๐๐๐ป ๐ฅ๐ฒ๐๐ฟ๐ผ๐๐ฒ ๐๐๐๐ฒ๐น๐ณ ๐ช๐ถ๐๐ต ๐๐ ๐ถ๐ป ๐ต ๐๐ฎ๐๐
Bun just translated its entire codebase from Zig to Rust. An LLM wrote most of it. This process took 9 days. The pull request added over 1 million lines of Rust code.
Rewriting software is different from translating software.
A rewrite means humans rethink the architecture. They fix old mistakes. Translation means you push code through a machine. You hope the output works.
This creates massive risks for developers.
- LLMs match patterns but do not reason. They might use Zig patterns in Rust that break the ownership model.
- Edge cases often fail in translation. An LLM does not know about the bugs that took months to find in the original code.
- Nine days is not enough time to test a runtime. Production environments require deep testing.
I call this tech debt laundering.
You had a Zig codebase with known issues. Now you have a Rust codebase with unknown issues. The debt did not go away. It just moved to a new language.
The old code was battle-tested. The new code is battle-translated. Those are not the same thing.
Companies depend on Bun for their infrastructure. They are now running code that was mostly generated by a machine rather than engineered by humans.
I use LLMs every day. I use them to write small functions. I do not use them to translate entire systems that thousands of people rely on.
Speed is the problem. Nine days is enough to see if code compiles. It is not enough to see if humans understand the code.
Would you run LLM-translated code in your production environment? What would make you trust it?
Optional learning community: https://t.me/GyaanSetuAi