About
Why Dark Factory exists, and what it is trying to find out.
Dark Factory is an experiment, not a finished product — an attempt to find out how far AI agents can be trusted to build real software under a governed process, with a human retaining final authority throughout.
It is not intended to eliminate people from software development. The motivation is the opposite: to find a meaningful future for development teams, in which developers move from much of the repetitive implementation work toward higher-level engineering — architecture, design, problem-solving, and ownership of risk and outcome. If that works, the value of a developer’s time goes up, not down: the same person who once spent it producing boilerplate spends it deciding what to build and whether it is safe to ship.
This site is a public record of the experiment as it happens. It is written with a discipline about what it claims: a human-confirmed fact, a recorded Factory stage result, an observed issue, and a plan or hypothesis are each worded so that a reader can tell which kind of statement they are reading. That discipline is part of the point — a project about trust in AI-driven delivery should not itself make claims its evidence does not support.
The questions the project is trying to find out, and why they stay open, are the substance of the experiment. Some of them:
- How much of software delivery can move to autonomous, agent-run execution before the human role stops being a bottleneck and starts being a different kind of work — and how much has to stay human, for good?
- Does a process built on separated roles, evidence, and gates actually hold up once the content is substantive rather than structural, or does it only look rigorous on a small slice?
- What does a meaningful future for development teams look like in practice — which tasks move to agents first, and what do developers do with the time they get back?
The Journal is where the answers, as far as they exist yet, are recorded with dates.