The Fed builds an AI-jobs panel + Meta's $10 billion compute bet

Under new Chair Kevin Warsh, the Federal Reserve created five external task forces to rethink how it studies the economy — and handed a co-lead seat on its Productivity and Jobs panel to Andreessen Horowitz co-founder Marc Andreessen, alongside Stanford economist Charles Jones and Microsoft executive Asha Sharma. The group will assess how AI is reshaping employment, productivity, and price growth, with recommendations due by the end of 2026. It is the first time the central bank has built a formal structure to study AI's macroeconomic impact — and the first time it has given a prominent venture capitalist a seat at that table.

Meta will build a one-gigawatt data center north of Edmonton, Alberta — its first in Canada and 33rd worldwide — in a project the company values at about $10 billion. The site, expected online within three years, anchors a striking strategy shift: Meta plans to rent surplus graphics chips to outside customers between its own training runs, treating idle compute like empty airline seats. It is a notable pivot for a company that has historically hoarded its own capacity, and a sign that the largest AI builders now see spare compute as a product.

Separately, Meta had a rougher week on the consumer side. It removed a feature from its new Muse Image generator that let users alter photos by tagging any public Instagram account, pulling in that account's images without notifying the owner. Critics — including the talent agency CAA — warned the tool invited non-consensual edits, echoing AI's history of fake intimate imagery. Meta said the feature "missed the mark" and is no longer available.

French lab Mistral entered embodied AI with Robostral Navigate, an 8-billion-parameter model that steers robots through unfamiliar buildings using only a single standard camera — no LiDAR, no depth sensors, no pre-built map. Given a plain-language instruction like "leave the lobby and stop at the second shelf," it plots a path, and Mistral says the same model works across wheeled, legged, and flying robots. Trained entirely in simulation, it is the company's opening move toward a unified "embodied agent" for warehouses, delivery, and logistics.

OpenAI signaled a push toward household users, posting for a product manager to build ChatGPT experiences for families, caregivers, and older adults. The move follows recent parental controls and sensitive-conversation routing, and tracks a real demographic shift: Sensor Tower data shows ChatGPT users aged 35 and over climbed from 26 percent to 31 percent in a year, with parents now making up roughly a quarter of US adult smartphone users. It is an early, hiring-stage signal rather than a shipped product.