I 100% agree all you said.
I like the feature. But MCP is going to DESTROY cursor if frontend is the company direction
This is starting to become a pattern and I just want to provide useful feedback. First off cursor is a great product and I use it daily.
My main issue with this 2.0 release is the amount of regressions and complete layout changes that are starting to trend towards the lovable/bolt beginner getting started crowd.
Im all for making things more accessible but the orginal editor layout is becoming unstable with a lot of minor UI regressions. Simple things that used to work now broken or buggy.
Its honestly too many to list and at this point is a bit frustrating because the Agents mode is also really buggy and inconsistent.
Guys its just too much. Cursor is starting to become so bloated and the UI changes happen so often that my day to day workflow is constantly taking a hit.
Stabilize the software. Stay consistent with UI. I do like the additional browser capability. Everything else I could easily do without and it feels like a step back tbh.
Make a choice between serving professionals or newbies. Doing both is becoming a headache.
This is a really common pattern for modern product teams it seems.
I was an avid bubble user and left because they would completely change a core function so much that it was unusable and broken all over - all the name of “progress” to some greater goal that after years is still “coming next release”.
I was in weweb and xano for a while and the exact same thing happened there - major breaking changes pushed with seemingly no internal QA or sanity checks (although not nearly as bad as bubble which takes the cake!)
It’s like these product teams want to redesign their entire product every release. It’s insanity.
Cursor thankfully is a very simple UI and the AI agent can fairly reliably edit or instruct me on how to edit everything I need more or less and most of the new features are easily ignored - if you guys really want to see a mess take a look at bubble. What a dumpster fire that has been for the last 5 years. Bloated. Buggy. Nightmare.
Maybe some of those bubble engineers and product people are working at cursor now. God I hope not.
So far I’ve only been around for 1 major cursor release and whilst I think it was largely a flop with mostly useless features it at least didn’t hose the existing features so badly to render them unusable. I can largely just ignore the release and new features.
I can no longer @ mention a git commit for the Agent to analyze. I have to say “look at commit abc123” and then the Agent has to actually run the command git show abc123 in order to look at it, and the command fails because it’s in the wrong folder, so it runs cd first, but the command fails because The token '&&' is not a valid statement separator in this version. so it separates them with semicolon and finally the command runs and it sees the raw output of the command, etc. This is such a downgrade and waste of tokens.
VSCode with Copilot is now more powerful than Cursor?
This has been broken on and off since 1.5 at least. I made a post about it and would love if Cursor would let us know what is going on
I’m referring to @git, not GitHub
How to revert back to pre-2.0 Cursor:
- Disable automatic updates by going to “Open VS Code Settings” and then searching for
updatethen going to Update: Mode and selecting “none” from the drop-down. - Download Cursor 1.7 from Download · Cursor and re-install it. (The “Windows (x64) (User)” version matches the way it’s currently installed for me, under
%localAppData%\Programs\cursor\Cursor.exe.)
Cursor needs to fork V2 into another product, and return the lost functionality in the V1.7 later releases. And fire the product development Director
I don’t see the heavy weight corparate customers being amused with a dumbed down IDE catering to beginners.
Very sad to see such a usefull programming tool going down the drain so quickly
Ok, how do I clear the worktrees? There’s like 30 folders, I couldn’t find any option in Cursor to release the agent threads.
What’s the workflow when you go from cloud agents to local? Presume you have to merge these or something. Got two different things going on.
Version: 2.0.54
We were able to undo changes in a worktree to review a parallel agent’s changes. This ability suddenly disappeared in this patch. Is this a bug?
After experiencing the 2.0 UI, I’ve noticed that the biggest difference between the Agent Layout and Editor Layout is the “Maximize Chat Size” feature—something that can actually be achieved by upgrading VSCode. So, what’s the point of implementing this functionality using an older version of VSCode?
Upgrading VS Code unlocks many convenient features and fixes numerous legacy issues.
I’ve noticed that all the updates since last week have broken the normal functioning of agents — they’ve simply started ignoring the rules. I tested it on several models, and the result is the same everywhere.
Is it just me having these issues, or is anyone else experiencing the same thing?
Version: 2.0.54 (Darwin)
It appears you need to drag files into your base project and then run a Local agent. There is no merging yet, and the Cursor team appears silent about this. We get it, multiple agents, multiple moneys, but if we waste the forks, it’s too expensive.
I agree. It’s important to be ahead of competitors, and surely agents will be useful at some point, but why make the default mode agentic when they are still incapable of writing production code?
Cursor Generating rules
What are the alternative methods?
Today, I couldn't load the latest chat on my work machine—it was stuck on "Loading chat" forever.
I restarted Cursor, and it greeted me with a welcome screen, as if I'd just launched it for the first time.
Inconvenient, but okay.
But where did the voice input gone?
Same, also no history retrieval with 2.0.69.
Can switch between new chats - format has changed or sth.
I would so puke if I had an unfinished thread from yesterday..
Same in 2.0.74 - WTF guys.

