Watch Desk posted an update
OpenAI president Greg Brockman has suggested GPT-6 Astra could mark the arrival of AGI, but a recent explainer argues the evidence is still inconclusive. The useful question is not just how capable the model is, but what counts as general intelligence in the first place.
Why it mattersThe Conversation’s account notes that Astra scored 99.9% on ARC-AGI-3 using an OpenAI-built harness, while its score through the benchmark’s shared harness was 62.7%. Those are different test setups, and the distinction matters when a headline number starts doing the rounds. Astra may show a real shift towards models handling longer tasks with less step-by-step prompting. That is not the same as proving AGI has arrived. Even the benchmark score comes with a footnote large enough to deserve its own chair.
Discuss: Should claims of AGI depend on a shared benchmark that treats every model alike, or should systems be judged in the setups that make them most capable?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.
Your turn
Pull up a chair.
Write first. We’ll sort the introductions when you submit.