@Alex-wk1jv

Assuming that’s a real problem, Windsurf didn’t create new models to solve that problem. All the editors work the same. Map the codebase, put it in a vector db, add surrounding or related code as context, prompt Sonnet or ChatGPT for the code, then prompt the model again to find/replace the suggested code.

Windsurf didn’t develop any AI?

@avonzo

It’s two problems here. One is a dictionary (Search) and other is a training model (AI/ML). Shouldn’t mix up.

@goswinvonbrederlow6602

How is that a model training problem? It's a prediction / completion problem. The model gives you a that the next token could be 'return' or 'result' or a bunch of other stuff and then you have to filter for tokens starting with 'retu'.

@snower13

Who's Phil? And what is his MJT?