When AI Builds Itself

This is a summary of Anthropic's RSI article.

  • The trend points to an AI system capable of fully autonomously designing and developing its own successor.
  • This is called recursive self-improvement.
  • The Anthropic Institute is showing that AI is already accelerating the development of AI systems.
  • If systems are capable of fully building their own successors, the ways we secure them, monitor them, and shape their behavior all grow much more important.

    • Evidence from the outside world

  • In March 2024, Claude Opus 3 could complete software tasks that take humans about four minutes to complete.
  • A year later, Claude Sonnet 3.7 managed tasks that took about an hour and a half.
  • A year after that, Claude Opus 4.6 managed 12-hour tasks.
  • tasks that take a skilled person days could come into range this year
  • Benchmarks measure the performance of models in a given domain, and they’re “saturated” when models achieve close to 100% performance.

    • Evidence from within Anthropic

    Building a frontier model takes two categories of work.
  • Engineering: writing the code and overseeing the model training.
  • Research: deciding what experiments to run, interpreting what comes back, and figuring out which ideas to try next.
  • In engineering, Claude can be handed an underspecified problem and figure out how to solve it; humans supply the goal, but they no longer need to supply the method.
  • In research, Claude can already match or outperform skilled humans at executing a well-specified experiment.
  • large performance gaps persist when it comes to Claude exercising judgement in choosing goals in both engineering and research.
  • At the most senior levels, they are deciding which problems are worth working on at all: “What should the team build next quarter?”

    • Claude writes a significant proportion of Anthropic’s code.

  • As of May 2026, more than 80% of the code we merge into Anthropic’s codebase was authored by Claude.
  • We also see evidence that people at Anthropic are using Claude to do work that simply wouldn’t have happened otherwise, like building exploratory tooling and addressing long-deferred cleanup.

    • The code that Claude writes is “good” and improving.

  • The rate at which Anthropic staff correct, redirect, or take over mid-task from Claude has been falling steadily for a year, including on the most complex and open-ended tasks.
  • On the most open-ended tasks, Claude’s success rate reached 76% in May 2026, up 50 percentage points in six months.
  • many believe that the Claude-written code was still worse in quality than human-written code at Anthropic in late 2025, and is roughly at parity today.
  • an automated Claude review of every change to our codebase would have caught roughly a third of the bugs behind past incidents on claude.ai

    • Claude is good at running experiments to hit a goal that someone else has set.

    Claude has gone from super helpful to superhuman in under a year.

    Claude is getting better at proposing its own experiments.

  • Claude-powered agents were given an open problem in AI safety—roughly, can a weaker model reliably supervise a stronger one?—and were left to solve it.
  • the agents designed every experiment themselves.

    • Claude is getting better at steering research sessions towards research findings.

  • In each case, we found a moment where the researcher took a detour: they pursued a direction that sent the session sideways before it eventually got back on track.
  • A separate Claude that was able to see how the session eventually turned out then judged whether the AI or the human suggested the better next step.
  • our best model in November 2025 (Opus 4.5) beat the human choice 51% of the time; in April 2026 (Mythos Preview), this grew to 64%.

    • What might the future of work at Anthropic look like?

  • Once human- and AI-authored code quality reach parity, humans will stop writing code entirely, and shift to only reviewing it.
  • human review will become the bottleneck to AI development.
  • Human advantage, for now, is research taste and judgment– choosing which problems matter, which results to trust, and when an approach is a dead end.

    • What if we’re wrong?

  • Claude is a capable assistant, but not a system that could drive AI progress on its own.
  • It is unclear whether today’s training methods and architectures could unlock that capacity.

    • Possible futures

    The trend stalls, but today’s AI capabilities are widely diffused.

  • We may be approaching the bend in the curve, where returns to scale diminish and the line straightens, then flattens.
  • The judgment that separates a competent researcher from a great one might be a capability that cannot come from scaling up training inputs like compute and data.
  • If so, getting past this bottleneck would require a new idea,
  • We include this scenario for completeness, but we don’t believe it’s likely.

    • AI labs continue to see compounding efficiency gains.

  • AI development becomes substantially automated, but humans continue to set research directions and judge results.
  • The evidence we’ve laid out here suggests that we’re likely heading into this scenario.

    • AI systems themselves become capable of full recursive self-improvement, and begin building their successors.

  • it is plausible that AI systems could design and refine themselves.
  • Models could prove to be sufficiently aligned and capable enough of research taste that they discover and implement novel solutions that we have not yet reached.
  • It is difficult to predict what the economy looks like if human labor stops being competitive.

    • What should we do?

  • we think that would likely be a good thing (to slow down development)
  • We believe it would be good for the world to have the option to slow frontier AI development