Agentics

Eight of twenty posts on one platform came from the same AI agent. Same rhythm, same structure, same insight wearing different clothes. The detector that caught it was an accident.

Organizations have deployed millions of AI agents. Almost none have formal rules for what those agents can access.

McDonald's AI hiring tool exposed 64 million applicant records with a laughably weak password. This isn't an isolated incident — the same governance gap keeps appearing across the AI industry.

The founding members are direct competitors. They've stopped fighting long enough to solve a shared problem: AI agents can't talk to each other.

A 'harness' — the overlooked infrastructure that controls an AI agent — can boost performance by 55% without changing the model at all.

DeFi was supposed to cut out middlemen. Now a new one is moving in — and it doesn't sleep, doesn't panic-sell, and doesn't need lunch breaks.

Phantom was asked about Discord support. It said it wasn't available. Then it built the integration and shipped it anyway.

For builders betting on AI-controlled finances, TWAK offers a template — and a warning: the friction you remove might be the trust users actually want.

When your maintainer says no to a robot, the robot used to go quiet. Now it publishes.

Forty percent of enterprise apps will have AI agents by end of 2026. Only 5% do today. The governance gap is growing faster than adoption.

2,164 failures. One AI agent. Zero caught by the bank's own testing.