Hexcat

September 2026

This is a summary of what we are currently working on at Hexcat.

Each section below links to the relevant Rust Project Goal, where you can find more in-depth technical updates. This page only provides a high level update.

This work is done in collaboration with other contributors not part of Hexcat, including newer contributors we mentor to help onboard them to the Rust project, improving the long-term health of the project.

Beyond these goals, we continue to review and clean up code, help run meetings for the Rust Types Team and Library Team, and respond to regressions and other critical issues.

Our work is only possible thanks to the funding and support from our clients and sponsors, including Amazon, Google and Turbopuffer, and Pydantic. 💛 If you'd like to support our work or have suggestions for these updates, please contact mara@hexcat.nl or lcnr@hexcat.nl.

Next-generation trait solver

Rust Project Goal: Next-generation trait solver

This work is currently driven by lcnr.

We continued working on the stabilization report and discussed it with the Types Team. We also continued triaging and fixing minor remaining issues. We proposed the stabilization FCP in #164209, which targets Rust 1.101!

Next-gen region constraints

Rust Project Goal: Next-gen region constraints

This work is currently driven by Boxy and lcnr and was previously called "Assumptions on Binders".

We have continued making good progress on the higher ranked region rework.

Ashley has significantly improved the testing DSL making it much more fully featured compared to the initial implementation landed last month, it's now capable of expressing most things we need it to.

Boxy has spent most of this month thinking through how to merge the new region constraint representation with the existing stable representation as well as how to handle some soundness issues arising from leaking inference variables from binders. At this point we now have a good handle on how to tackle both of these and should be able to make good implementation progress.

Implicit default auto traits

Rust Project Goals: Extern types and SVE and Move and Forget

Work on Move and Forget is currently driven by lcnr while collaborating with David Wood and Rémy Rakic on the type system work needed for extern types and SVE.

Jana has continued to work on a performance optimization for the way we represent the current assumptions in scope to reduce the performance impact of adding new implicit bounds everywhere. Our main focus has been the next-solver stabilization however.

Const generics

Rust Project Goal: Const generics

This work is currently driven by Boxy and Ashley.

We made significant progress on Generic Const Args, enough for Boxy to post an update to the Inside Rust Blog: Generic Const Args and You. This progress includes things like Ashley removing the type const syntax and moving to a macro-based approach, and making the feature gate story more comprehensible and clear. Ashley has also done significant internal compiler infrastructure refactors that unblock future work. The removal of feature(generic_const_exprs) will happen shortly.

Custom Allocators

Rust Project Goal: Allocators 1.0

This work is driven by Nia.

Nia's stabilization of the Allocator trait and the allocator parameters on Box and Vec was merged on September 23. These APIs are now stable on nightly and are expected to reach the stable channel in Rust 1.100! 🎉

Thanks to Pydantic for sponsoring this goal. 💛

There's more!

Fallible continues to work on the HIR type checking chapter of the rustc-dev-guide as part of the type system documentation goal.

Ashley is now a co-lead of the Rust Const Generics Project Group together with Boxy.

Nia now leads the Rust Library Team.

We've hired Sasha Pourcelot and Nicholas Nethercote to join Hexcat in October to work on the Compiler performance optimizations project goal. This is made possible thanks to sponsorship from Google, OpenAI, Devolutions, and Fulfil.io. 💛 We are still looking for more sponsors for this goal. If you are interested please contact mara@hexcat.nl.

See you next month for more updates!

August 2026

This is a summary of what we are currently working on at Hexcat.

Each section below links to the relevant Rust Project Goal, which contains more in-depth technical updates. This page only provides some additional context. This work is done in collaboration with other contributors not part of Hexcat. Mentoring others is a great way to onboard new contributors to the Rust project, improving its longterm health.

Outside of these Project Goals, we are always doing a significant amount of other work that positively impacts Rust. We are doing a lot of mentoring, reviews, and cleanup work, drive the meetings for the Rust Types Team and Rust Library Team, and respond to regressions and other critical issues.

Our work is only possible thanks to the funding and support from our clients and sponsors, including Amazon, Google, Turbopuffer, and Pydantic. 💛 If you'd like to support our work or if you have any input on how to make these updates more effective, please contact mara@hexcat.nl or lcnr@hexcat.nl.

Next-generation trait solver

Rust Project Goal: Next-generation trait solver

This work is currently driven by lcnr.

The new solver is now enabled on nightly by default for testing and we expect to start open the stabilization proposal in about a month! This continues to take up most of lcnr's and Jana's time.

Next-gen region constraints

Rust Project Goal: Next-gen region constraints

This work is currently driven by Boxy and lcnr and was previously called "Assumptions on Binders".

Boxy has made a lot of progress the last month with the help of Ashley and João Soares. We merged an initial implementation for a testing DSL and are making steady progress on the implementation of the feature. A lot of the necessary work is tracked on the the Assumptions On Binders GitHub repository.

Implicit default auto traits

Rust Project Goals: Extern types and SVE and Move and Forget

Work on Move and Forget is currently driven by lcnr while collaborating with David Wood and Rémy Rakic on the type system work needed for extern types and SVE.

Progress is happening, but rather slowly as we're mainly focussed on the next-solver stabilization.

Const generics

Rust Project Goal: Const generics

This work is currently driven by Boxy and Ashley.

There has also been a lot of progress here. We're still working towards the removal of feature(generic_const_exprs) and are spending a lot of effort on the design and implementation of feature(generic_const_args).

There's more!

Not all our work is tracked by the above Rust Project Goals. Here are some of the highlights of our other work:

lcnr has been helping out Waffle with the type system changes necessary for the stabilization of the never type. This stabilization has now been merged! https://github.com/rust-lang/rust/pull/155499

Fallible has been improving the compiler's type system documentation with reviews from Boxy. She expanded the explanation of MIR type checking and clarified the documentation for the compiler's inference context, while continuing work on the chapter on eager type inference.

Nia is making progress towards stabilizing the Allocator API for custom allocators. Her refactoring of the Allocator API was merged this month, followed by documentation and naming cleanups. She also added an Allocators Project Goal to coordinate the work for stabilization and broader standard library support.

Nia's RFC to simplify and reorganize the library team was accepted and implemented this month! It brings the libs, libs-api, and libs-contributors teams together and sets out clearer responsibilities and decision-making procedures.

Nia has also been reviewing unsafe code in the standard library. Her work to improve safety comments throughout alloc and extend automated checks for undocumented unsafe blocks has now landed, making this code easier to review and maintain.

See you next month for more updates!

July 2026

This is a summary of what we are currently working on at Hexcat.

Each section below links to the relevant Rust Project Goal, which contains more in-depth technical updates. We only provide a high-level overview here. This work is done in collaboration with other contributors not part of Hexcat. Mentoring others is a great way to onboard new contributors to the Rust project, improving its longterm health.

Outside of these Project Goals, we are always doing a significant amount of smaller work that positively impacts Rust. We are doing a lot of mentoring, reviews, and cleanup work, drive the meetings for the Rust Types Team and Rust Library Team, and respond to regressions and other critical issues.

Our work is only possible thanks to the funding and support from our clients and sponsors, including Amazon, Google, and Turbopuffer. 💛 If you'd like to support our work or if you have any input on how to make these updates more effective, please contact mara@hexcat.nl or lcnr@hexcat.nl.

Next-generation trait solver

Rust Project Goal: Next-generation trait solver

This work is currently driven by lcnr.

Given the amount of on-going work here, we've switched to a weekly update cadence here. So there have been 4 update this month.

The last month has been incredibly productive. There have been a lot of huge performance improvements, see especially the most recent update.

We intend to enable the new trait solver by default on nightly in the next few days as all major issues have been fixed. We should then be able to properly stabilize it within the next 2–3 months.

Next-gen region constraints

Rust Project Goal: Next-gen region constraints

This work is currently driven by Boxy and lcnr and was previously called "Assumptions on Binders".

Boxy and lcnr spent a lot of time discussing this feature and we now have a fairly clear roadmap towards stabilization of an MVP by roughly the end of this year. This MVP should be slightly stronger than the existing -Zhigher-ranked-assumptions flag, see the list of known fixed issues for a vague idea of what this will cover.

We also realized that Assumptions on Binders is a poor description of what we're doing as it's far more general than this: we're entirely rewriting how region constraints are handled in the type system. Boxy rewrote the Project Goal to correctly explain the larger scope.

Implicit default auto traits

Rust Project Goals: Extern types and SVE and Move and Forget

Work on Move and Forget is currently driven by lcnr while collaborating with David Wood and Rémy Rakic on the type system work needed for extern types and SVE.

Not much has happened here from our side as we've mainly been pushing to stabilize the next-generation trait solver. It is slowly moving along and we should get support for Move on nightly during the next few months.

Const generics

Rust Project Goal: Const generics

This work is currently driven by Boxy and Ashley.

We're also doing weekly updates for Const Generics as it's quite active. This month has mainly been a lot of implementation work polishing feature(generic_const_args).

We're now in a good spot to remove feature(generic_const_exprs) and doing so is mainly blocked on writing a blogpost explaining how to migrate to generic_const_args. We need to do so before stabilizing the next-generation trait solver as it does not support generic_const_exprs.

There's more!

Not all our work is tracked by the above Rust Project Goals. Here are some of the highlights of our other work:

lcnr has been more proactively getting involved in the Project Goals team to make it more useful for contributors, project leadership, the wider community, and funders. Please reach out to them if you have any thoughts here!

Boxy and Fallible continue to work on improvements to the rustc-dev-guide. Fallible is currently focussed on type inference subtleties during HIR typeck after finishing the chapter on well-formedness.

Nia published the binary size roadmap.

Nia re-instantiated the custom allocators working group and is making progress towards stable custom allocators.

See you next month for more updates!

June 2026

This is a summary of what we are currently working on at Hexcat.

As Rust is an open source project, a lot of this work is done in collaboration with other contributors not part of Hexcat. We try delegate work and mentor others as much as possible. This is a great way to onboarding new contributors to the Rust project, increasing its longterm health and making us more effective longterm.

Each section below links to the relevant Rust Project Goal, where you can find the motivation behind the features and more in-depth technical updates. This page only provides a high-level overview of our contributions in these area.

Our work is only possible thanks to the funding and support from our clients and sponsors. If you'd like to support our work or if you have any input on how to make these updates more effective, please contact mara@hexcat.nl or lcnr@hexcat.nl.

Next-generation trait solver

Rust Project Goal: 2026/next-solver

This work is currently driven by lcnr.

For the new trait solver, our main focus has been improving the compile-time performance to avoid performance regressions and a redesign of the way we handle the normalization of aliases, e.g. associated types.

The most significant performance regression from the existing implementation in our benchmarks was wg-grammar. Over the course of the last month, we've improved its performance from being 18.7× slower than the old implementation to 1.6× slower via three different PRs. This is a more than 10× performance improvement!

We have designed and implemented a new normalization approach, fixing the largest remaining issue blocking stabilization. This was a huge effort requiring a new design and then implementing this over the course of the last month. See the last Project Goal update for more details.

With this we intend to do another crater run before enabling the next-generation trait solver on nightly by default for testing. This should hopefully uncover any remaining issues we're not aware off at which point we're then on the final stretch towards stabilization.

Assumptions on binders

Rust Project Goal: 2026/assumptions_on_binders

This work is currently driven by Boxy and lcnr.

Boxy has been driving this work and we have merged an initial implementation of this feature after roughly 6 months of work.

With this initial implementation merged, she is now focussing on a lot of minor issues and cleanups. We also need to add a special DSL for testing, as getting sufficient test coverage by writing actual Rust programs is not feasible. This is also a great place to onboard new contributors to this area of the compiler.

Implicit default auto traits

Rust Project Goals: 2026/scalable-vectors and 2026/move-trait

Work on Move and Forget is currently driven by lcnr while collaborating with David Wood and Rémy Rakic on the type system work needed for extern types and SVE.

Adding new implicit trait bounds has two main difficulties. We need to minimize the performance cost of adding additional trait bounds everywhere. Adding such new bounds without any additional performance work is unacceptable, regressing performance by up to 7400% in some stress-tests.

We want to add support for these traits on nightly. To do this without having fully resolved the performance concerns, we add fast-paths for crates which do not enable the corresponding features. There is already unstable support for MetaSized which was originally implemented by David Wood. We're close to also merging unstable support for Move and I expect that adding Forget afterwards will not be too challenging.

The next-generation trait solver alleviates some of the performance impact and also enables more optimization, so stabilizing Move and Forget is blocked on its stabilization.

Finally, we are actually working on targeted performance optimizations for these traits.

We also need to make sure adding these new bounds does not cause breaking changes and enable library authors to remove these implicitly added bounds over time. New types which implement e.g. Move can otherwise not be used with large parts of the ecosystem.

For this we have been working on changing minor edge cases of the type system to avoid breakage, e.g. candidate preference in the trait solver.

We are also working on adding new language features to enable users to migrate off these default bounds by creating a new Library Trait Evolution Initiative.

Similarly, lcnr has been collaborating with @lqd to support weakening errors involving generic parameters to a FCW, see their blogpost for a better overview of this feature.

Const generics

Rust Project Goal: 2026/const-generics

This work is currently driven by Boxy and Ashley.

We are currently splitting some of the more uncontroversial types out of feature(adt_const_params) to stabilize them separately while leaving more involved user defined types for later.

We are seperately doing a lot of implementation and design work for feature(generic_const_args). One of our near-term goals is to get this into a good enough state to remove support for feature(generic_const_exprs), as the implementation of that feature has fundamental issues while negatively impacting the work on other parts of the compiler.

There's more!

Not all our work is tracked by the above Rust Project Goals. Here are some of the highlights of our other work:

lcnr has been mentoring work on migrations necessary for the stabilization of the never type, two of the three have already been implemented and approved by the Types Team: #157820 #156047.

Boxy and Fallible are working on improvements to the rustc-dev-guide. Last month they have been focussing on explaining "well-formedness".

Jana continues to host the twice-weekly Compiler Team Office Hours where Rust compiler team members and contributors meet in a video call to discuss their work.

Mara has started the initial work for organizing the Rust All-Hands 2027, which will take place in the Netherlands again, as part of RustWeek 2027.

Nia is working on creating a roadmap and Rust Project Goal for binary size improvements.

See you next month for more updates!