It’s Monday
Here’s what’s worth your time today.
Elon Musk says Tesla and SpaceX may start manufacturing their own chips through a proposed “Terafab” facility near Austin. The reason is simple: existing suppliers aren’t producing enough compute to meet his AI and robotics ambitions.
Musk claims the facility could eventually support 100–200 gigawatts of compute annually on Earth, plus even more in space. It’s an enormous vision, though details and timelines remain unclear. The broader signal is familiar: as AI demand explodes, even the biggest companies are starting to vertically integrate and take control of their own chip supply.
Supported by Lightfield
After every call, you’re supposed to log notes, update six fields, create follow-up tasks, and draft an email. You have another call in 10 minutes. So you skip it. Context disappears. Follow-ups slip. Deals go cold.
Lightfield captures every interaction automatically — calls, emails, meetings. It suggests CRM updates you accept with one click. It drafts follow-ups using your customer’s actual words.
Migrate from any CRM in under an hour. just connect your inbox, upload your data, and watch your CRM build itself.
Use code CTRLALTDEL33 for 3 months off
Cursor admitted its new Composer 2 model was built on top of Moonshot AI’s Kimi 2.5, after users spotted references to the model in its code. The company says it added significant additional training, with most of the compute coming from its own work, and that performance differs meaningfully from the base model.
The bigger takeaway is how AI models are increasingly layered rather than built from scratch. But it also highlights a more sensitive dynamic: even top U.S. startups are relying on Chinese open-source models under the hood, raising questions about transparency and how the AI “stack” is actually being built.
Anthropic claims the Pentagon told the company it was “very close” to an agreement on AI use just days before labeling it a national security risk. In new court filings, Anthropic argues the government’s case relies on claims that were never raised during negotiations, including concerns about control over military operations.
The filings also push back on technical arguments, saying Anthropic has no ability to remotely access or alter its models once deployed in secure government systems. The broader signal is escalating tension between AI labs and governments, where safety positions, military use, and infrastructure control are now colliding in public and legal battles.
Mark Zuckerberg is developing a personal AI agent to help him operate as CEO of Meta, using it to quickly retrieve information and cut through layers of management. The tool is part of a broader push to make the company “AI-native,” with employees increasingly using agents to automate tasks, query internal data, and even communicate with each other.
The bigger shift is structural. Meta is flattening teams, embedding AI into performance reviews, and encouraging employees to build their own tools to speed up work. The signal is clear: AI isn’t just a feature inside companies anymore, it’s starting to reshape how companies themselves are run.
OpenAI plans to nearly double its workforce to 8,000 employees by the end of 2026, expanding across engineering, research, product, and sales. The company is also adding “technical ambassadors” to help enterprises adopt its tools more effectively.
The hiring push comes as competition intensifies, particularly from Anthropic, which is gaining traction with new enterprise customers. The signal is clear: winning in AI isn’t just about models anymore, it’s about distribution, support, and embedding into real-world workflows.