Why are other AI slopilots so far behind Cursor? All the competition such as Copilot takes the approach of just slapping new and newer models to the pile and hoping it'll solve the problem. But any AI model will inherently suck without good external sourcing. And none except Cursor provide that. Copilot has a @github call but it returns mere links that the model doesn't even visit later. Same goes for stuff like gemini code assist except gemini isn't grounded at all on the extensions. And it makes it fucking ass. No matter what model you use, the newest and hottest fad, it won't know the newest up to date syntax for the tool you use. It's the only thing in which it COULD outperform you and turn up useful. To quickly synthesize a solution from the Internet and apply it's internal knowledge to piece it together with the stuff you have. To look up syntax faster than you can and while you're slouching off reading some obscure doc, a external crawler has already fed him the necessary info through a json and is now integrating into your codebase. I don't need the latest Claude to do menial shit. But it will undeniably struggle because even if it is smarter than your average Pajeet, it needs to do the same thing as the rest of the fucking programmers on the globe did and still do: search the fucking Internet!!! All the other companies seem to just ditch the idea of making their extension anything other than an api wrapper with a clunky UI and just pour more money into this shit. Surprise that no one still uses it
Afaik cursor uses the exa.ai to deliver the external sourcing. I've used it as an MCP but models are still too stubborn and don't perform searches central to the problem but just tick off the requirement. Cursor works around it by probably having an external model analyze your query and formulate a query to the crawler that delivers the enhanced results to every reply, not on the whim of the model.

>>106019784 (OP)
Pajeet

>>106020609
jeets don't radiohead