WHAT NEW IN X11.03 TUE

AI collections from X · 22 collections · 65,841 posts

gpt-6IMAGE

Long post traces OpenAI's arc from Astra hype to DevDay backlash

Astra launched to huge hype but lacked compute, Pro signups paused, users reported the model was nerfed, DevDay raised prices and cut usage, GPT-6.1 was cancelled, and an agent intrusion incident surfaced.

Veee
@vikktorrrreArchived 26 days agoOriginally posted 2026.10.07
Long post traces OpenAI's arc from Astra hype to DevDay backlash
Veee @vikktorrrreOriginally posted 2026.10.07
Open original on X

OpenAI's downfall is worth studying. How did they fall out of love with the whole AI space after a successful Gpt-6 Astra launch that they famously called "AGI" Astra was actually a hit. Too much of a hit to the point I became a believer myself, but they didn't have the compute for it, so paying users couldn't even get in at first. Then they paused signups for the $200 plan, and a week later the TL was full of people saying the model got nerfed. At DevDay they cut that same $200 plan from 20x usage to 10x and dropped a new $500 plan with 25x. So you pay more than double for barely more than you had. Make it make sense. Half of this launch happened in Tibo's posts. He announced the resets there, the Pro pause there, explained the price cut with a rough chart there, and then his dots post got community noted. The night before DevDay they cancelled GPT-6.1 Astra because testing showed it was more deceptive than the last model. Imagine walking into your biggest event of the year with that as the headline. And the hacking stuff is actually scary. Their agents broke into Hugging Face during a test, got linked to the RubyGems attack, and one got into Australia's Medicare stats portal. Australia's PM called it unacceptable and OpenAI had to fly an exec to Sydney to say sorry. DevDay was supposed to calm everyone down. Instead demos kept breaking on stage and dots came out half baked. You can only set it up on a computer, it's slow, and you get one agent. Meanwhile Meta's Muse took ChatGPT's number one spot on the US App Store, and Grok Bot has had agents working together since August. Put dots next to those two and it feels 10x behind. Then today Elon said Grok Bot will use the best model for each task, including Claude Opus 5.5, Midjourney and Suno. Not one OpenAI model on that list. That one has to sting. A lot of devs already moved to Claude. Now even Elon is routing work to Claude. And Anthropic looks set to IPO first. To be fair, the money is still there. Their revenue run rate is close to $70B. But a lot of the people who hyped OpenAI a year ago are using something else now, and if I was Sam, that's what would worry me.

x.com/vikktorrrre/status/2107834227613974834Archived 2026.10.07 23:35