Remix.run Logo
embedding-shape 4 days ago

> a two sentence prompt fixes it in a million line codebase in thirty minutes.

Could you please create a verifiable and reproducible example of this? In my experience, agents get slower the larger a repository is. Maybe I'm just very strict with my prompts, but while initial changes in a greenfield project might take 5-10 minutes for each change, unless you deeply care about the design and architecture, you'll reach 30 minute change cycles way before you reach a million lines of code.

fooker 4 days ago | parent [-]

Your observation was valid in 2025.

This is largely a solved problem now with better harnesses and 1M context windows.

embedding-shape 4 days ago | parent [-]

No, it's valid still in 2026, I'm literally using most agents on a day to day basis.

As mentioned, I'm happy to be proven otherwise, otherwise please stop just trying to say "It's not like that anymore" when supposedly it's so easy to prove. PoC||GTFO, as we used to say back in the day when they people shipping software actually knew how to write software.

fooker 4 days ago | parent [-]

> I'm happy to be proven otherwise

I am getting the impression that you'd be moving goalposts :)

I just checked out clang+llvm, 24 million lines of code, and implemented a simple C++ language extension with claude 4.6 opus high in less than five minutes. Complete with tests. Debugging and optimizations working seamlessly.

(It's a simple pattern matching implementation, if anyone's curious.)

embedding-shape 4 days ago | parent [-]

> I am getting the impression that you'd be moving goalposts :)

Nope, the original goal post you set was "implementing large changes are easy [...] a two sentence prompt fixes it in a million line codebase in thirty minutes", I'm more than happy for you to prove just that, no moving.

> I just checked out clang+llvm, 24 million lines of code, and implemented a simple C++ language extension with claude 4.6 opus high in less than five minutes. Complete with tests. Debugging and optimizations working seamlessly.

Awesome! Please share the steps to reproduce, results and the prompt?