tsrs: Cloudflare and VoidZero port TypeScript to Rust, claiming up to 8x speed and 70% less memory
TL;DR
tsrs is an experimental Rust port of the TypeScript compiler from Cloudflare and VoidZero, built with the TypeScript team rather than as a fork. Cloudflare reports up to 8x faster and about 70% less memory than TypeScript 7.0. The benchmarks are Cloudflare's own.
On this page
tsrs is a Rust port of the TypeScript compiler, announced by Cloudflare and VoidZero. TypeScript is the typed layer on top of JavaScript, and its compiler (tsc) checks your code for type errors. Per Cloudflare's post, tsrs aims to be a faithful port that follows upstream behaviour, and it is built together with the TypeScript team. The speed and memory figures below are Cloudflare's own and have not been independently reproduced.
- Project
- tsrs, a Rust port of the TypeScript compiler
- Built by
- Cloudflare and VoidZero, with Max Schwenk of Owner.com, in collaboration with the TypeScript team
- Home
- Incubated in the Oxc JavaScript toolchain
- Status
- Experimental. Install with
npm install tsrs; a WebAssembly build ships as@tsrs/wasm - Licence
- Not stated in the announcement
- As of
- 20 Oct 2026
What it does
It type-checks TypeScript the way the official compiler does, written in Rust. Cloudflare says it reports up to 8x faster runs and about 70% less memory than TypeScript 7.0 (the Go-based rewrite of the compiler), with the same number of type-check workers. The post also says TypeScript 7.1 beta is up to 2.5x faster than 7.0, and that some of those gains came from work done in tsrs.
- faster than TypeScript 7.0
- up to 8x
- less memory than TypeScript 7.0
- about 70%
- TypeScript 7.1 beta vs 7.0, partly from tsrs work
- up to 2.5x
Read the benchmarks with care
Cloudflare says it agreed the configuration with the TypeScript team so both sides run the same setup, and that the benchmarks live in the project's bench directory. It also criticises other ports for benchmarking on larger machines with non-default options. The post says tsrs by default picks a configuration that adapts to machine size, which uses more memory and runs faster. The figures are charts in the post, with no exact numbers in the text.
On the Hacker News thread, some commenters questioned the 8x figure and what it is measured against, and one noted that TypeScript 7.0 already checks large projects quickly. Treat "up to 8x" as a best case on Cloudflare's chosen workloads.
Why Rust, not Go
Commenters on the thread say the aim was a compiler that runs inside Cloudflare's WebAssembly-only sandboxes. Per the post, compiling the Go compiler to WebAssembly means shipping Go's garbage collector with it, while Rust's memory model makes it simpler for the compiled module and its host to share data. Speed was a side benefit in that reading, and at Cloudflare's scale the compute savings matter too.
What is missing
- Done: Type checking from the command line (
npx tsrs) - Done: WebAssembly build (
@tsrs/wasm) - Not done: Same diagnostic output order as
tsc(known differences) - Not done: Language Server Protocol support, which editors use for hints and errors
- Not done: Content mapper support
- Not done: Stable release
Cloudflare plans to keep tsrs in sync with upstream TypeScript using an agent-driven process like the one it used for Vinext, to contribute performance fixes back to the TypeScript Go codebase, and to merge the separate tsc-rs project led by Theo Browne into the same package.
Try it
- Open a projectAny folder with a
tsconfig.json. - Run it
npx tsrs, which fetches the experimental package from npm. - CompareRun
tscon the same project and diff the diagnostics. Output order may differ.
Best for / skip if
Best for
- teams with large TypeScript codebases or tooling authors who need a compiler that runs in WebAssembly, and anyone willing to test an experimental build against
tsc.
Skip if
- you need editor support, a stable release or identical diagnostics today, or want the speedup confirmed by someone other than the vendor first.
Sources
- Cloudflare blog: TypeScript in Rust: what tsrs is, who built it, the benchmark claims, install commands and known gaps
- Hacker News discussion: reader reactions, benchmark scepticism and the WebAssembly rationale
Continue reading
DeepInfer: DeepSeek open-sources its inference engine, but it runs one model, offline, for now
DeepInfer is DeepSeek's open-source inference engine (Apache-2.0, built mainly on vLLM). The release runs only DeepSeek-V4.1-Flash, offline, on NVIDIA SM100 GPUs or Huawei Ascend 950 NPUs. It has no KV cache system and no online serving. The $38 per GPU-hour figure is unverified.
Vercel eve deep dive: real deployments, what it costs on Vercel versus Google Cloud, and the alternatives
eve is Vercel's Apache-2.0 agent framework, in public preview since 17 June 2026. Vercel runs 100+ agents on it, and an outside company reports about 20 on one 4-core server. In an illustrative model, Vercel hosting costs less than a Google Cloud VM until roughly 13,000 to 26,000 runs a month.
pstack: a Cursor engineer's open-source skill pack that makes agents prove their work
pstack is a free, MIT-licensed plugin of agent skills from Cursor engineer Lauren Tan. Its /poteto-mode skill picks one of 22 playbooks for a task, applies engineering principles such as proving the work on the real app, and can have several models review a diff. It installs in Cursor and Claude Code.