📊 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 reports measurable acceleration in AI’s ability to automate its own development tasks, with internal data indicating significant progress toward recursive self-improvement. The evidence suggests AI is already taking on more research and coding roles, but key decision-making remains human-controlled.
Anthropic has released new internal data indicating that AI systems, specifically their models, are already automating a growing portion of their own development tasks, including coding and experimentation. This suggests progress toward what is known as recursive self-improvement, a process where AI could potentially improve itself at a pace faster than human intervention. While the authors emphasize that this is not yet happening at an exponential or fully autonomous level, the evidence points to rapid advancements that could reshape AI development timelines.
The report from Anthropic’s Institute presents data showing that AI models, such as Claude, are increasingly capable of performing tasks traditionally done by human researchers and engineers. For example, more than 80% of new code integrated into Anthropic’s projects since early 2025 has been authored by AI, up from single digits before that period. Public benchmarks like METR, SWE-bench, and CORE-Bench demonstrate that AI models are rapidly improving in tasks like code fixing, experimental reproduction, and complex problem-solving.
Inside the labs, Anthropic’s researchers observe that models can now handle tasks that previously required days of human effort, with some models capable of managing 12-hour tasks and potentially longer in the near future. Data also shows that models are improving in the ‘middle rungs’ of research — designing experiments and interpreting results — but still lag in making high-level strategic decisions about which problems to pursue. The authors highlight that this ongoing progress could, if sustained, lead to a point where AI begins to autonomously design and improve its successors, although this remains a conditional possibility.
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
AI Engineering: Building Applications with Foundation Models
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
AI Systems Performance Engineering: Optimizing Model Training and Inference Workloads with GPUs, CUDA, and PyTorch
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 Shift Toward Autonomous AI Development
This development matters because it suggests that AI systems are increasingly capable of automating significant portions of their own creation and refinement processes. If these trends continue, it could accelerate the timeline toward AI systems that self-improve without human intervention, raising questions about control, safety, and the future pace of technological change. The evidence from inside Anthropic indicates that the bottleneck is shifting from technical capability to decision-making, which remains largely human-controlled for now.
Current State of AI Self-Development Capabilities
Over the past few years, AI research has shown steady improvements in models’ abilities to perform complex tasks, with benchmarks indicating rapid progress. Anthropic’s internal data reveals that these improvements are now translating into increased automation in research and development activities. The concept of recursive self-improvement has been discussed theoretically for decades, but concrete evidence of it happening at scale has been limited. This report provides some of the first data suggesting that AI is already capable of automating parts of its own development, though not yet at a fully autonomous or exponential rate.
“The internal data from Anthropic suggests that AI models are increasingly taking over tasks that were once exclusively human, hinting at the early stages of recursive self-improvement.”
— Thorsten Meyer, AI researcher
Unconfirmed Aspects of Fully Autonomous Self-Improvement
It remains unclear whether AI can autonomously decide which problems to pursue or design entirely new architectures without human input. The current evidence shows progress in automating research tasks but does not confirm that AI systems are capable of recursive self-improvement at an exponential or fully autonomous level. Experts caution that the bottleneck in decision-making and strategic planning still largely resides with humans, and whether this will change remains uncertain.
Next Steps in Monitoring AI Self-Development Trends
Researchers and industry observers will closely watch internal performance metrics and benchmark results from labs like Anthropic to determine if the trend of increasing automation continues. Future developments may include more autonomous AI systems capable of designing experiments, optimizing architectures, and perhaps even creating new AI models without human guidance. Public disclosure of internal data and transparency about capabilities will be critical to assess the trajectory of recursive self-improvement.
Key Questions
What is recursive self-improvement in AI?
Recursive self-improvement refers to an AI system’s ability to improve its own design and capabilities autonomously, potentially leading to rapid, exponential progress without human intervention.
Does the new data confirm that AI is already self-improving?
The data indicates AI systems are automating many research and development tasks, but it does not confirm they are fully self-improving or capable of designing and implementing entirely new architectures independently.
Why is this development important for the future of AI?
If AI systems can autonomously improve themselves, it could accelerate technological progress dramatically, raising questions about control, safety, and the pace of change in AI capabilities.
Are there risks associated with AI self-improvement?
Potential risks include loss of human oversight, unpredictable behavior, and rapid escalation of capabilities. Experts emphasize the importance of monitoring and regulation as progress continues.
What should we expect next from AI research labs?
Labs will likely publish more internal metrics and benchmarks, and we may see the development of more autonomous AI systems capable of designing and optimizing their own architectures, pending further evidence.
Source: ThorstenMeyerAI.com
