Levien's Next-100 Rust GUIs Post Counts About 20 New Toolkits and Bets Against egui's Immediate Mode
Software / analysis
Levien's Next-100 Rust GUIs Post Counts About 20 New Toolkits and Bets Against egui's Immediate Mode
The Linebender author credits AI rewrites for the surge; egui's own README concedes the trade-off he is pressing on.

Raph Levien's case that "building high-performance UI in Rust is a practical alternative" to Electron and the platform toolkits comes with a number attached: roughly 20 Rust GUI toolkits have been announced since his last survey, and he still expects more.
Levien published "Advice for the next 100 Rust GUIs" on the Linebender site on November 21, 2026. It follows his July 2022 post, "Advice for the next dozen Rust GUIs", and opens a longer bet than that one: "I predict there will be a small explosion of trying lots of different things." He credits "a recent rapid acceleration of interest in Rust GUI, fueled by AI."

What Levien claims about AI and Rust rewrites
The post's most quotable line is a causal one: "Rust is rapidly becoming the language for AI rewrites of existing codebases." His reasoning is that "the LLM cost of writing Rust code is not significantly higher than other languages", so a team already rewriting for speed pays little extra to choose Rust.
He names two examples as his own reading of events. He writes that "GPUI is emerging as a major choice for AI-driven rewrites, most notably Discord," and that Zed editor is "built on their own GPUI framework." He says Shopify made a related call: "abandoning React Native."
Those are claims in a blog post by a toolkit author. The post does not link a Discord or Shopify announcement for them, and this piece did not find one, so treat them as his account.
The second 90%: why Levien bets against immediate mode
The sharpest passage is about egui, an immediate-mode toolkit that has 32.4k stars on GitHub. Levien calls it "a really good implementation of the immediate mode paradigm in Rust" and then argues the model does not scale: "In short, it's good at the first 90%, but it makes the second 90% harder."
The second 90% is the job of anyone shipping a design tool or an editor. Immediate mode redraws the interface description every frame, and Levien says production use needs "stable widget identities" and cached layers that the model pulls against. He writes that this "is a problem for egui in particular, as it has to build and diff a tree each animation frame."
He is not dismissive about every use. "For some applications (most especially development support for games) it works well." But "it wouldn't be my first choice for productivity and design applications," and he adds that he is "not impressed by the choice of the ArtCraft projects to use egui."
What egui's own README says back
egui's repository, under the emilk account on GitHub, does not claim otherwise. Its README states the trade-off in its own terms: "immediate mode GUI libraries are easier to use, but less powerful," and it names layout as the main disadvantage. It does not hide the cost of the model it is built on.
It is also candid about maturity. "egui is still work in progress," it says, and "If you want something that doesn't break when you upgrade it, egui isn't for you (yet)." The repository shows 32.4k stars and 974 issues on the page captured November 22.
- egui (immediate mode)32.4 stars (thousands)
- Xilem (retained, Linebender)5.6 stars (thousands)
Source: GitHub repository pages for emilk/egui and linebender/xilem, accessed 2026-11-22
Stars are not adoption. They show which project people noticed, and the immediate-mode one has nearly six times the stars of the Linebender one. Levien's argument is about what survives a second year of feature requests, which no star count measures.
What Xilem is, and who is better off
Levien's post sits on Linebender's own blog and Xilem is a Linebender project, so the post is an argument from inside one camp and not a neutral survey. Xilem, its reactive layer, describes itself in its README as "An experimental Rust architecture for reactive UI," built on the Masonry retained widget toolkit. The page lists 106 open issues, 51 open pull requests, an Apache-2.0 licence and a minimum Rust version of 1.96.
| Project | Licence | README status wording |
|---|---|---|
| egui | MIT OR Apache-2.0 | "still work in progress", breaking changes in new releases |
| Xilem | Apache-2.0 | "An experimental Rust architecture" |
Both describe themselves as unfinished. A developer picking one today for a dashboard or a debug overlay gets a working result sooner with egui. A developer planning a two-year design application should read the post's other argument, that text layout and accessibility are largely solved in Parley and AccessKit while the system compositor is not.
What is still open after Levien's post
Levien lists the compositor as the unsolved layer: hardware-accelerated layers matter for video and low-end devices, and he says toolkits rarely integrate them. He also asks directly, "Can we get silky smooth UI even on low end Android phones?" and does not answer with a benchmark.
No framerate, memory or startup figure appears in the post. The next test is whether any of the roughly 20 new toolkits ships a customer product with numbers attached, and Levien gives no date for that. See how star counts misled on another popular repo and what a 13,884-star agent actually shipped before reading too much into either number.
Sources
More in Software
- 01C++26 Is Reopened: The 13–9–12 Ballot, the Enforced Label That Answers It, and What Is Still OpenSutter's Búzios report says the committee kept contracts, added one paper, and will send C++26 to a second ballot with no published date.
- 02OpenAI Says Codex Cannot Run in Docker for iOS Builds, and Docker Launches It With Codex's Sandbox OffBolin's November 18 post lists Seatbelt, Bubblewrap and MXC as Codex's three sandboxes; Docker's page defaults to --dangerously-bypass-approvals-and-sandbox.
- 03CLI-Anything Has 62,119 Stars and a README With Three Different Test TotalsThe HKUDS project turns desktop software into agent-callable CLIs; its own per-harness lines sum to 158 more tests than it reports.
- 04Brazil's Receita Federal Posts Tax Calculator Under MIT With 16 Open IssuesThe agency published three repositories on Nov. 16 for the new consumption taxes, and outside developers had filed bug reports against the first within days.