An institute for one question.
The Centre for AGI exists to track a single thing rigorously: the rise of artificial general intelligence and superintelligence — and whether the methods we have today can actually get us there.
Discussion of AGI is awash in confident timelines and competing narratives, most of them untethered from anything you can check. We think the field deserves the same treatment serious empirical questions get everywhere else: clear definitions, tracked metrics, honest extrapolation, and a willingness to be proven wrong by the data.
Modelled on the empirical, data-first spirit of organisations like Epoch AI — but focused exclusively on AGI and superintelligence — we publish independent analysis and curate the papers that genuinely advance the question.
What we do
We do three things, in order: measure capability with defensible metrics, forecast where the trends point while flagging every extrapolation as a hypothesis, and inform the researchers and decision-makers who need to read the frontier accurately.
Principles
Evidence over narrative
Every claim is tied to a measurable proxy and a stated source. Where we speculate, we say so and attach our uncertainty.
Independence
No lab affiliation, no product to sell. Our only incentive is to read the frontier accurately, including when that means saying progress has stalled.
Legible methodology
If you can’t see how a number was built, you can’t trust it. We publish the composition of every index and the reasoning behind every forecast.
Graceful definitions
Benchmarks wear out. We track capabilities, not leaderboards, and replace instruments without moving the conceptual goalposts.
Who’s behind it
The Centre is an independent project founded in 2026 by a machine learning researcher. Replace this section with your own bio — your background, research interests, and what drew you to the AGI question.
Get in touch
Corrections, collaborations, and pointers to papers we’ve missed are all welcome.