← Back to digest
Models & Releases

The rollout was a mess

What's this about?

The launch itself stumbled: the press embargo lifted at 11am Pacific, articles went live — and OpenAI's own blog post wasn't up for nearly an hour, apparently knocked off schedule by the morning's outages. Access rolled out to limited orgs first; sama apologized twice; OpenAI promised paying subscribers one banked usage reset for every day they lack access.

What exactly is a press embargo?

Journalists get materials early and agree to publish at a set time, so coverage lands simultaneously. It assumes the company's own announcement lands at the same instant — when it doesn't, the world reads about a launch the company hasn't made yet.

What's the community saying?

HN 248 points on the rollout thread alone
  1. Live confusion, documented in real time: Reuters at 2:03pm, no blog post at 2:40pm, commenters unable to verify the model existed beyond past-tense press claims.
  2. The irony, in one comment: "on the one hand we have AGI; on the other, the release page is returning 500s."
  3. A pre-written launch checklist as satire: "1) Astra will win all benchmarks. 2) The pelican will have a basket. 3) Cyber is too dangerous to release."
  4. Quota math from heavy users comparing how many days a 20× subscription survives each model tier.
← Read the comparison table: Fable still leads three indexesNotes across context windows in Codex →