"in the future, Agents should act for you without being asked" OpenAI Isa Fulford says the next big focus is improving how we collaborate with AI agents Multi-turn chats work well, but the goal is smoother, more continuous interactions personalization and memory will be key.

this should be a bigger deal than winning IMO openAI researcher said: "their IMO model "knew"; it didn't have the right answer" does he mean hallucination-free? if models know what they know and when to ask for human help — we could finally trust them with real-world tasks

AI progress is speeding up at all major labs: xAI, openAI, Google, China no single lab will dominate because new methods will quickly be copied or reverse-engineered there is no wall. and it's only a matter of time until intelligence is fully solved

Zuck seems desperate: - meta $300m/4yr offer still isn't enough to poach some openAI top researchers - zuck even pitched Mark Chen (oAI CRO) with a $1B offer woah... that shows panic mode and the level of resistance suggests openAI is working on something big

Grok 4 feels much more human than other AIs after using gemini 2.5 pro, o3, and others, Grok felt more natural and less robotic DeepSearch is highly powerful and unbiased, no point in using google search anymore these are the things that matter most to non-technical users

i'm convinced we've already crossed the AGI threshold the term "AGI" has lost its meaning every time we get close, the goalposts shift maybe only a superintelligence can look back and define what AGI truly was — otherwise, we'll be stuck debating it forever

openAI Noam Brown confirmed 2 key points: 1) the model that reached IMO gold-level performance is general-purpose — not fine-tuned for math, and achieved it without using tools 2) openAI developed new techniques that make LLMs a lot better could it be a new RL breakthrough or an algorithmic improvement?

BIG NEWS openAI experimental reasoning model reached IMO gold-level performance — and they are also releasing GPT-5 soon new Q-star moment? looks like they're using new reasoning methods that go beyond standard CoT techniques it also confirms they're running internal models months ahead of anything public

Eric Schmidt asks what happens when AI creates millions of polymaths History's greatest discoveries came from experts applying ideas from one field to another — something AI can't do yet "the rules keep changing" solving non-stationarity might bring breakthroughs, but it'll take years to know

wait... openAI employees said this is a new model, not just a new feature we trained and fine-tuned it with RL to handle agent-like tasks. so, does that mean it's GPT-5 or not?