I rewrote 65,000 lines of Go to Rust with Fable. Invoice: $400

An engineer tasked Fable, Anthropic's specialized code transformation model, with migrating a 65,000-line Go codebase to Rust. The result? It works, it cost $400, and it raises real questions about the future of large-scale codebase migrations.

Case study: Migrating 65,000 lines without breaking the team

Iurii, a backend developer, had a Go project that worked—but he wanted to switch to Rust. Reason: performance, memory safety, and an ecosystem that had attracted him for some time. The classic problem: 65,000 lines of code (65kLoC) would take two to three years of part-time manual migration, or a dedicated team that most startups don’t have.

His solution: entrust the project to Fable, Anthropic’s model specialized in large-scale code transformation. Result: the migration happened. Total token cost: $400.

What Fable does (and doesn’t do)

Fable isn’t a simple “Go→Rust translator.” It understands program semantics—Go idioms (goroutines, channels, implicit interfaces) and their Rust equivalents (async/await, traits, ownership). It produces idiomatic Rust, not a literal transliteration.

What remains manual:

• Test validation (Fable generates, tests pass or fail—it’s up to you to debug)
• Architecture decisions (keep a Go channel = tokio::sync::mpsc or restructure?)
• Third-party integrations (C bindings, drivers, specific system calls)

The real cost: $400, but how many hours?

Iurii doesn’t give the exact human time, but mentions “a few weeks” of supervision. The migration isn’t “zero shot”: there are back-and-forths, corrections, and blocks where Fable gets stuck. The token cost is low; the cost in human attention remains non-zero.

For comparison, a manual migration of 65kLoC Go→Rust by a senior developer experienced in both languages would be estimated at 3–6 months of full-time work.

What this means for teams

Three key takeaways:

1. The viability threshold collapses. A migration no one dared to start due to resource constraints becomes feasible for a 2–3 person team.
2. The bottleneck shifts. It’s no longer about “writing the code” but “validating the produced code.” Testing, code review, and PR reviews become the real work.
3. Accumulated technical debt becomes attackable again. Bases like Python→TypeScript, PHP→Go, Ruby→Rust… projects frozen for years re-enter the realm of possibility.

Key takeaways

Fable at $400 for 65kLoC Go→Rust is a tipping point: large codebase migrations are no longer reserved for teams with substantial R&D budgets. The developer’s real job becomes supervision, validation, and architecture—not transliteration. A shift in posture as much as a productivity gain.