How Anthropic runs large-scale code migrations with Claude Code

Anthropic's playbook for using Claude Code to run large-scale code migrations, built from two real-world examples: migrating Bun from Zig to Rust (~1M lines in under two weeks, $165K in API cost) and migrating a Python codebase to TypeScript (165K lines over a weekend).

The central philosophy is shifting effort from fixing individual bugs to fixing the process that produces the code. A six-step pipeline runs from building a "judge" (portable test harness) through rulebook creation, stress-testing those rules with mini-migrations, fan-out translation by many agents, then compile/run/behavior-match loops with progressively less human judgment.

Key practices: adversarial review paired with mechanical verification, smaller models for high-volume implementation and largest models reserved for reviewers and rule-writers, and a mechanical work queue where "done" means the output file exists on disk. Human attention stays on patterns, not individual failures.

The Bun results in production: memory dropped from 6,745 MB to 609 MB in one benchmark, binary 19% smaller, and 2–5% faster across real-world workloads.

Sources: raw/ai-code-migration Last updated: 2026-08-01