How We Migrated 2M Lines of TypeScript to Strict Mode in 6 Months

Stripe's TypeScript monorepo had grown to over 2 million lines of code across hundreds of packages, but strict mode had never been enforced globally. The result was a patchwork of type safety — some packages were strict, others relied heavily on `any` types and implicit conversions that masked real bugs.

The migration team (3 engineers, part-time) started by building a dependency graph of all internal packages and identifying leaf nodes — packages with no internal dependents. These were migrated first using automated codemods that handled the most common patterns: adding explicit return types, replacing `any` with `unknown`, and fixing nullable access patterns.

The codemod pipeline processed approximately 400 files per week. Each batch went through automated testing, then a lightweight human review focused on semantic correctness rather than syntax. The team found that 92% of transformations were correct on the first pass.

The critical infrastructure decision was the CI ratchet: once a directory was marked as strict-compatible, any PR that introduced a strict-mode violation would fail CI. This prevented regression and created natural pressure for teams to migrate their own code proactively.

By month 4, momentum had shifted from push to pull — teams were requesting migration support rather than resisting it, largely because the IDE experience in strict packages was noticeably better. Auto-imports worked more reliably, refactoring tools caught more issues, and code review became faster.

The final metrics were compelling: 38% fewer type-related production incidents, 15% improvement in IDE autocompletion accuracy (measured via internal telemetry), and a measurable reduction in PR review cycles for strict packages. The team estimates the migration paid for itself within 3 months of completion.