Author here: I wrote this peice after trying to use AI to build an internal tool and running into a failure mode I didn't expect.
Curious if others have his the same "expansion without convergence" issue with using LLMs for coding.
dbinky•1h ago
Have you tried using Ralph Wiggum Loops to reach convergence? I use the Ralph-O-Matic on Github to run automated, iterative refinement loops on my codebases. It works very well!
vanyaland•1h ago
agree, I also used Ralph loop /umputun/ralphex on GitHub
aurareturn•1h ago
How has Ralph loop helped you?
vanyaland•1h ago
The only thing that’s helped me is adding hard constraints: spec, architecture, small verifiable steps, and explicit decision logs.
wespiser_2018•1h ago
Me too! I'm really interested in seeing an LLM coding platform treat those documents (spec, architecture, decision logs) as first-class objects: collaborative, persistent, and outside the codebase itself.
Right now they live in repos and it's easy for them to get lost later, and harder to share them.
wespiser_2018•1h ago
Curious if others have his the same "expansion without convergence" issue with using LLMs for coding.