AivexaNewsSearch
AI news for builders and product teamsChecked every hour

Anthropic co-founder Jack Clark moves his prediction for self-improving AI forward to 2027

Collected Oct 25, 2026

Anthropic co-founder Jack Clark has moved his prediction for self-improving AI forward, now expecting it could arrive by 2027. The revised timeline follows results from a new benchmark by P-Zero Research that measures how well AI models design and evaluate experiments on their own, and which The Decoder reports are improving quickly enough to shift Clark's view.

Self-improving AI generally refers to systems that can carry out parts of AI research itself, such as proposing experiments and judging their results, rather than only solving tasks set by people. The P-Zero Research benchmark targets that capability directly: whether a model can design and evaluate experiments without human direction. The reported pace of improvement on this measure is what prompted Clark to bring his forecast forward.

The announcement does not detail the benchmark's scoring method, the models tested or specific scores. What is clear is the direction of travel: capability on autonomous experiment design and evaluation is climbing fast enough that a prominent industry figure has shortened his timeline to 2027.

Why it matters: developers and product teams should expect shorter planning horizons for AI research tooling and for systems that take on more of their own evaluation. If progress on benchmarks like this continues, the practical questions shift toward oversight, reproducibility and how much autonomy teams grant such systems. The 2027 date is a prediction, not a demonstrated capability.

Read at The Decoder

Based on reporting from the original publisher. Visit the source for full context and later updates.

Publisher excerpt

A new benchmark by P-Zero Research measures how well AI models design and evaluate experiments on their own. The results are improving so fast that Anthropic co-founder Jack Clark now thinks self-improving AI could arrive by 2027. The article Anthropic co-founder Jack Clark moves his prediction for self-improving AI forward to 2027 appeared first on The Decoder .