• Hitachi opened a literal software factory in 1969: a building where code was produced under statistical quality control, defect rates measured per thousand lines, managers accountable for output. Toshiba, NEC and Fujitsu copied it.
• The dividing line is liability. Ford sells you a car and recalls it when it fails; it doesn't sell you a wrench and wish you luck. A factory stands behind its output, a developer tool never has to.
• The input test: if what goes in is a Jira ticket written by an engineer for an engineer, you have a power tool bolted to an old process. A factory takes business requirements and works out production itself.
• The coherence test: intent, spec, code, tests and production behavior must move as one governed object. Change the requirement and code changes; hotfix the code and the requirement updates. Ask a vendor to demo that loop live.
• Third test is person-independence. Hand the same coding agent to two engineers and output quality diverges with who writes the prompts and who catches the mistakes. A production system can't depend on that.

• AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.

If you're building or evaluating AI-assisted development pipelines, this piece gives you the historical and conceptual vocabulary to distinguish real industrial-grade systems from the current wave of hype — and explains why 'guaranteed output' is the standard enterprises are actually trying to reach. The COBOL reverse-engineering case study makes the abstraction concrete.

• postSoftware Factory BuilderPierson Marks
• clipThe software factory loop: signal to deployAI Engineer
• videoFactory's Forward Deployed Engineering Model for DroidAI Engineer

Checking sign-in…

Loading comments…