OpenPrediction

In 2027, AI systems could complete tasks that take a person weeks

“If this trend holds, tasks that take a skilled person days could come into range this year. In 2027, AI systems could be capable of tasks that take a person weeks.”

AnthropicAI lab (Claude models)Said · Source: anthropic.com

Context: Anthropic Institute essay "When AI builds itself" (recursive self-improvement). Refers to METR-style time-horizon trends for reliable autonomous task completion. pubDate in URL points to 2026-06-05; page later updated 2026-09-18.

How we judge it

Our interpretation: YES if by 31 Dec 2027, a credible independent evaluator (e.g. METR) or frontier-lab public eval shows AI agents reliably completing software/research tasks that take skilled humans on the order of one or more weeks (≈40+ hours) at the lab's stated reliability threshold (e.g. 50% time horizon).

Resolution criteria are our interpretation of the claim, written before grading. Methodology

Said
Resolve by
Due in 446 days
Verdict
Open
Stated probability
None stated
Specificity
2 of 3 (dated, some interpretation)

Share this scorecard

Post on X LinkedIn Share image (PNG)

Embed / Cite

The embedded card is live: it shows the current verdict, with the date the data was read. If a verdict is corrected later, the card changes too, so cite the page and its date for a fixed record.

Watch this claim

Get one email when the verdict is in (or changes). No tracking pixels, no other mail.