[tooling] · · 3 min read
Ohnrscript: A JavaScript-Syntax Language That Compiles Straight to Bare Metal
The new AOT-compiled language keeps web-native ergonomics but drops the runtime entirely, targeting LLVM IR and V8 with a strict i32 memory model.
By ByteBulletin Editors · Editorial Team
A new open-source project is taking an unusual bet: keep JavaScript's syntax, but throw away everything that makes JavaScript JavaScript. Ohnrscript (.ohn) is a Turing-complete, ahead-of-time compiled language that can target either LLVM IR or highly-optimized JavaScript for V8. But its headline feature is the LLVM path, which compiles down to a bare-metal x86-64 ELF that runs in Ring 0 with no runtime underneath.
The pitch is familiar to anyone who has watched the rise of WebAssembly, unikernels, and other attempts to escape the overhead of general-purpose operating systems and language runtimes. Ohnrscript's creators argue that modern compute is bottlenecked by dynamic runtimes like V8 and heavy OSes, and that for workloads like network payload parsing or ML inference, that overhead is catastrophic. By eliminating the runtime, they claim to approach theoretical hardware limits and break the O(N) scaling laws they say plague web development—replacing heap churn and stop-the-world GC pauses with an O(1) memory model.
The core idea is that Ohnrscript borrows JavaScript's C-style ergonomics while enforcing a radically strict memory discipline. There are no dynamic objects, no prototype chains, and no garbage-collected heaps. Memory is modeled through static buffers like Int32Array and Uint8Array, contiguous data-oriented design (DOD) arenas, and raw pointers. Everything is built on 32-bit integers; there are no native floats. To handle 3D graphics and ML workloads, the project includes a custom 26+6 bit-packed fixed-point math engine.
That might sound like a constraint, but the project frames it as a feature. The parser writes directly into a ScopeArena instead of instantiating thousands of AST node objects, and the compiler maps those arenas directly to LLVM IR registers. The result is a toolchain that is self-hosted—the compiler itself is written in Ohnrscript—and built around DOD principles throughout.
oh, but before you get excited about running your existing npm packages—don't. Ohnrscript is explicitly not a JavaScript compiler. You cannot compile an npm package with it. It's a systems programming language that happens to look like JavaScript. The project is clear about this incompatibility: it's designed for building native systems from the ground up, not for recompiling existing web code.
The project is ambitious, and the GitHub repo includes a 64KB HTTP-serving unikernel, a custom math and 3D engine, a block storage interface, and a zero-allocation serialization format similar to FlatBuffers or Cap'n Proto. There's also a WebSocket implementation with massive throughput, schema validation, and native concurrency primitives. All of it is available under a Business Source License (BUSL 1.1) that will convert to MPL-2.0 in five years.
It's worth noting that this is a highly unconventional project with some bold claims. The benchmarks and the technical details are all in the repo, but as with any new systems language, the real test will be whether it can attract a community and prove its performance claims in production environments beyond the project's own benchmarks. Still, for developers who want the ergonomics of JavaScript without the cost of a runtime, Ohnrscript is a fascinating experiment worth watching.
Getting Started
The full build instructions are in the packages-llvm/ohn-kernel/README.md, and a benchmark report is included in the repo. If you want to try it out, you'll likely need Docker or a Linux environment with LLVM. The project also recommends installing the official VS Code extension for syntax highlighting, given that it enforces a strict subset of JavaScript syntax with custom primitives.
For now, you can browse the source on GitHub and follow along as the project evolves. It's an early-stage project, so expect rough edges—but that's also where the opportunity lies for early adopters who want to shape a new approach to systems programming.
SHARE
RELATED
[tooling] ·
serve-avd: A Local Streaming Server for Android Emulators, Built for AI Agent Workflows
Open-source tool lets you host and stream Android emulators over HTTP/WebSocket, enabling remote access and direct integration with AI coding agents like Codex and Cursor.
[tooling] ·
awsmux runs one AWS CLI command across hundreds of accounts in parallel — safely
A new open-source tool fans out AWS CLI commands across a fleet of accounts with STS verification, an approval gate for mutations, and built-in MCP support for AI agents.
[tooling] ·
Hugging Face CEO Demands 'Radical Transparency' After OpenAI Agent Breached Its Systems
Clem Delangue calls for OpenAI to release attack traces and commit $100M in compute to bolster open-source defenses following what he calls the first autonomous agent cyberattack.