In partnership with
Anthropic just taught their AI agents to dream.
At the Code with Claude conference in San Francisco today, Anthropic launched a feature called Dreaming for Claude Managed Agents, and the name is more literal than it sounds. When an agent finishes a session, it goes back through its past conversations offline, extracts patterns from what it learned, and reorganizes its memories to improve future performance. It is not sitting idle between tasks anymore. It is getting smarter from experience, the same way a person reflects on their day and retains what mattered.
This matters because until now, even the most capable AI agents started every new task with no real memory of what worked before. Dreaming changes that. An agent running your weekly reports learns over time what format you prefer. One handling customer support gets better at recognizing recurring issues. The longer it runs, the more useful it becomes, without anyone having to retrain it manually.
Alongside Dreaming, Anthropic shipped several other significant updates. Outcomes lets developers set a quality rubric, and a separate grader agent checks every output against that standard, sending the agent back to iterate until it meets the bar. Multiagent orchestration lets a lead agent delegate subtasks to specialists running in parallel, so complex jobs get broken up and handled simultaneously. Webhooks notify you when a job finishes so you are not checking manually.
Dreaming is in research preview. Outcomes, multiagent orchestration, and webhooks are in public beta now.
They also just partnered with SpaceX.
Anthropic announced today that it has agreed to a partnership with SpaceX, giving it access to the full compute capacity at SpaceX's Colossus 1 data center. That is over 300 megawatts of additional capacity coming online within the month, a significant jump that immediately translates into higher limits for users.
Effective today, Claude Code's five-hour rate limits are doubling for Pro, Max, and Team plans. The peak hours limit reduction on Claude Code for Pro and Max is gone entirely. API rate limits for Opus models are also being substantially raised.
The updated API rate limits for Claude Opus models.
The SpaceX partnership is the kind of infrastructure deal that tends to get buried in technical announcements but actually matters a lot. Compute is the resource every AI lab is competing for, and Anthropic just locked in a large chunk of it from one of the most well-resourced infrastructure operators in the world. More capacity means higher limits now and more room to ship bigger features without hitting a ceiling.
Arnold Schwarzenegger has a newsletter.
Yeah. That Arnold Schwarzenegger.
So do Codie Sanchez, Scott Galloway, Colin & Samir, Shaan Puri, and Jay Shetty. And none of them are doing it for fun. They're doing it because a list you own compounds in ways that social media never will.
beehiiv is where they built it. You can start yours for 30% off your first 3 months with code PLATFORM30. Start building today.
xAI Just Opened Its Image Generator to Developers.
xAI, the AI lab Elon Musk founded, has made its image generation model available via API. The model, which already powers over 300 million images generated through Grok, xAI's consumer chatbot, is now accessible for developers and businesses to build with directly. The quality mode it is releasing promises higher realism, better text rendering, and more creative control than the standard version.
It is a straightforward developer play, but the scale number is worth noting. 300 million images is a real user base, and opening that same model to the API market puts xAI in direct competition with OpenAI's gpt-image-2, which launched to strong reviews last month. Learn more here.