Skip to content
SUPREME AI AGI

Method

How we predict

Anyone can make claims about AGI. We try to make ours checkable: evidence first, plain statements, confidence levels and public reviews.
01

Follow the evidence

We read papers, benchmarks and independent reports from teams around the world. A result is listed only with a working link to its source.

02

Say it plainly

Each prediction is one sentence that can later be judged right or wrong, with a time horizon and a confidence level.

03

Show the basis

Every prediction links to the public results it rests on and says what would prove us wrong.

04

Review in the open

When new results appear we update the status and the review date. Wrong predictions stay on the site, marked wrong.

Time horizons

  1. Near termWithin about 3 years of the date the prediction was made.
  2. Medium termAbout 3 to 10 years out.
  3. Long termMore than 10 years out, or no clear timeline yet.

Prediction status

  1. OpenNot yet decided either way.
  2. On trackNew evidence points the way we predicted.
  3. HappenedIt came true. The review note says when and how.
  4. WrongIt did not happen or evidence went the other way. We keep it listed.

Stages of our own research

Every project we list carries one of these stages.

  1. Research interestA question we want to study. No experiments yet.
  2. Work in progressBeing built or trained now. No results claimed.
  3. Experiments runInternal experiments done. Results not yet published.
  4. Documented resultMethod, data and limits written up and public.
  5. Peer reviewedReviewed and published by an independent venue.

How we publish

  • We list only real work. No invented papers, numbers, partners or results.
  • Other teams’ results are described in our own words and always credited and linked.
  • Predictions are labelled as our view, never as fact.
  • Our own projects carry an honest stage; we do not claim results we have not published.
  • Every publication names its authors and date and states its method and limitations.
  • When AI tools substantially helped write a text, the page says so.
  • User chats and files are private and never used for research or training without opt-in.