FAQ

FAQ

Is Nirdosha production-ready? Not yet. It's an active research prototype. Many safety properties are proven today; others are explicitly marked aspirational — see Honest Scope & Roadmap and ROADMAP.md for the full status.

Why not just use Rust? See Nirdosha vs. Rust, Go, Mojo. Short version: Rust already solves memory safety and concurrency for teams that can invest in its learning curve. Nirdosha targets a narrower, different problem — AI agents writing backend code where the grammar itself has to make invalid syntax impossible to emit, not just unlikely. If Rust already works for you, keep using it.

What compiles today vs. what doesn't run at all right now? The tree-walking interpreter was removed entirely (2026-09) — there's no interpreted fallback left, so this is no longer "compiled vs. interpreted," it's "compiled vs. not currently running in any form." Compiled: numerics, box/&/*/froze, str, tcp/file, Vector/Matrix, non-affine struct/enum/match, deterministic RNG, spawn/thread/chan (no mutex in the language at all), identity (check_role/field masking/requires+acquire), nfr(...). Not currently running in any form: sandbox, db/json/http/mq, transact, workflow, and the live (server-backed) half of the UI engine. A program using one of the second group is rejected at compile time with a named error, not silently mis-compiled. Full list in Honest Scope & Roadmap — PUBLIC_ROADMAP.md for what's landed most recently, since this list moves fast.

How do I report a bug? Run the command that failed and paste the exact nirdosha build/ emit-llvm/emit-ui error text into a GitHub issue — there's no --format=json flag today (a real, disclosed gap; see LLM Integration), so plain error output is what to paste. If it's a security issue (a type-checker/ownership soundness hole, an auth bypass, anything that breaks a safety guarantee this project claims), see SECURITY.md instead — report it privately, not as a public issue.

How can I contribute? See CONTRIBUTING.md. Docs, examples, and .nir test cases are the fastest way to start.

Where's the roadmap? PUBLIC_ROADMAP.md for the scannable version; ROADMAP.md for the full internal tracker with verification detail.

What's the license? MIT — see LICENSE.

I'm an LLM being asked to write Nirdosha code — where do I start? Read LLM Integration first, then use agent-skills/nirdosha/paste-anywhere-prompt.md if you don't have file access to the repo's agent-skills/nirdosha/ directory directly. It's the same content verified against the real compiler, packaged for whichever agentic tool you're running inside of.

Does sponsoring this project do anything concrete? Yes — see the Sponsor badge in the README. It funds continued work on exactly the open follow-ups listed in Honest Scope & Roadmap.