are we AGI yet?

Not yet,
but we are close.

There are still things humans do better.

Five stages of a blue circle labelled 'Tasks of a human job' with a pink blob labelled 'Tasks an AI can do' growing inside it, from 'The AI is a fun toy' through 'The AI has a jagged frontier' (marked 'We are here') to 'AGI', where the spiky pink shape swallows the circle entirely.
Jagged Frontier - Tomas Pueyo, X (2025)

In recent news...

  1. A proposed proof of Navier–Stokes

    OpenAI’s unreleased model proposes a proof for the Millennium Prize problem, with a paper and Lean formalization.

    OpenAI

  2. A map of 9 billion DNA changes

    DeepMind’s AlphaGenome Atlas predicts the molecular effects of 9 billion possible DNA changes.

    Google DeepMind

  3. AI is helping build the next AI

    OpenAI reports agents handling research tasks that take humans days, with human direction.

    OpenAI

  4. Astra makes a leap in unfamiliar games

    ARC-AGI-3: 99.9% with a provider-adapted setup; 62.7% with the standard setup. Neither proves AGI.

    ARC Prize

  5. Claude advances a Riemann-related bound

    Anthropic reports a new bound of 67.2%, up from 41.6%—progress toward the Riemann hypothesis, not a solution.

    Anthropic