Dark Factory
A fully automated pipeline with no humans inside it.
The end point of the software-factory idea: work that flows from ticket to deployment with no human in the loop at all. Named after factories that run without lights because nobody is inside.
Achievable today only for narrow, verifiable, low-blast-radius tasks. As a general claim about software engineering it is mostly aspiration — verification, not generation, is the bottleneck.
Reachable today only within narrow boundaries: high test coverage, reversible changes, a mature rollback path, and a class of work whose failures are cheap. Everywhere else the honest architecture is the same pipeline with a human gate, and the gate belongs where it is cheapest to be wrong.
Where the human gate still belongs
Fully automated dependency updates are working well.
The success invites extending the same autonomy to schema migrations, where a wrong change is not reversible in the same way. The boundary is not the difficulty of the work; it is the cost of the failure and whether rollback restores the previous state.
- blogEffective harnesses for long-running agents ↗
The initializer-plus-incremental-sessions pattern for work that outlives a context window, and the four ways it fails: doing too much, losing context between sessions, declaring victory early, not testing end to end.
- reportState of AI-assisted software development 2025 ↗
Frames AI as an amplifier of the organisation it lands in: the same tooling helps strong delivery cultures and compounds problems in weak ones.
Software Factory
A pipeline where agents take tickets and produce code, with checks along the way.
AFK · Away from keyboard
Running without a person watching.
Automated Check
A machine verdict the agent cannot talk its way past.
Human Review
A person reading the work, with judgment as part of the job.