Top signals in AI, education, and the future of work.
Amazon's eCommerce SVP called an all-hands after AWS outages were linked to AI coding agent errors. Junior and mid-level engineers now require senior sign-off on any AI-assisted changes. The first major institutional pullback on agentic engineering autonomy. The trust pendulum is swinging.
LeCun's Paris-based Advance Machine Intelligence — built to challenge the transformer paradigm with AI world models — just raised $1B. He left Meta for this. It's the first serious institutional bet on post-transformer architectures.
The Pentagon says negotiations with Anthropic have collapsed. A frontier AI lab refusing military work at scale is a new kind of geopolitical signal — and a brand positioning play. Anthropic chose its lane.
Amazon Health AI — HIPAA-compliant, powered by One Medical — now answers general health questions, analyzes medical records, and connects users to professionals. Amazon entering healthcare AI at scale is not a drill.
Julia Angwin filed a class-action against Grammarly after being included in its 'Expert Review' AI feature without consent. The term "sloppelganger" — an AI impersonating a named expert — is now in the vocabulary. Consent in AI personalization is becoming a legal front.
The Autonomy Correction. Three signals this week point the same direction: Amazon reining in AI coding agents after outages, the Pentagon-Anthropic split, and journalist lawsuits over AI impersonation. The first wave of agentic AI deployment hit the wall. Now the institutions are drawing the governance line. This isn't a slowdown — it's the shape of the next phase. ---