## Overview AI daemons are a distinct product category: always-on background AI processes that autonomously handle engineering maintenance tasks (PR hygiene, bug triage, documentation updates) without human prompting. Named after Unix daemons — persistent background processes that serve requests without direct user interaction. ## Core Framework **Three AI product categories by interaction model:** | Category | Trigger | Scope | Example | |----------|---------|-------|---------| | Copilot | User keystroke | Line/block completion | GitHub Copilot | | Agent | User prompt | Task execution | Claude Code, Codex | | Daemon | None (continuous) | Maintenance/hygiene | AI code reviewer, doc updater | The key differentiator: daemons run without human prompts. They watch for drift (stale docs, failing checks, style violations) and fix it autonomously, the way Unix daemons handle system services. ## Cross-Domain Applications **Software engineering**: Continuous codebase maintenance — lint fixes, dependency updates, test gap detection, dead code removal. The daemon watches CI signals and repo state, not user commands. **Knowledge management**: A vault daemon could detect orphaned notes, broken wikilinks, stale project references, and fix them on schedule — the autonomous PARA agent concept is essentially a knowledge daemon. **DevOps/SRE**: Toil reduction agents that watch monitoring dashboards and auto-remediate known failure patterns. Distinct from alerting (which notifies humans) — daemons act. ## References - [[AI Daemons - A new category of AI for engineering teams.]] — Riley Tomasek tweet, saved 2026-03-25 - Unix daemon pattern: background process, no controlling terminal, serves requests autonomously