A study found that Anthropic's AI models improved in performance nearly twice as fast when assisted by another AI.

What it means

An independent study by the research group METR found that Anthropic's AI models improved in capability roughly twice as fast over an 11-week period compared to the previous year, with much of that acceleration attributed to the models themselves helping with coding and experimentation tasks during development.

Accuracy & sourcesGrounded

The claim originates from METR, a credible AI-safety research organization, and is reported by multiple outlets (Reuters, AP) that describe confidential access to Anthropic's performance data. As a secondary report without direct citation, the specific methodology and effect size should be treated as illustrative rather than precise, but the core direction is well-sourced.

How new

Genuinely new — the METR study and associated news articles were published on November 5, 2026.

Go deeper

• Channel News Asia — first to publish