- Agent safety and governance: Essays on keeping AI agents inside their authority: models that propose, tools that check scope and live state, harmless retries and grading on the end state.
- Agentic dev pipelines: How to build a delivery pipeline for coding agents you can trust: spec first, hooks for hard rules, held-out acceptance, gates that read what ran, and merge risk math.
- AI adoption and team enablement: How to take AI from demo to a job a team relies on: one problem and one number, the smallest proving scope, risk as a measured range, staged gates and a stop button.
- Build or adopt for AI systems: How to decide whether to build or adopt AI agents, gateways and infrastructure: measure what you already run, build only on a named trigger, keep the build thin.
- Evals and measurement: How to measure whether an AI system is right: answer keys built first, lower bounds instead of headlines, severe-error gates, slices, consistency and calibrated judges.
- Running AI in production: Essays on running AI after launch: blind audits, earned confidence, hand-off rules, measured reviewers, release gates and an honest range on every rate.
Topics.
Topic hubs that group the field essays: Agent safety and governance, Agentic dev pipelines, AI adoption and team enablement, Build or adopt for AI systems, Evals and measurement, Running AI in production.