Blog

Alphabet's worst day on record came with 25% revenue growth and $568bn of backlog. The first projects cut will be inside companies that never sold AI.

April's model risk rewrite put AI out of scope and deleted the sections that named an owner. When an AI system fails, blame goes to whoever is closest.

88% of organizations had an AI agent security incident last year. The guardrails they bought wrap exactly one agent, and the tier above it is missing.

OpenAI launched a $4 billion company to deploy the models it already built. The model was never the bottleneck. The scarce resource is a person who owns the outcome.

Europe put record money behind sovereign AI and, the same week, its central bank warned it could be cut off. Trust is proven, not promised. Inspectable wins.

China's agentic AI pays at scale. The hard part is authority: who lets it spend, and who answers when it's wrong. One market solved it via escrow rails.

Agent swarms and sub-agents fail when they carry everything: every responsibility on the split, every bit of context loaded at once. The fix is structure.

Creator-vetting tools track what's moving, which category is hot, and how many followers a creator has. They leave out which creator converts sales for a shop.

Europe's AI Act leans on human oversight that cannot scale to its adoption goals. The missing mechanism is how trust gets built in, industry by industry.

"Good enough" AI models made the whole industry affordable in 2026. That victory didn't make it manageable, because the model was never the expensive part.

China is building new power and water for AI. The US is answering with software and a new engineer. The advantage isn't the model. It's what surrounds it.

Capability and authority are separate design decisions. A framework for what an agent is authorized to decide, drawn from Meta, Claude Mythos, and Workday.

The Seven Dimensions of Enterprise AI names the seven decisions behind an AI product that works in production. Containment is where an agent's authority ends.