GitHub Uses Copilot to Rewrite 800,000 Lines of TypeScript into Rust

A GitHub engineering case study became one of the week's most concrete demonstrations of AI-assisted software migration at scale: Copilot was used to convert 430,000 lines of TypeScript into roughly 800,000 lines of Rust, at a reported cost of about $120,000. The write-up is candid about the division of labor — examining where the AI genuinely accelerated the port and where human engineers still had to intervene to fix, verify, and architect.

The example matters because language migration is a notoriously expensive, error-prone undertaking that has historically required large teams and long timelines. A ~$120,000 AI-assisted port suggests a new cost curve for modernization work, particularly for organizations sitting on large legacy codebases. The near-doubling of line count (430K TS to 800K Rust) also reflects the languages' different verbosity and Rust's explicit memory and type semantics.

The honest framing — 'here's what AI actually did' — is what gives the story credibility with a skeptical developer audience that has grown wary of inflated AI-coding claims. It fits the week's broader coding-agent theme: xAI's Grok 4.7 targeting CursorBench, AWS's open-weight coding-agent guidance, and OpenAI shipping Sol/Luna into Codex. The caveats are real: an 800K-line Rust codebase still needs thorough testing, security review, and long-term maintenance, and the write-up reportedly notes areas where human oversight was indispensable — architecture decisions, tricky type translations, and correctness verification. For engineering leaders, the takeaway is that AI can compress the grunt work of migration dramatically while shifting, not eliminating, the human burden toward review and judgment. Whether the resulting Rust is idiomatic and maintainable — not just compiling — is the question that determines if this is a template or a cautionary tale.