Gemini just landed in the Chrome sidebar for everyone, not only paid AI users. You can now summarize any page, ask questions about what you are reading, or draft an email without leaving the tab. It also pulls context from the page itself, so prompts are not starting from zero. For anyone building on top of AI workflows, this is a major distribution event. Google just put an assistant in front of roughly three billion Chrome users by default, and you do not need an account to use it.

xAI released the base weights for Grok 3 and put them on Hugging Face for free download. Most frontier labs keep the big models locked up. xAI is going the other way and betting that open weights win the developer mindshare race. If you have been waiting to run a real frontier model on your own hardware, this is the moment. Pull the weights, load them in your stack of choice, and start tinkering this weekend.

Claude Cowork is now on your phone. Anthropic shipped the mobile and web versions today. The background agent that runs multi-step work on your desktop is now reachable from iPhone, Android, or any browser. For builders who live in their phone half the day, this is the unlock. Start a research job at the desk, walk out, and the result is waiting when you tap the app. A genuine shift in how an ambient agent should feel.

Microsoft is starting to swap OpenAI and Anthropic out of its own products. Bloomberg reported that Microsoft is putting its own in-house models into Copilot features across Office and Windows. At Microsoft scale, paying per token to another vendor stops adding up. If you build with Copilot, do not assume the model underneath will stay the same for long. Build your prompts and workflows so they are easy to retarget when the engine changes underneath you.

GPT-5.6 is finally rolling out wide in the US, no more staggered drops. OpenAI got the all-clear from Washington this morning. After months of waiting and limited access, anyone who wants to build on 5.6 can start this week. If you have been holding off on GPT-5.6 because access was patchy, the gates are open. Worth a serious test drive if your stack can use the extra reasoning power.