What a risk desk taught me about evaluating AI agents

Before I wrote a line of agent code, I spent a decade deciding which trades to let through and which to halt. The instinct that transfers most cleanly isn't technical at all. It's the habit of asking what an automated system does on its worst day, not its average one.

An AI agent in production is a risk surface. The failure modes that matter aren't the ones in the demo; they're the long-tail inputs nobody scripted. That's the lens I bring to building them, and it's the same discipline that kept a balance sheet intact through more than one market shock.

Seed post. This is a placeholder draft to validate the writing pipeline; the full essay will follow.