research and development processes, potentially leading to rapid self-improvement cycles. Such a shift could accelerate AI capabilities beyond current expectations and challenge existing frameworks for safety, oversight, and regulation. Understanding whether and when this might happen is critical for policymakers, researchers, and industry leaders concerned with AI safety and control.
Internal Data and Public Benchmarks Show Rapid Progress
Anthropic’s report builds on public benchmarks like METR, SWE-bench, and CORE-Bench, which track AI’s ability to perform increasingly complex tasks, from coding to research reproduction. These metrics reveal exponential growth in AI capabilities over the past two years, with models now handling tasks that previously required days or weeks. Internally, Anthropic’s data indicates that AI models like Claude are automating a growing share of engineering work, with more than 80% of code merged by AI in 2026, up from single digits in early 2025. This trend aligns with broader industry observations of AI’s accelerating performance, though the pace of internal research automation remains less transparent and more difficult to quantify precisely.
“The evidence from Anthropic suggests that AI is already automating significant parts of its own development, but the critical question is whether it can fully automate the decision-making that guides its progress.”
— Thorsten Meyer, AI researcher
Uncertainties Surrounding Full Self-Improvement
It remains unclear whether AI will eventually automate the decision-making processes involved in research and development entirely, or if human oversight will persist. The evidence shows rapid capability gains but does not confirm that AI can autonomously design its successors or determine research priorities without human input. The timeline and likelihood of such a breakthrough are still uncertain and depend on future technological and organizational developments.
Monitoring AI Development and Regulatory Responses
Further transparency from labs regarding internal progress and benchmarks will be crucial to assess the risk of autonomous self-improvement. Researchers and policymakers will likely focus on establishing safeguards and monitoring mechanisms, especially if AI begins to demonstrate capabilities for autonomous research decision-making. The next milestones include observing whether internal data continues its upward trend and how the industry responds to these developments in terms of regulation and safety protocols.
Key Questions
What evidence supports the claim that AI is automating its own development?
Anthropic’s internal data shows that AI models like Claude are now responsible for a majority of code contributions, and public benchmarks demonstrate rapid capability improvements in tasks related to research and coding, indicating increasing automation.
Can AI fully automate recursive self-improvement now?
No, the report states that while capabilities are advancing rapidly, the critical bottleneck—decision-making about research goals and priorities—remains human-controlled. Full automation of self-improvement is not yet achieved.
Why is this development significant for AI safety?
If AI begins to autonomously improve itself at accelerating rates, it could outpace human oversight, raising concerns about control, safety, and alignment. Understanding the current pace helps inform safety measures and regulations.
What are the main limitations of the current evidence?
The internal data on AI-driven research progress is less transparent and harder to verify independently. Benchmarks measure task performance but do not directly capture the internal pace of development within labs.
What should we watch for next in AI development?
Future indicators include continued increases in internal automation metrics, new benchmarks measuring AI’s ability to set research goals, and regulatory or organizational responses to these technological advances.
Source: ThorstenMeyerAI.com
