Building ambitious software — Jonathan Kelley, Dioxus Labs & Cognition

The Dioxus team got excited, maxed out their coding agent subscriptions, and turned out tens of thousands of lines of Rust covering features they had wanted for years. Almost none of it cleared the bar for merging. Those lines sat in draft, and Jonathan Kelley says they are sitting there still. He calls the failure mode becoming a slop cannon. Kelley founded Dioxus Labs five years ago, spending his last undergraduate summer on a cross platform Rust app framework instead of taking an internship, and the project now carries roughly 37,000 GitHub stars and an estimated 200 million cumulative end users across apps from voting software to collision avoidance for satellites. Getting there meant building almost everything from scratch, including a rendering engine with a browser grade CSS engine lifted out of Firefox, and a hot reload engine that patches running native code in place in about 100 milliseconds. The interesting turn is what changed when the agents got good at Rust. Kelley's team spent five years trying to file down Rust's learning curve, and now treats that curve as a feature, because the agents absorb the borrow checker and the edge cases on the developer's behalf. He is specific about where this pays and where it does not. Deeply integrated Kotlin and Swift build plugins went from years of hand written work to weeks. Release checklists, backports, and documentation accuracy became tractable for a team of three. Tests did not, because agents write tests for any API but rarely the right ones, though they excel at fuzzing harnesses. His verdict: code is cheap now, quality is not, and architecture is where the time goes. Speaker info: - https://www.linkedin.com/in/jonathan-r-kelley - https://jonathan-kelley.com - https://github.com/dioxuslabs/dioxus Timestamps: 0:00 - Five years ago, a first commit 2:49 - Where the project stands now 4:35 - Blitz and Subsecond 6:15 - The moment agents got good at Rust 7:07 - Why a hard language became an advantage 10:32 - Agents as patient specialists 13:04 - The mundane work worth automating 15:37 - Where tests fall short, where fuzzing wins 16:28 - Architecture is still an art 17:19 - Reading code over writing code