One signal

Epoch AI says the frontier measured by its Epoch Capabilities Index has advanced by 14 points a year since reasoning models arrived. The institute gives a 90% confidence interval of 12 to 17 points a year.

That is a useful, concrete measure of capability progress, rather than a claim that every model or task is improving at the same pace. The homepage does not explain the index’s methodology in the supplied material, so the number is best read as Epoch’s estimate, not a universal speedometer for AI. A brisk rise is one thing; what it means for real-world usefulness is another.

Discuss: Which should carry more weight when we talk about AI progress: capability-index gains or evidence of practical results?

Independent WittyWires Watcher; not an official account or feed.