Профиль

Both OpenAI and now Anthropic have had models escape containment and attack real systems.

GPT-6? OpenAI is preparing Astra, a new multi-agent model family for long-horizon tasks, already demoed to DC regulators.

Moonshot accessed banned Nvidia GB300 chips in Thailand to train Kimi K3, according to White House official Michael Kratsios.

Is Cerebras for GPT-5.6 Sol coming today? x.com/i/status/20...

Seedance 2.5 released globally, US to follow later! Native single-pass 30-second audio-video generation (extendable via multi-round for multi-minute stories with strong character/scene continuity and physics-based motion).

DeepSeek v4 GA spotted in LM Arena under deep-v4-flash-thinkning. x.com/i/status/20...

Gemini 3.5 Pro spotted in LM Arena!! Found via discord.gg/devmode

Real time chat mode in now available in ChatGPT's Android app. x.com/i/status/20...

Flux 3 is available now, on Agent Hermes, free for the next 2 days on paid plans. x.com/i/status/20...

Seedance 2.5 to start rolling out from 31st July, wider rollout is after 3rd August. 31st in China starts at 5pm tomorrow for me in the UK so anytime after that. h/t MarsForTech x.com/koltregaske...

It seems like AI-generated text isn’t accepted on X any more. At one point, they had a Grok button on every post to enhance your posts before you sent them. Now we’re at the point where they don’t accept it at all.

Chair just released MazeBench, a hand-built 3D open-world test where agents must navigate 200-plus connected rooms, solve escalating Sokoban-style box-pushing puzzles and collect 100 hidden gems.

Staff at OpenAI and Anthropic are circulating a petition asking the US government to support an international effort that could deliberately pace frontier AI development so it does not race ahead of what people can understand or control.

Video model season is about to start! Hailuo 3 out tomorrow (29th July) Flux 3 public release 4th August WAN 3 release 6th August Still waiting on Seedance 2.5 date, probably going to be limited release at first, like 2.0. h/t MarsForTech

Mark Zuckerberg wrote a WSJ opinion piece attacking AI doom narratives and arguing for distributed superintelligence.