Operator Notes

Field notes on running product and building software with agents doing more of the work.

2026

I ran a bake-off for my meeting notes. Price didn't help.

Public benchmarks measure general skill. A test on five of my own meetings, graded against facts I could check, told me which model to trust with my agent's memory. Price had little to do with it.

Tokens got cheap. Tasks didn't.

Per-token prices fell 50× in three years. The bill for a finished agent task went the other way, and the spread is the number nobody benchmarks.

The three guardrails I run my agentic OS by — and why the limits made it more useful

Each rule trades a capability I rarely need for supervision I'd otherwise owe constantly.

The job description for an AI-fluent PM in 2026

The falsifiable proxy: has this person built and lived with an agent of their own, unattended, for a week.

Six months building an agentic OS. What it changed about how I think about product orgs.

The rubric replaced the PRD, the cost meter is about to go live, and most org charts haven't priced in either.