In my previous post about Ergonomic Ref Counting, I talked about how, whatever else we do, we need a way to have explicit handle creation that is ergonomic. The next few posts are going to explore a few options for how we might do that.

This post focuses on explicit capture clauses, which would permit closures to be annotated with an explicit set of captured places. My take is that explicit capture clauses are a no brainer, for reasons that I’ll cover below, and we should definitely do them; but they may not be enough to be considered ergonomic, so I’ll explore more proposals afterwards.

This post presents a proposal to extend Rust to support a number of different kinds of destructors. This means we could async drop, but also prevent “forgetting” (leaking) values, enabling async scoped tasks that run in parallel à la rayon/libstd. We’d also be able to have types whose “destructors” require arguments. This proposal – an evolution of “must move” that I’ll call “controlled destruction” – is, I think, needed for Rust to live up to its goal of giving safe versions of critical patterns in systems programming. As such, it is needed to complete the “async dream”, in which async Rust and sync Rust work roughly the same.

Continuing my discussion on Ergonomic RC, I want to focus on the core question: should users have to explicitly invoke handle/clone, or not? This whole “Ergonomic RC” work was originally proposed by Dioxus and their answer is simple: definitely not. For the kind of high-level GUI applications they are building, having to call cx.handle() to clone a ref-counted value is pure noise. For that matter, for a lot of Rust apps, even cloning a string or a vector is no big deal. On the other hand, for a lot of applications, the answer is definitely yes – knowing where handles are created can impact performance, memory usage, and even correctness (don’t worry, I’ll give examples later in the post). So how do we reconcile this?

This post describes SymmACP – a proposed extension to Zed’s Agent Client Protocol that lets you build AI tools like Unix pipes or browser extensions. Want a better TUI? Found some cool slash commands on GitHub? Prefer a different backend? With SymmACP, you can mix and match these pieces and have them all work together without knowing about each other.

This is pretty different from how AI tools work today, where everything is a monolith – if you want to change one piece, you’re stuck rebuilding the whole thing from scratch. SymmACP allows you to build out new features and modes of interactions in a layered, interoperable way. This post explains how SymmACP would work by walking through a series of examples.

Right now, SymmACP is just a thought experiment. I’ve sketched these ideas to the Zed folks, and they seemed interested, but we still have to discuss the details in this post. My plan is to start prototyping in Symposium – if you think the ideas I’m discussing here are exciting, please join the Symposium Zulip and let’s talk!

There’s been a lot of discussion lately around ergonomic ref-counting. We had a lang-team design meeting and then a quite impactful discussion at the RustConf Unconf. I’ve been working for weeks on a follow-up post but today I realized what should’ve been obvious from the start – that if I’m taking that long to write a post, it means the post is too damned long. So I’m going to work through a series of smaller posts focused on individual takeaways and thoughts. And for the first one, I want to (a) bring back some of the context and (b) talk about an interesting question, what should we call the trait. My proposal, as the title suggests, is Handle – but I get ahead of myself.

This blog post gives you a tour of Symposium, a wild-and-crazy project that I’ve been obsessed with over the last month or so. Symposium combines an MCP server, a VSCode extension, an OS X Desktop App, and some mindful prompts to forge new ways of working with agentic CLI tools.

Symposium is currently focused on my setup, which means it works best with VSCode, Claude, Mac OS X, and Rust. But it’s meant to be unopinionated, which means it should be easy to extend to other environments (and in particular it already works great with other programming languages). The goal is not to compete with or replace those tools but to combine them together into something new and better.

In addition to giving you a tour of Symposium, this blog post is an invitation: Symposium is an open-source project, and I’m looking for people to explore with me! If you are excited about the idea of inventing new styles of AI collaboration, join the symposium-dev Zulip. Let’s talk!