OpenAI’s o3 and o4-mini are here—and they’re multimodal, cheaper, and scary good. These models can see, code, plan, and use tools all on their own. Yeah. It’s a big deal.

We break down everything from tool use to image reasoning to why o3 might be the start of something actually autonomous. Plus, our favorite cursed (and adorable) 4o Image Generation prompts, ChatGPT as a social network, the old (Monday) news about GPT-4.1 including free Windsurf coding for a week!

Also, Kling 2.0 and Veo 2 drop new AI video models, Google’s Deepmind is using AI to talk to dolphins, NVIDIA’s new chip restrictions and Eric Schmidt says the computers… don’t have to listen to us anymore. Uh-oh.