While Anthropic and OpenAI trade blows over IPO filings, custom chips and defecting Gemini researchers, the more interesting AI story is happening a level down, where a privacy-first chatbot just hit unicorn status, a Montreal startup is quietly fixing the timing chips inside data centers, and a UK law firm proved a bot still cannot be trusted to draft a will. Even today's applied wins, an AI-designed coronavirus vaccine clearing human trials and a 2,000-year-old Vesuvius scroll finally readable, are reminders that AI's most consequential work rarely comes with a press release. The giants get the headlines, but the smaller builders are the ones actually shipping.
A 15-person startup founded by a former Amazon and Pinecone researcher raised seed funding to build a search engine that returns machine-ready passages in under 200 milliseconds instead of ranked links for humans to skim.
.jpg)
Acquired in an all-stock deal worth 3.9 billion dollars, as Qualcomm builds a silicon-agnostic AI software layer to challenge Nvidia's CUDA.
A tiny macOS menu bar app that keeps your lid-closed Mac awake only while AI coding agents are actively working, then lets it sleep normally.
Ben Thompson argues Anthropic's safety messaging conveniently licenses business moves, like data retention and API restrictions, that serve its bottom line as much as its mission.
