Skip to content

Blog

Strong automated gates separate a dark factory from a bullshit factory

Above a certain complexity, human code review lowers quality at agent speed. Quality inversion, the automation ratchet, and what a software dark factory needs instead of approvals.

Written by

Software Factory

Published: 31 March 2026 Updated: 1 min read

Most teams building with AI agents are accidentally building bullshit factories. High output, weak quality gates, garbage at scale.

The dark factory

Manufacturing solved this decades ago. They call it a “dark factory”: lights-out production, no humans on the floor, a factory that runs itself.

Now we are building dark factories for software. Agents write, test, deploy and maintain code, with no human intervention on individual steps. The human builds the factory, not the software.

Quality inversion is real

Here is the uncomfortable part. Above a certain complexity threshold, human code reviews decrease software quality. The system holds more codified knowledge than any reviewer can activate in review time. You are rubber-stamping, and rubber-stamping behind a human approval gate is worse than automated fitness functions.

What makes a bullshit factory

  • High-volume agent output with weak quality gates. Agents optimize to pass gates, so weak gates mean garbage at scale.
  • Human approvals that give false confidence. “A senior reviewed it” means little at agent speed.
  • Measuring lines of code generated, not defects prevented.
  • No fitness functions, just vibes.

What makes a dark factory

  • Every human correction becomes a test or a fitness function. The ratchet never goes backward.
  • Programmatic tests and fitness functions are the source of truth, not the specification.
  • Quality gates are automated, measurable and continuously tightened.
  • The fix is the system upgrade.

The automation ratchet

Every time a human intervenes, that intervention gets codified as a rule, a test or a fitness function. It is irreversible, and there is no known upper limit.

This is what separates agentic engineering from vibe coding. Vibe coding ships fast with weak gates. Agentic engineering builds the factory that ships fast with strong gates.

Are you building a dark factory, or a bullshit factory?

Further reading

  1. Build an auditor that improves your automations' prompts An automation that rewrites other automations' instructions sounds risky. Built as a loop of real runs, explained verdicts and a reviewed diff, it is continuous improvement with one gate. 2 min readShort post
  2. AI modernization works as a delivery problem, not a code translation Translating COBOL to Java one to one reproduces the old debt. Treated as delivery, with architecture decisions, encoded context and a migration factory, AI-driven modernization works. 2 min readShort post
  3. Every company will build its own software factory The question shifts from which editor a team uses to what its factory looks like. Four stages lead from autocomplete to a factory that improves itself, and no team can buy its factory off the shelf. 2 min readShort post

All 23 posts on Software Factory

ClientsClient work

Test the constraint in your system

The Factory Readiness Assessment connects this research to your value stream, your data, and the next decision.

Training on this topic: Agentic Software Factory: Advanced Team Training
See the assessment

Want to check one team first? The self-check: twelve yes or no questions, three minutes, instant result, no account. Take the self-check