Tree News

DECRYPT · 2026-11-18 16:57 UTC

'It Speaks in Tongues That Only 6.2 Astra Understands': Why Early Testers Hate OpenAI's New Model

OpenAI released GPT-6.2 Astra and Sol on Tuesday, promising smarter agents, better coding and creative abilities. Early testers such as Theo Browne called Astra hard to work with, saying its outputs were unparsable, it looped without progress, and it made a compiler 56% slower. Theo and another tester got near-identical 'Bellwether' water-and-bell levels for different games, and Matt Shumer called the model 'utterly unusable' for serious work.

Loading the story on Tree News…