Rust app frameworks and developer tools
Dioxus Labs
Dioxus Labs develops Dioxus, an open source Rust framework for building web, desktop and mobile applications from one codebase. Developers write interfaces with React-like syntax, style them with HTML and CSS, and add backend functionality through server functions integrated with Axum. Its CLI handles development and application bundling, while signals-based state management supports reactive interfaces.
Founded in 2023 by solo founder Jonathan Kelley, Dioxus Labs grew from a college side project into a full-time engineering team. Its tools address both iteration speed and rendering: Subsecond hot-patches running Rust code without restarting the application, allowing edits while retaining app state. Blitz is a modular HTML/CSS rendering engine underlying Dioxus Native, which renders interfaces without a webview. Blitz remains in beta, with bugs and missing features.
The Dioxus team joined Cognition in September 2026 to work on Devin’s virtual machine, computer use and testing capabilities. Cognition already used Dioxus in Devin, including its CLI terminal renderer. The transition includes continued support for Dioxus, Blitz, Taffy and Subsecond, with increased investment in native rendering. Dioxus Labs described the move as allowing continued open source development without the financial pressure to monetize its framework.
Explore the recordings
The supplied Dioxus Labs archive contains one recording. Use the paths below to explore its application-building foundations and its lessons from coding agents. Technical capabilities, adoption figures, and team experiences are claims reported in the recording, not independently verified facts about Dioxus Labs today.
Start with the application stack
Building ambitious software follows Jonathan Kelley’s account of starting Dioxus in 2021 and building supporting infrastructure for cross-platform Rust applications using HTML and CSS. Follow this path for his discussion of Blitz’s HTML/CSS rendering and Subsecond’s native-code patching. Kelley reports roughly 100-millisecond patching; treat that as a recorded performance claim rather than a current benchmark or product guarantee.
Follow the coding-agent lessons
Return to the same recording for the team’s reported shift from generating large volumes of code to using agents for specialized integrations, debugging, release work, and documentation. Kelley describes generated code that failed the team’s merge-quality threshold and contrasts API-test generation with more useful fuzzing harnesses. This path is useful for examining the talk’s central argument: faster generation still requires deliberate architecture, maintainable foundations, precise communication, and line-by-line review. These are Kelley’s reported experiences and conclusions, not a claim about every coding agent or today’s tools.
1 talk
Newest first
1 speaker at AIE
Affiliations reflect their AIE appearances, not necessarily current employment.