Amazon, Adobe, Atlassian, Citi, and others are quietly capping AI usage as costs spiral beyond projections. The gap between the AI productivity narrative and the actual P&L reality is closing fast — and it's not closing in a good direction.
Researchers found people rated AI impersonators as more authentic, coherent, and relevant than the real politicians. This isn't a future problem — it's an active threat to how the public processes political reality.
A flaw in Apple's privacy feature could allow attackers to unmask the real emails behind relay addresses. Users who relied on it for anonymity should treat that protection as potentially compromised.
Current AI, the $400M non-profit launched at the Paris AI Action Summit, has published a map of where open-source AI is weakest relative to closed frontier models. A rare attempt to treat open AI infrastructure as a public good rather than a competitive moat.
Mirendil cofounders — ex-Google and Anthropic researchers — are building AI systems designed to contribute to their own development. The self-accelerating AI thesis is moving from academic paper to funded startup.
After the US government sent Anthropic a letter pulling Fable 5 access, Isenberg makes the case for local AI — intelligence on your own hardware that survives bans, stays private, and runs free after hardware cost. Government-model friction just got real.
Kobeissi dropped a sharp data point: companies with the highest AI spending are seeing +10.2% headcount growth, inverting the narrative that AI kills jobs. The firms actually deploying AI at scale are hiring more, not less — suggesting the replacement story may be lagging the expansion story by several quarters.
Ethan Mollick floated an underrated idea: what if the frontier model itself becomes the router, delegating tasks across an agent stack rather than relying on external orchestration? As models get smarter, the meta-layer collapses into the model itself — a structural shift that would reshape how agent frameworks are built.
The Neuron flagged the overnight shutdown of Fable 5 by government mandate — Anthropic's most capable model launched, then vanished. This is the first time a frontier model has been pulled mid-deployment by state action, and it sets a precedent that every AI lab and enterprise buyer should be stress-testing their continuity plans against.