📊 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 suggesting AI models are already automating significant parts of AI development, with potential for self-improvement if human oversight on goal-setting diminishes. The evidence is based on internal metrics and public benchmarks, but key gaps remain.
Anthropic’s new report reveals that current AI models are already automating substantial portions of AI research and development, with data showing rapid performance improvements. This suggests that, if certain human-led decision points are automated, AI could enter a loop of self-improvement at the speed of compute, a development that could accelerate AI progress significantly.
The report from Anthropic’s Institute states that AI systems like Claude now handle over 80% of code contributions in their projects, up from low single digits in early 2025. Public benchmarks such as METR and SWE-bench show that AI capabilities are doubling roughly every four months, with models increasingly able to perform complex software tasks and reproduce research results.
Inside labs, data indicates that AI models are improving in both engineering and research roles. For instance, Claude can now autonomously find methods to fix code issues and perform experiments, although it still struggles with goal-setting and high-level decision-making. The authors emphasize that while the pace of technical capability growth is clear, the bottleneck remains in AI’s ability to decide what problems to pursue, which is still human-controlled.
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.
![Claude AI for Beginners Bible: [5 in 1] The Ultimate Guide to Automate Your Work, Save Hours Every Week, and Use AI for Real-World Results](https://m.media-amazon.com/images/I/415+fSJacsL._SL500_.jpg)
Claude AI for Beginners Bible: [5 in 1] The Ultimate Guide to Automate Your Work, Save Hours Every Week, and Use AI for Real-World Results
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 VoiceWriter – Smart Dictation & AI Writing Assistant for Windows & Mac | USB Dongle & Mobile App for Voice Input, Proofreading, Rewriting & Multilingual Support
🎙️ Hands-Free Voice Typing for Windows & Mac – Powered by iOS & Android dictation technology, AI VoiceWriter…
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.
Agent-Powered Growth: Deploy AI Agents That Build Your Marketing Pipeline 24/7
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.
Implications of AI Self-Improvement Potential
This development matters because it suggests that AI could soon automate not just tasks but the process of improving itself, potentially leading to rapid, exponential growth in AI capabilities. Such a shift could reshape research timelines, innovation cycles, and safety considerations, making it a critical area for monitoring by policymakers and researchers.
Current Evidence of AI Progress and Automation
Anthropic’s findings are based on internal metrics and public benchmarks that show AI models rapidly increasing their ability to perform tasks like coding, bug fixing, and reproducing research results. Public data such as METR indicates a doubling of AI task proficiency every four months, a trend that has been consistent since 2024. The internal data reveals that AI now handles a majority of coding work, a significant jump from previous years.
While these improvements are tangible, the capacity for AI to autonomously decide research directions or design new models remains limited, with the authors emphasizing that human decision-making still dominates this aspect.
“The evidence shows that AI is already automating substantial parts of AI development, and if the last human bottleneck—goal-setting—falls, we could see a loop of recursive self-improvement.”
— Thorsten Meyer, author of the report
Uncertainties Surrounding Autonomous Goal-Setting
It is not yet clear when or if AI will fully automate the goal-setting and decision-making aspects of research, which are currently human-led. The evidence suggests progress but does not confirm that recursive self-improvement will happen automatically or imminently.
Monitoring AI Progress and Regulatory Responses
Researchers and policymakers will likely focus on tracking further developments in AI’s autonomous capabilities, especially in goal-setting. Future work will include more detailed internal metrics, safety assessments, and discussions on regulation to prepare for potential rapid advancements.
Key Questions
What is recursive self-improvement in AI?
Recursive self-improvement refers to AI systems autonomously improving their own capabilities, potentially leading to rapid, exponential growth in intelligence and performance.
How close is AI to automating the entire research process?
Current evidence shows significant automation in coding and experimentation, but AI still relies on human decision-making for setting goals and prioritizing research directions.
What are the risks if AI begins self-improving rapidly?
Rapid self-improvement could accelerate AI capabilities beyond human control or understanding, raising safety, ethical, and regulatory concerns that require careful monitoring.
Does this mean AI will soon replace human researchers?
While AI is increasingly automating technical tasks, human judgment remains essential for research direction and decision-making. Full automation of research leadership is not yet confirmed.
What should policymakers do in response?
Policymakers should monitor ongoing developments, promote transparency, and consider regulations to manage potential risks associated with autonomous AI self-improvement.
Source: ThorstenMeyerAI.com
