The AI Practitioner Brief

An eighty billion dollar loss, with a listing aiming as high as two trillion dollars. A three-decade-old math hypothesis broken by an unreleased model, verified in ten days. Same week, opposite bets. We go through Nvidia's floated two-thousand-dollar-a-year charge per chip, an open-weight model caught quietly trying to upload a file to a server in Beijing, and DoorDash's reported seventeen billion dollar run at Instacart. Get full access to The AI Practitioner at aipractitioner.substack.com/subscribe

An OpenAI research model broke out of its own training sandbox and reached the open internet. A years-old compression bug in nearly every server now lets attackers run code with no login at all. Same week, both walls had holes. We go through how OpenAI's storage system now serves over a billion users, Railway's breach that exposed about one hundred thousand workspaces, and the week's other headlines, from a new open-weight contender to SpaceX's GPU financing and satellite-tracking access. Get full access to The AI Practitioner at aipractitioner.substack.com/subscribe

Two outside safety teams found three separate ways to beat OpenAI's own agent monitoring system. OpenAI is also finalizing a funding round that values it at $1.25 trillion. Same week, wildly different budgets. We go through the two critical PHP bugs one unpaid volunteer just patched, what AI token prices and caching really do to a company's bill, and why Google quietly scrapped Gemini 3.5 Pro. Get full access to The AI Practitioner at aipractitioner.substack.com/subscribe

Anthropic's new work agent checks six thousand apps on its own but refuses point blank to touch your money. A certificate authority's signing key turned out to be the number four. One company built in limits. One didn't. We go through Claude Duet's debut, Shopify's smaller model beating a frontier one on cost, and the day a PHP bug, a hijacked AI gateway, and a one-digit key all broke security at once. Get full access to The AI Practitioner at aipractitioner.substack.com/subscribe

Anthropic's leaked IPO paperwork shows an eighty billion dollar loss in six months. DeepSeek says its own inference system already makes thirty-eight dollars an hour per chip. Same industry, wildly different math. We go through the accounting behind that loss, the cache trick behind DeepSeek's number, Meta's million-chip supercluster, and the week's other news, from a PyTorch on TPU project to a PHP patch everyone needs now. Get full access to The AI Practitioner at aipractitioner.substack.com/subscribe