~/reejangopan.me — Reejan R Gopan, home
Download CV
Enterprise engineering

How do multiple developers or AI agents work on one codebase safely?

By Reejan R Gopan · Full-Stack Software Engineer, Dubai, UAE

One accountable technical lead (the orchestrator) owns requirements, scope, acceptance criteria, architecture decisions, integration order and production changes. Specialists (security, database, performance, frontend/QA, SRE) can audit and build independently, but:

  • One branch or worktree per task with assigned, non-overlapping file ownership
  • Order of work: contracts/ADR → database migrations → API → frontend → integration tests
  • One migration authority who serialises schema changes and replays them on a fresh database
  • Only the integration owner merges or cherry-picks — a cherry-pick copies a commit, not its dependencies or migrations
  • Protected main, CODEOWNERS, required CI, reviews, environment approvals — no silent emergency merges, no force-push to main
  • Treat AI-written code as untrusted until reviewed; limit agent credentials and separate dev from production secrets
  • Write a spec first: user story, scope exclusions, threats, invariants, permissions, API/DB contract, error behaviour, migration compatibility, tests and rollback plan

All guides