Where is adding past Chats!!! huge LOSS
Hey, see my reply on this here:
1 Like
I agree with the opinion that 2.0 has broken development productivity.
Iām not a vibe-coder myself; I adhere to an AI-driven development approach. That is, I have minimal control over the finished code, but I try to monitor architectural decisions and processes. So the new interface suited me just fine.
But all the other changes are a complete bummer:
- AGENTS.md seems to have been broken ā it now has a very low priority, and suddenly it went from mandatory to recommended, but in reality, it simply wasnāt used in 90% of situations. Even when I moved its contents into .cursorrules, it didnāt help much.
- The worktree feature almost drove me crazy. When using a single agent thread, I donāt understand why itās even necessary. And I spent a lot of time and energy trying to figure out the agentās phantom behavior.
- And the hallucinations ā there are more of them in automatic mode. Previously, the agent, if it didnāt complete the assigned task, would usually complete it in one additional iteration. Now, even 4 might not be enough. And itās not a context issue ā it might not even be half loaded and compressed.
2 Likes
I can no longer use the plugin to run unit tests in Java anymore. The 2.0 update broke it.
Also, I can no longer put summarized past chats in the agent context. This is sad.
Also again, Iām not able to disable this TODO feature.
Stop removing things that work and are useful. This does not make sense
3 Likes
Good afternoon, gentlemen.
Iāve been using Cursor for almost six months. Iām not a developer, but Iām learning. I led the development, building the architecture and logic, delving into the details, and selecting methods. I only checked the code superficially and simply monitored the models.
After update 2.0, itās impossible to work! The cursor agent doesnāt clear its cache automatically or forcefully, which caused the model to freeze during operation. After creating a new chat and transferring information from a previous chat, the model froze again, completely breaking everything that had previously workedā¦
I spent the entire day fixing it, to no avail.
This was very unexpected for me! Iām worried about the future of Cursor and my projectā¦ Will they have a future?
Even within a single chat, the model has a very low context threshold! Now the cursor works no better than a regular AI bot.
Agent? I donāt think so! BOT!
3 Likes
1 Like
Submit a bug report if you can. The developers do read and address them. Bugs can get lost in these unofficial mega threads.
2 Likes
@andrewh I just downgraded from 2.0. Removing allow-lists from auto-run non-sandboxed mode nuked my productivity.
2 Likes
Thanks @jrista for advocating here. I will not be upgrading until this is fixed unfortunately.
I think if you toggle this on, it will give you the Pre 2.0 allow-lists:
1 Like
Yeah, I donāt know how much more I can handle approving every single command. Its driving me crazy and has been a productivity killer. Worst change they have made since 1.7.
1 Like
@jrista The screenshot I shared just above shows how to go back to the old allow-list.
I turned off auto update for the IDE and downloaded the install files for 1.6 and 1.7 - reverting to earlier versions is the solution for me.
@jrista ā In Cursor: File ā Preferences ā Settings ā search āupdate modeā ā set to ānoneā
Everything was working great before v2 and I refuse to give up the excellence of prior functionality.
Too many Chefs have spoiled the soup. Wake me up when its over.
1 Like
Allowlist issues: The same that I reported here?: Dangerous change in last update! - #4 by RafeSacks - They added a (badly named) āLegacy Terminal Toolā option to roll back to the older version.
Aside from that, I honestly have been ok using 2.0 in the normal Editor mode. Browser got better for front-end work, which is really great. Is there a central list of things that are no longer there? There is a very good chance I just havenāt reached for one I will miss yet.
Read sticky note on sandbox vs allowlist
The sandbox is completely broken at the moment. Iāve seen the agent pipe dynamic commands into a python REPL. Iām now approving EVERY command manually (which starts to doubt whether Cursor is still the tool I want to to with, to be honest). 2.0 has been a huge disappointment and productivity disruptorā¦
My experience has been even worseā¦ Itās not just the removal of features I relied on! Iām experiencing frequent crashes, and most critically Iāve had instances where it has **actually corrupted my chat and code, doing things it is not supposed to do. **
The impact on my productivity has been devastating. Iām not exaggerating when I say that tasks which previously took 20 minutes now frequently escalate into 2-to-4-hour debugging sessions, simply to manage the toolās failures or fix what it broke ā¦. WTF is that!?
1 Like
Sorry to continue bashing on this, but what actually makes the situation with the Sandbox even worse is that the latest update introduced a little overlay with the edited files that (for me) often blocks the approve or reject button of the tool. More than once I stopped the generation because I wasnāt able to click the āRunā button.
6 Likes
I actually just found your original post. Many thanks for finding a solution.
Could you please clearly let us know when these issues will be resolved? Itās really frustrating that we keep encountering major bugs almost every day. Weāre paying for the $200 plan, and todayās new problem is that the thread stops working in the middle of a task and becomes unresponsive, forcing us to revert everything and start over!