Another feature release that target a select customer base while crippling bugs are being ignored or made worse. For the past month or so, this terminal crashing bug was an issue maybe every 3 or 4 chats. Now it happens after every 2 or 3 calls within a chat. I am having to restart Cursor every 10 minutes or less now. Wasting my time. Wasting my tokens. Wasting my money.
What is going on? Is AI in charge of your releases now? Is anyone actually handling quality control?
I like Cursor but lately is really hard even to recomend it for someone. Every feature has a bug (or even more than one)
Vibe hiring 
They have been pretty open about their devs using Cursor to develop Cursor.
I think there is a non-negligible possibility that the backbone of Cursor is quite spaghetti, as in there is no one that is quite sure how everything actually works or interacts.
I wouldn’t be at all surprised if they straight up doesn’t have QA as part of some kind of vague design philosophy from the top akin to
“Issues are resolved as they appear, leaving more room for our developers to create rather than fix. Something something fluid design pattern, something something evolutionary production flow, AI revolution, Agentic, efficiency cascade” → Investors fighting for their life to throw their money at them.
They simply fix issues when they have users complaining about it, if no one complains there is no issue, so why QA test when you can just ship it and fix what doesn’t work? 
There is one article about their hiring strategy. Between the lines you can read that it’s creative chaos of bright minds, and no QA at all.
With 2.0 there was a brief moment of concerted direction, which was great.
Now it seems “everything basic is solved, let’s brainstorm to somehow get an advantage again”.
I predict some multimodal canvas-like input experience soon, and Composer-2 in a month (new chinese models).
I think they neglect the harness too much, the results are pretty bad compared to original harness. Droid and Pi show that it can be done better.