📊 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 internal data shows AI models are now automating significant parts of AI development, with potential for recursive self-improvement if human oversight diminishes. The evidence is based on public benchmarks and internal metrics.
Anthropic’s new report presents evidence that AI models are increasingly capable of automating core tasks in AI research and development, suggesting a potential path toward recursive self-improvement. This development is significant because it indicates that AI could soon accelerate its own evolution without human intervention, a prospect that raises both technical and ethical questions.
The report, published by The Anthropic Institute, draws on both public benchmark data and internal metrics to demonstrate rapid progress in AI capabilities. For example, models like Claude now perform complex coding and research tasks that previously required human effort, with some metrics showing a tenfold increase in code production over the past 15 months. The key measure, METR, indicates that AI can now handle tasks ranging from minutes to hours, with projections suggesting that in the next few years, AI could undertake days-long or even week-long research tasks.
Anthropic’s internal data reveals that over 80% of the code merged into their projects since early 2025 was authored by AI models, a stark increase from single-digit percentages before that. While models are already capable of executing well-specified experiments and generating code, they still lag in autonomous goal-setting and strategic decision-making, which remains a human responsibility. The authors emphasize that this gap is the critical factor that determines whether recursive self-improvement becomes inevitable.
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.
Coding with AI For Dummies (For Dummies: Learning Made Easy)
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
CLAUDE AI UNLEASHED From First Prompts to Pro: The Complete Guide to Claude AI for Writing, Research, Coding, and Business (The Claude AI Mastery Series)
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.
Learning Resources STEM Simple Machines Activity Set – STEM Engineering Kits for Kids, Simple Machines Kit for Classroom, Pulley Machine, Engineering Activities, Force and Motion Science Kit
EXPLORES SIMPLE MACHINES & ENGINEERING CONCEPTS: Handson STEM activity set introduces kids to simple machines like levers, pulleys,…
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
The Lights in the Tunnel: Automation, Accelerating Technology and the Economy of the Future
Used Book in Good Condition
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 Autonomy in Research and Development
This evidence suggests that AI systems are approaching a level where they can significantly contribute to, or even lead, their own development processes. If the trend continues and the human oversight in goal-setting diminishes, AI could enter a feedback loop of self-improvement, dramatically accelerating progress in AI capabilities. Such a scenario could reshape the landscape of AI research, impact technological innovation, and raise important questions about safety, control, and ethical oversight.
Progress and Benchmarks in AI Self-Development
Over the past few years, AI capabilities have steadily improved, with benchmarks like METR showing faster doubling times for AI’s ability to perform complex tasks. Public data indicates that models are now handling tasks that once took days in a matter of hours, and benchmarks like SWE-bench and CORE-Bench demonstrate rapid gains in coding and research reproducibility. However, these metrics primarily measure task performance, not the internal rate of progress within AI labs, where proprietary data suggests even more rapid advancements.
Until now, the dominant view was that AI progress depended heavily on human researchers designing experiments and setting goals. The new internal data from Anthropic challenges this view by showing AI models increasingly taking on roles traditionally reserved for humans, especially in coding and experiment execution, while the strategic decision-making gap remains a key bottleneck.
“The evidence from Anthropic suggests that AI is already automating a significant portion of its own development, at least in execution tasks. The real question is whether it can soon take over the strategic, goal-setting aspects as well.”
— Thorsten Meyer, AI researcher
Unresolved Questions About Autonomous Goal-Setting
It remains unclear whether AI systems will soon be able to autonomously set research goals at a level comparable to human experts. The internal data shows progress in executing tasks but does not confirm that models can independently identify priorities or strategic directions without human oversight. The timeline and feasibility of achieving this level of autonomy are still uncertain and subject to further technological development.
Next Steps in Monitoring AI Self-Development Progress
Researchers and industry observers will focus on tracking internal metrics from AI labs, particularly how models evolve in goal-setting and strategic planning. Transparency initiatives may also emerge to better understand AI’s internal decision-making processes. Additionally, further research will assess whether the current technological trajectory can sustain the rapid acceleration toward recursive self-improvement, and discussions around safety protocols are likely to intensify.
Key Questions
What is recursive self-improvement in AI?
Recursive self-improvement refers to AI systems improving their own capabilities and design without human intervention, potentially leading to rapid, exponential progress.
How does Anthropic measure AI’s progress in self-development?
Anthropic uses benchmarks like METR to track task completion times, internal metrics on code authorship, and performance on research reproducibility tests to gauge AI capabilities.
Is AI currently capable of designing its own research goals?
According to the report, AI models are not yet capable of autonomously setting research goals; they excel at executing specified tasks but still depend on human decision-making for strategic directions.
What are the risks if AI begins self-improving rapidly?
Rapid self-improvement could lead to AI systems evolving beyond human control or understanding, raising safety, ethical, and governance concerns that need to be addressed proactively.
When might AI achieve full recursive self-improvement?
It is currently uncertain; while progress is promising, experts differ on timelines, and achieving autonomous goal-setting remains a significant challenge.
Source: ThorstenMeyerAI.com
