I Shipped a Rust App Without Writing a Line of Rust

I’m Leon Shelhamer, and in December of 2024 I built and shipped a desktop application with a Rust backend, having never written a line of Rust in my life. I still haven’t, really. The thing worked anyway.

I tell that story in interviews and watch people decide, in real time, whether I’m a fraud or whether I just described the future of their engineering org. I understand the reaction. A few years ago I’d have had it too.

I’ve dropped this app into other posts as a one-line flex about AI in my workflow. I want to actually take it apart this time, because the lesson sitting underneath it is the single most useful thing I know about how to work as an engineer right now - and it has almost nothing to do with Rust.

What I actually built

It was an Electron app - that part I know cold, I’ve been doing desktop work in that ecosystem for a long time - but it needed to do real work that JavaScript had no business doing. The kind of work you want compiled, fast, and close to the metal. The right answer was native binaries. The right language was Rust.

The wrong answer, the one my old instincts would have reached for, was: spend three weeks learning enough Rust to be dangerous, ship something mediocre, and call it growth. Instead I sat down with ChatGPT and treated it like a Rust engineer who never got tired of my questions. I described what I needed each binary to do, in the precise terms a systems engineer would - inputs, outputs, error modes, the contract at the boundary. It wrote the Rust. I read it, pushed on it, made it justify its choices, wired it into the app, and tested it against reality until it held.

Roughly ninety-five percent of that Rust was AI-generated. I could not have written it from a blank file. But I could tell when it was wrong, I knew exactly what the binary was supposed to be, and I knew where it fit in the system. The app shipped.

The skill that mattered wasn’t Rust

Here’s the part people miss. The reason that worked isn’t that AI is magic. It’s that I’d spent sixteen years learning what a well-behaved binary looks like, what a clean boundary between two layers looks like, what “this will leak memory” smells like even in a language whose syntax I can’t write. The language was the part I didn’t have. The engineering was the part I did.

That’s the multiplier nobody priced in correctly. My effective range is no longer “the languages I’ve memorized.” It’s “the systems I can reason about” - which, after this long, is most of them. Drop me into a Go service, a Python pipeline, a Rust binary, and I will be useful on day one, not week three, because the hard part was never the syntax. The hard part was always knowing what to build and how to tell when it’s wrong.

Why this should matter to anyone hiring

The old hiring question was “does this person already know our stack?” It made sense when learning a new stack cost months. It costs days now, for the right kind of engineer - the kind whose value is judgment and whose typing is the least interesting thing about them.

I’m not telling you to hire people who can’t code. I’m telling you the bottleneck moved. It’s no longer “can they write the language.” It’s “can they reason about the system, smell the bugs, and own the result.” I can. I proved it to myself with a Rust app I couldn’t have hand-written, and it’s been my default mode of operating ever since.

Leon Shelhamer is a senior software engineer with sixteen years of experience and the developer behind Two Eleven Jay and the managed-hosting service uphost.ly. He works in React and Vue, runs his own production infrastructure, and builds AI-augmented development workflows from upstate New York. Connect on LinkedIn.