PROJECT LIGHTHOUSE
Open
The next generation of AI models (post-2025) will outperform human experts on average on the METR long-task agentic benchmark.
- Expert
- Ethan Mollick
- Prediction date
- Sep 29, 2025
- Deadline
- Dec 31, 2027
- Category
- AI & Technology
- Explicit confidence
- Not stated
Original statement
“If the current patterns hold, the next generation of AI models should beat human experts on average in this test.”
Normalized claim
The next generation of AI models (post-2025) will outperform human experts on average on the METR long-task agentic benchmark.
A falsifiable interpretation, separate from the exact source text.
Resolution criteria
METR or equivalent independent evaluations showing frontier AI models exceeding average human expert performance on multi-hour realistic tasks by end of 2027.
Market context
No high-confidence matching market is linked. Loose thematic matches are not shown.
Resolution
This prediction remains open.
Audit trailView full history
- Source recorded2025-09-29 00:00 UTC
- Prediction verified and frozen2026-09-11 04:36 UTC