kara

Core idea

Kāra is a systems programming language positioned for the age of AI written code. The design premise is that a developer declares intent and the compiler handles the parts LLMs tend to get wrong, specifically memory layout, ownership and concurrency, while emitting every decision as structured output that agents can consume. The compiler's analyses are not trapped in human oriented text; they are queryable as JSON. The README frames this as the whole point: rather than trusting inferred behavior, a user or an agent asks the compiler exactly what it decided and why.

Three distinguishing features

The language lists three things it does that others leave to the programmer. First, effects become automatic concurrency: functions declare what resources they touch, and the compiler runs independent work in parallel with no async, await, colored functions or thread plumbing. Second, ownership works without lifetime annotations; parameter modes are declared at the function signature and the rest is inferred. Third, every compiler decision, covering effects, ownership, concurrency and fixes, is queryable as JSON. The README shows a sequential looking program whose two data source reads the compiler proves independent and runs concurrently, then verifies with a karac query command.

Tooling and output

The karac query interface exposes effects, ownership, concurrency, cost summary, monomorphization, affected-by and attributes, letting tooling and agents inspect the compiler's reasoning per function or file. The language targets WebAssembly in addition to native compilation through LLVM, and the examples emphasize parallel data source reads and structured diagnostics with machine applicable fixes. The README invites design feedback through GitHub Discussions and points contributors to a CONTRIBUTING.md covering the build, the gates and the invariants worth knowing before patching.

Editorial conclusion

Kāra is Apache-2.0 licensed, implemented in Rust, and reported forty stars at time of indexing, with its homepage at karac.dev.