The quiet return to Go- and Rust-native development
Why the next decade of infrastructure — and increasingly, of product — will be written in Go and Rust, and what that means for how we build.
The pendulum is swinging back. For a decade, the default answer for a new backend was TypeScript on Node, or Python if there was a whiff of data. That answer optimised for velocity, and it was the right answer at the time. But the constraints that made it right — cheap compute, forgiving latency budgets, small teams shipping small services — are quietly disappearing.
Compute is no longer free. Every AI-native product has an inference bill that dwarfs its infrastructure bill, and every operator is now looking at the runtime layer with fresh eyes. A Go binary that idles at 12MB of memory and starts in 80ms is not a stylistic preference anymore — it is a line item on a P&L.
Rust, meanwhile, has stopped being the language you learn to feel smart and started being the language you reach for when correctness is a feature. The tooling — cargo, clippy, the analyzer — is now boring in the way great tooling should be. The moment you ship a data plane, a proxy, a signing service, or anything that sits in the hot path, Rust stops being an aesthetic choice.
The interesting shift is not that Go and Rust are winning benchmarks. They have won benchmarks for years. The shift is cultural: the industry has re-learned that boring, deterministic, statically-typed systems compound. They make on-call quieter. They make refactors safer. They make the cost curve of a company legible.
None of this is a repudiation of TypeScript. The product surface still belongs to it, and probably always will. But the layer underneath — the queues, the workers, the gateways, the schedulers, the AI orchestration — is quietly, unmistakably, being rewritten.
The teams I respect most are already there. They ship a thin TypeScript edge, a fat Go core, and a small handful of Rust services where the physics demand it. That is the shape of the next decade, and it looks a lot like the shape of the best systems of the last one.