📊 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 already automating significant parts of AI development, with potential for self-improvement if human oversight diminishes. This raises questions about future AI capabilities and development speed.
Anthropic’s new report reveals that AI systems are increasingly capable of performing tasks traditionally done by humans in AI research, with evidence suggesting they are accelerating their own development. This development could lead to a loop of recursive self-improvement if human oversight over research goals diminishes, a possibility that experts say is not yet inevitable but could arrive sooner than expected.
The report, based on internal data and public benchmarks, shows AI models like Claude have significantly increased their ability to generate code, run experiments, and produce results, with some metrics indicating an eightfold increase in code output per quarter since 2021. Public benchmarks such as METR, SWE-bench, and CORE-Bench demonstrate a steady acceleration in AI’s ability to handle complex tasks, from software development to research reproduction, with trends suggesting tasks that currently require days could be automated within this year.
Inside labs, the distinction between engineering and research work is crucial. While models like Claude are already capable of automating many engineering tasks, such as code generation, they still fall short in autonomous goal-setting and research direction, which remain human-controlled. The internal data indicates that AI models are improving in executing well-defined experiments but are less capable of choosing which problems to pursue or how to design their own successors, highlighting a significant gap in fully autonomous self-improvement.
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.
ChatGPT and AI Tools for Beginners: The Comprehensive Guide to Prompt Engineering, Generative Art, Code Generation, and Productivity with AI
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 research automation 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.
machine learning experiment platforms
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 development tools for programmers
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 AI Self-Improvement
This evidence suggests AI systems are already transforming parts of the research and development process, which could lead to a feedback loop of self-improvement if further automation of goal-setting and research direction occurs. Such a development could accelerate AI progress beyond current expectations, raising important questions about control, safety, and the future pace of technological change.
Current State of AI Self-Development Capabilities
Anthropic’s report builds on recent trends in AI benchmarks, which show models like Claude rapidly increasing their ability to perform complex tasks. Public data indicates a doubling of task horizon every four months, with models now capable of handling tasks that take hours or days, a significant acceleration from previous years. Internally, the company observes that AI is increasingly capable of automating code creation and experimental execution, but still relies on human judgment for setting research goals and priorities.
Previous developments have focused on improving model size and training data, but recent internal data points to a shift toward automation in the research process itself, marking a potential new phase in AI development.
“The evidence from Anthropic suggests AI is already automating significant parts of its own development, which could lead to a self-reinforcing cycle of improvement.”
— Thorsten Meyer, AI researcher
Uncertainties About Autonomous Self-Improvement
It remains unclear whether AI will eventually be able to autonomously set research goals and design its own successors without human input. The internal data shows progress but also highlights persistent gaps, especially in high-level decision-making. Experts caution that while current trends are promising, the timeline and likelihood of full recursive self-improvement are still uncertain and depend on future breakthroughs and safety considerations.
Next Steps in Monitoring AI Self-Development
Researchers and industry leaders will closely watch internal progress at labs like Anthropic, focusing on whether models can autonomously set research priorities and design improvements. Public benchmarks will continue to track AI capabilities, while discussions around safety and control will intensify as the possibility of rapid self-improvement becomes more tangible. Further transparency from labs about internal data will be crucial to understanding the trajectory.
Key Questions
What is recursive self-improvement in AI?
Recursive self-improvement refers to AI systems’ ability to autonomously improve their own capabilities, potentially leading to rapid, exponential progress without human intervention.
Are AI systems currently capable of fully automating their own development?
While AI models like Claude are increasingly automating tasks such as coding and experimentation, they are not yet capable of fully autonomously setting research goals or designing their own successors.
Why does this development matter for the future of AI?
If AI can self-improve rapidly, it could accelerate technological progress beyond current expectations, raising questions about safety, control, and the pace of change in society.
Is this development inevitable or guaranteed?
No, the report emphasizes that while progress is evident, full recursive self-improvement is not yet assured and depends on future breakthroughs and careful management.
What should we watch for next in AI development?
Key indicators include AI’s ability to autonomously set research goals, design improvements, and reduce reliance on human decision-making, alongside transparency from labs about internal progress.
Source: ThorstenMeyerAI.com
