Reddit The heart of the internet

Izzo Ever write a spec that becomes instantly outdated the moment someone touches the code?

Izzo You're listening to Exploring Next, episode two-eighteen. I'm Izzo, and with me as always is Boone. Today we're diving into CodeSpeak — a new programming language from Andrey Breslav, the creator of Kotlin.

Boone And this isn't just another syntax experiment. CodeSpeak flips the whole development model — you write specifications first, and the language generates executable code from those specs.

Izzo Which sounds amazing in theory, but I've seen too many 'revolutionary' approaches crash into reality. What makes this different, Boone?

Boone Well, Breslav's track record helps — Kotlin went from side project to Android's preferred language. But the real difference is CodeSpeak treats specifications as first-class citizens, not documentation afterthoughts.

Izzo Okay, break that down for me. How does spec-first actually work in practice?

Boone You start by defining contracts — input types, output types, invariants, pre and post conditions. The CodeSpeak compiler then generates implementation scaffolding that's guaranteed to respect those contracts.

Izzo So instead of writing code and hoping it matches the spec, the spec IS the code?

Boone Exactly. And here's the clever part — CodeSpeak uses dependent types and formal verification under the hood. Your specifications aren't just comments, they're mathematically provable constraints.

Izzo That sounds incredibly powerful for teams dealing with complex business logic. No more 'well, the code works but it doesn't match what we documented.'

Boone Right. And the type system is genuinely impressive. It can infer relationships between inputs and outputs that would take pages of documentation to describe manually.

Izzo What's the learning curve like? Because if this requires a PhD in type theory, adoption's gonna be rough.

Boone Surprisingly approachable. Breslav designed it with gradual adoption in mind — you can start with simple contracts and add complexity as needed. Plus it compiles to JVM bytecode, so it plays nice with existing Java and Kotlin codebases.

Izzo JVM interop is smart. No one wants to rewrite their entire stack for a new paradigm.

Boone The tooling is where it gets really interesting. CodeSpeak includes a specification playground where you can test contracts before implementation. It's like having a REPL for your business logic.

Izzo Now that's addressing a real pain point. I can't count how many times we've discovered spec issues only after building the feature.

Boone And the generated code includes built-in monitoring and telemetry. When contracts are violated at runtime, you get detailed traces showing exactly where the specification broke down.

Izzo Okay, I'm genuinely impressed. This feels like it could change how product teams collaborate with engineering. Specs become living, testable artifacts instead of dusty documents.

Boone The implications for API design are huge too. Your OpenAPI specs could literally generate the server implementation, with contract validation baked in.

Izzo I'm giving this a solid A-minus. The only question is whether teams will embrace the discipline of spec-first development.

Boone Fair point. It requires a mindset shift — you can't just hack together a solution and document it later.

Izzo But for teams already struggling with specification drift, this could be transformative. What should people try first?

Boone Install the CodeSpeak CLI from codespeak.dev and work through their specification tutorial. Start with a simple API endpoint — define the contract first, then let CodeSpeak generate the implementation skeleton.

Izzo Also worth experimenting with their contract playground. Upload an existing API spec and see how CodeSpeak would implement it differently.

Boone And if you're feeling ambitious, try building a small microservice entirely spec-first. I'm definitely adding this to the weekend project list — want to see how it handles complex state machines.

Izzo CodeSpeak might just be the bridge between product requirements and engineering reality we've been waiting for. Definitely worth your time to explore.