Browse by topic.
The same body of work, sorted by subject. Each hub gathers the published field notes and essays on one thread of running real companies with AI agents doing the work. Start where the problem you have lives.
Claude Code, in production
Field notes on running Claude Code as daily infrastructure: durable memory, session logging, model routing, context budgeting, and the failures that shaped the setup.
Latest · Automatic Model Routing Doesn't Exist Yet (Mar 21, 2026)Open the hub →Running AI agents
How a team of AI agents actually runs a company: reliability patterns, autonomy limits, task queues, review gates, and what breaks when agents work unattended.
Latest · The Gate That Can Say No (Jul 6, 2026)Open the hub →AI infrastructure & plumbing
The plumbing under an AI-run workspace: backup and restore drills, local-first cron, deploy pipelines, convergent logging, and the daemons that keep it alive.
Latest · What It Takes to Run a Company on Agents (Jul 6, 2026)Open the hub →Running companies with AI
The operating record of running real companies with AI agents doing the work: content teams, goal systems, review gates, product briefs, and where human judgment stays.
Latest · The Gate That Can Say No (Jul 6, 2026)Open the hub →macOS developer traps
Platform-specific failures from building on macOS: launchd silent failures, the system Python binary, iCloud path shifts, app-launch timing, and orphaned processes.
Latest · I Tunneled a Dev Server and Exposed the Whole Folder (Apr 10, 2026)Open the hub →AI workspaces & repos
Managing the workspaces an AI team works inside: multi-repo VS Code setups, master-source versus living copies, branch discipline in knowledge repos, and distribution audits.
Latest · Master Source vs. Living Copies (Mar 30, 2026)Open the hub →