LLM-built Rust port of the TypeScript compiler arrives with $400,000 in API-priced tokens and a nod from the TypeScript team
- The pingdotgg/ts-rust repository ports the TypeScript compiler, type checker, and language server to Rust, and the author reports spending over $400,000 in API-priced tokens with GPT-5.6 Sol and GPT 6 Astra, though the GitHub description puts the real bill at about $24,000.
- A TypeScript team member commented that this is impressive work and that three such ports have appeared in the last week, promising the team will have more to say soon.
- The repo carries 5,963 commits across 64 branches and publishes to npm as tsc-rs under MIT AND Apache-2.0, with built-in Effect diagnostics and a release rule where a stable version tag publishes to the latest dist-tag.
- Benchmarks in the real-world project section show compile times 1 to 3 seconds faster than tsc7, a gap one commenter noted looks far smaller in absolute terms than the percentages suggest.
- The readme uses a "slop line" marking everything below it as written by LLMs rather than the author, and commenters checking the source found no unsafe code, with avoiding unsafe listed as an explicit goal.
Hacker News opinions
Wait, you spent north of $400,000 on this?
That's $420,000 worth of tokens at API prices, but I'm pretty sure he ran a few 20x subs. The number is what it would have cost, not what he paid.
Most of it went to OpenAI tokens, and the GitHub description says the actual cost was $24k. People love flexing with API prices for tokens they got through subscriptions.
TypeScript team member here. This is impressive work, and it's honestly crazy that three of these ports popped up in the last week. I'm AFK right now but we'll have more to say soon.
Go looked like a very reasonable target when they ported originally, but the world is changing fast. Have an agent mine this port for ideas worth bringing back into the real compiler.
How much unsafe does it use? These ports either write idiomatic Rust or slap unsafe everywhere, and the readme says nothing about it.
Checked the repo search, no unsafe on a skim, and avoiding unsafe looks like an explicit goal. There's a goals file that says so.
Is this a serious project? No. Will it be maintained? Nope. It's like porting Doom to a microwave, but it shows what's possible. We'll look back on this fondly.
More unserious and unmaintained projects at $400k a pop. Sick.
LLM reimplementations checked against years of human tests will have holes where the original authors thought tests weren't necessary and common sense was enough. I say that partly to protect my ego, since I'm never reading all that code.
They'll have holes for about three days and then it gets fixed. For software with a thick test suite, humans create those gaps more often than the LLMs do.
This should be titled "Rust port of the Go rewrite of the TypeScript compiler."
In the real-world project section, compile times beat tsc7 by 1 to 3 seconds. Percentages look way more impressive than the absolute numbers.
Bun claimed an 11-day rewrite too, then spent three months of human labor fixing hundreds of issues before shipping it as a release. Good luck adding new code to a codebase nobody has read.
That comment kind of answers itself. In the future you don't add or read code, you add test cases and let the loops cover the new case.
What a waste of power. $424,000 of API-priced tokens is how many kWh burned? And how would you trust this to build your code?
Same way you trust any build of the official compiler: it passes all of the existing tests.