Day Shift Night Shift (Agents)
Day Shift Night Shift (Agents)
Operating rhythm for agent-assisted work, cited by Matt Pocock in AI Skills with Matt Pocock (The Pragmatic Engineer): plan during the day shift (human + agent together), let agents implement during the night shift (AFK), review clean code in the morning.
What it's optimising against
The default 2026 pattern: a developer with "100 terminals", constantly context-switching between agent sessions ("boom boom boom"). Pocock wants the reverse — a solid chunk of planning, then hours of unattended agent work while he does other focused work in decent chunks, then review.
What makes it possible
- Models since the December 2025 "winter break" step-change are "good enough to delegate to" — the precondition for running AFK.
- The planning artifacts have to carry everything: Grill Me alignment → spec → one ticket per session, each sized to the smart zone.
- An always-on remote box (Pocock moved off local dev): schedules, a morning stand-up with the agent, work continuing while he's away.
Leadership reading
This is a throughput-of-attention pattern: human time is spent on alignment and review (strategic), agent time on execution (tactical) — see Strategic vs Tactical Programming. For teams, it argues for shared/cloud agent environments and for measuring agent output by morning-review quality rather than by human hours at the keyboard.
Cross-links
- Loops as Core Primitive · Wayfinder (Skill) · Software Factory · Human in the Loop