About

Leonid with his sons

Leonid Bugaev — Human, Father, Nomad, Engineer

I’m Leonid — buger across most of the internet. For close to a decade I’ve led engineering at Tyk, the API management platform, building global teams and the systems that connect other people’s systems. Before and alongside that, I’ve been an open source maker: GoReplay, the traffic replay tool I started in 2013, became the de-facto standard for shadow testing with over 18,000 GitHub stars, and jsonparser is quite literally the fastest JSON parser in Go — a title it has held for years.

Proof: Verification Is My Main Focus

Founder, Proof — continuous correctness audits for critical infrastructure. Every finding backed by a runnable reproducer.

Proof is where nearly everything I’ve learned converges. AI now writes code faster than any team can honestly review it, and the industry’s answer — more tests, more eyeballs — doesn’t scale. Tests check behavior on specific inputs; they can’t tell you whether the code matches the intent behind it.

Proof takes a different path: it audits systems continuously, not at a point in time, and grounds the work in formal methods — requirements written as machine-checkable specifications, proven consistent and realizable with SMT solvers and realizability checking, and traced bidirectionally to code, tests, and documentation. When something drifts, you see the blast radius immediately.

And there’s one rule I hold above everything: every finding is backed by a runnable reproducer. No speculative warnings, no static-analysis noise to argue with — if Proof reports a problem, you can run it and watch it fail. That’s the bar I think correctness tooling has to meet, especially for the critical infrastructure — aerospace, automotive, medical, finance — where “probably fine” isn’t an answer. Proof generates the compliance evidence those industries need (NPR 7150, DO-178C, ISO 26262) as a by-product of verification, and brings capabilities like MC/DC coverage, historically locked in $50K/year aerospace toolchains, to modern languages.

Probe Labs

Proof sits on top of an ecosystem I build at Probe Labs:

  • Probe — a semantic code search engine written in Rust that combines ripgrep speed with tree-sitter AST parsing, so AI agents get complete, structured code context on multi-million line codebases with zero setup.
  • Visor — an AI workflow engine that orchestrates code review, assistants, and automation through YAML-driven pipelines, running the same config as a GitHub Action, CLI, Slack bot, or HTTP API.

Together they form a stack where AI can understand code, act on it, and — through Proof — have its output verified.

The Longer Arc

My whole career has circled one question: how do you know your software actually works? GoReplay answered it with real production traffic. Tyk answered it with battle-tested infrastructure serving billions of API calls. Proof answers it the way I believe the AI era demands: with mathematics, continuous audits, and evidence you can run.

Beyond Code

I’m a father first. My boys and I travel the world together — I write about nomad parenting, slow parenting, and why having kids is the best thing that can happen to your career. The balance between family, open source, and wanderlust isn’t a side note to the engineering — it’s the point.

Get in Touch

The best ways to reach me: