Unverified50% confidenceBenchmarkExact time
Meter's assessment found that the rate of capability improvement of Anthropic's models on well-defined tasks (such as tasks related to the AI R&D process) has reached about twice that of some previous historical interval, and the current rate may be closer to 1.3 to 3 times the historical rate.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
11/7/2026
First Seen
Valid until: 2/5/2027
Sources
Related Entities
Related Claims
UnverifiedThe maximum time AI can continuously make progress on a single task has grown from a few seconds to approximately one hour over the past one to two years.72% similarUnverifiedMETR conducted an eleven-week review of Anthropic's internal data and concluded that over the past year Anthropic's progress on AI testing amounted to the sum of the previous two years, with the R&D pace roughly doubling.69% similarUnverifiedGreenblatt's starting point for deriving the 4.5x acceleration is that current AI progress is overall about 2x faster because of RSI.68% similarUnverifiedThe evolved program is 3.8 times faster than the starting program.66% similarUnverifiedBrown observed that each time AI crosses a mathematical benchmark threshold, the human time required for the tasks it can handle increases by approximately 10x (GSM8K ~5 seconds, MATH ~1 minute, olympiad qualification level ~10 minutes).66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1146754API
curl https://kongchang.com/api/v1/knowledge/claims/1146754MCP
get_claim(id=1146754)