An application stack Claude coded directly in LLVM IR

Name: An application stack Claude coded directly in LLVM IR
Availability: InStock
Author: dboreham

by dboreham·Mar 12, 2026·10 points·0 comments

Visit Project View on HN

AI Analysis

●●●●GemZero to OneWizardryBig Brain

Writing working apps in LLVM IR for AI agents skips human-centric abstractions entirely.

Strengths

•Challenges assumption that high-level languages are necessary for AI code generation.
•Working demos prove agents manage semantic representations without human-readable syntax.
•Whitepaper formalizes the architecture instead of just dumping raw code snippets.

Weaknesses

•Purely experimental research artifact; no practical utility for human developers today.
•Debugging LLVM IR directly remains painful even if agents handle writing.

Post Description

This repo is the result of a debate about what kind of programming language might be appropriate if humans are no longer the primary authors. Initially the thought was "LLMs can just generate binaries directly" (this was before a more famous person had the same idea). But that on reflection seems like a bad approach because languages exist to capture program semantics that are elided by translation to machine code. The next step was to wonder if an existing "machine readable" program representation can be the target for LLM code generation. It turns out yes. This project is the result of asking Claude to create an application stack entirely coded in LLVM's intermediate representation language.