We pushed out ~half a dozen improvements towards the end of the month, largely focused on quality, some on speed!
The apply model generates full file diffs for chat, composer, agent (eg here for chat). We’ve consistently heard that people really love it when it works. But that patching the long-tail of performance was really important.
We’re not done and want to keep improving both quality and speed a lot! Please let us know if you have feedback. If you run into a bad generation, cmd + shift + p > “report ai action” to grab the request id would help us investigate.
@truell20 Thanks for the update on the apply for composer, one issue I still commonly observe is that i struggles to apply changes made by the composer when the file size is large, say more than 1k line of codes, some times it starts to remove existing code in the file that is not relevant.
Regardless this has been a great experience so far, eager for next update.
While 0.43 still does it for me sometimes, 0.44 will either completely ignore the context and start redeclaring files that already exist, or it will just make a few tiny edits that don’t at all achieve what I was asking for.
Can you expose the apply model via an API? It would be super useful for other agents use cases that want to update code without repeating the whole file
I’m late, but you can get to the changelog via a link at the bottom of the main site (or by just going to https://www.cursor.com/changelog)
Dunno if the reply about it being in the about section of the help menu is referencing something that used to be a feature but it isn’t there in the version I have… though it definitely should be.
