Rust Patch Set Headstart Claims 54% Faster cargo check, Has No Licence and Is Seven Days Old
Software / news
Rust Patch Set Headstart Claims 54% Faster cargo check, Has No Licence and Is Seven Days Old
The numbers come from the project's own benchmark page, run on a 16-core AMD EPYC machine. Its readiness document says the patches are not ready to merge.
A patch set called headstart makes cargo check on rust-analyzer 54% faster on a 16-core machine, according to benchmark tables the project published on its own repository. It changes both rustc and Cargo, it is not in any Rust release, and its readiness document says in so many words that it is "not yet ready to merge."
The repository, PowderworksCode/headstart, was created on September 27, 2026. GitHub's API lists its owner as an organisation, shows 14 stars and 0 open issues, and reports no licence. It reached 61 points on Hacker News on October 4 under the title "Emitting metadata early makes building/checking Rust up to twice as fast."
What headstart changes in rustc and Cargo
Today a Rust crate waits for every crate it depends on to be fully checked, function bodies included, before it starts. The headstart README argues that a dependent crate does not need those bodies to type-check itself, only the signatures.
The patches split the work. A new analysis_interfaces query in rustc finishes the interfaces first and writes an .early-rmeta file. Cargo, behind a -Zheadstart flag, starts downstream crates the moment that file exists, then pauses them and hands back job slots while it waits for the full metadata. The project counts six rustc patches and three Cargo patches.
Rust already does a version of this. Cargo's pipelined compilation starts a downstream crate when the .rmeta file appears, as the compiler team's April 5, 2019 notes describe. Headstart moves the signal earlier.
The benchmark numbers and the machines behind them
The project's results page reports clean builds on a Linux machine with an AMD EPYC 9554P and 16 jobs. Every figure below is the project's own measurement; no outside party has published a reproduction that we found.
| Project | cargo check saving | cargo build saving |
|---|---|---|
| rust-analyzer | 54% (90.1 s to 41.5 s) | 42% (137.9 s to 79.9 s) |
| polars | 49% | 29% |
| wasmtime | 47% | 36% |
| zed | 27% | 25% |
| lemmy | 26% | 30% |
- rust-analyzer54 %
- polars49 %
- wasmtime47 %
- zed27 %
- lemmy26 %
Source: headstart docs/results.md, project-supplied, accessed 2026-10-04
The README says the "up to 54%" and "up to 42%" figures cover 13 real projects, none of which slowed down, and that all 312 builds produced identical diagnostics in both modes.
The gain shrinks with core count. The README says the gains "come from cores the build would leave idle." On a 4-core Intel Xeon at 2.1 GHz with 15 GB of RAM, the results page lists rust-analyzer cargo check at 86.18 s against 65.36 s, a 24% saving. The same page labels that machine a "Claude Code container."
What the project says is not done
The README lists three costs: downstream work that gets thrown away when a dependency fails to type-check, errors that arrive slightly later, and more memory in use at once.
The readiness page is blunter. It says the series is "Ready for the design conversation, and for a draft PR to show with it. Not yet ready to merge." It lists a compiler-side Major Change Proposal still to be written and fresh 16-core and 8-core measurements.
It also names objections reviewers are likely to raise:
- The rule that early metadata can be swapped for full metadata is enforced only by a hand-maintained list of queries.
- Only Cargo can drive it, so sandboxed build tools such as Bazel and Buck2 cannot use it.
- Resumed compilations bypass jobserver tokens and can exceed a build's job limit.
- Saturated machines pay 3% to 7% extra CPU.
- It is untested on Windows and recent macOS.
The page says the design "answers the objection that stopped the 2019 attempt" but quotes no Rust maintainer, and nothing we read shows the compiler or Cargo teams have reviewed it.
How to try it, and what to wait for
Running it means building a patched rustc and Cargo with scripts/setup.sh, then calling cargo check -Zheadstart with RUSTC pointed at the patched compiler. With no licence file, it is a benchmark to read, not a tool to adopt.
The roadmap's next step is sending the first rustc commit as its own pull request, which the page says needs none of the rest.
Related: Dwarfstar and Agent-Reach, two other young repositories that drew stars quickly.
Sources
More in Software
- 01OpenCut Has 92,200 Stars, but the Editor People Use Is the Classic One and the Rewrite Is Not Taking ContributionsThe open-source CapCut alternative rebuilt its default branch in May. The README and a third-party walkthrough disagree on how much of the new code is Rust.
- 02Impeccable's Design Detector Runs Without a Model, but Its Open Issues Show Gaps Outside .htmlPaul Bakaus's design skill for coding agents ships 61 deterministic rules you can run from the command line. The bug tracker says where they are least reliable.
- 03A Hacker News Post Says Agents Need Documentation, Not Memory, and Its Author Wrote the Plugin That Does ThatKevin Liao's October 3 essay attacks snippet-recall memory plugins and promotes Operator Memory. A separate September essay argues the real gap is neither memory nor documents.
- 04Agent Reach, at 90,900 Stars, Reads X and Reddit for Your Agent Through Your Own CookiesThe MIT-licensed CLI routes agents to 20-plus sites with a backup backend per channel. Its README admits the login channels can get an account banned.