Why I Needed Rust, Finally
Infrastructure automation I can sleep on — Rustikon 2026, Warsaw
Why I Needed Rust, Finally
Rustikon 2026 · Warsaw · 19 March 2026.
▤ Read the slides — real text, not screenshots
The talk is the story of a long resistance to Rust and of what broke it. Provisioning is what came out the other side, so this page reads the talk from the platform’s side: which of its claims you can check in the code.
Failure moves to the schema
The argument of the talk is that Rust’s value for infrastructure is not speed: it moves the failure left, from the deploy to the compiler. Provisioning applies the same move to configuration. Every piece of infrastructure is a Nickel schema, validated before anything reaches runtime; TOML and YAML are generated output, never the source of truth. A malformed server, a missing dependency or an impossible plan is rejected at the schema, not discovered on a node.
Rust where the guarantees have to hold
The orchestrator, the Control Center and the platform services are Rust: provider calls, SSH verified against pinned host keys, Cedar policy decisions and state in SurrealDB. Operations report back as events on NATS JetStream, the single broker between services. That is the part of the system where “it compiled” has to mean something.
Infrastructure you can sleep on
The title is not a slogan. When the declaration is the source of truth and it is typed, and the platform refuses what does not fit, being on call stops being an act of faith. What is declared is what runs, and what runs is checked against what was declared.
The talk is in English. The slides are on this site as real text, in English and Spanish.
· reads