[▲ Vercel Community](/) · [Categories](/categories) · [Latest](/latest) · [Top](/top) · [Live](/live) [v0](/c/v0/59) # Bug0: the v0 team is going on a bug fixing speedrun 654 views · 56 likes · 36 posts Pauline P. Narvas (@pawlean) · 2025-08-19 · ♥ 4 If you could fix ONE thing in v0, what would it be? This week's focus is fixing bugs: reply here with your idea and the v0 team will ship as many of them as possible this week Tzuyi 1494 (@tzuyiyu-1494) · 2025-08-19 · ♥ 10 Hi, REVERT TO THE PREVIOUS MODEL. Ozce2300 7307 (@ozce2300-7307) · 2025-08-19 · ♥ 7 Remove the V0-agent, i will not use V0 anymore with this agent. Pierrunoyt (@pierrunoyt) · 2025-08-19 · ♥ 1 Im really sad that it can’t make project structure like new folders etc. It doesnt work. Now it think it createed the folder and the image is there.  Ash (@crossmedia) · 2025-08-19 · ♥ 2 Agent is really problematic. I understand some things take time to settle, but after hundreds of prompts, I’ve lost almost all faith in this model. I could say a lot, but the biggest issue is that it simply doesn’t do what you want it to do. It often thinks it has resolved a bug that it has not. The “hallucinations” like this, if we can call them that, are regular. In fact, they are common enough to prompt me (pun intended) to move on from v0 even though I’ve loved the platform. I do not know what is going on ‘behind the scenes’ but the agent just fails to do what the other models seemed capable of doing. I’d love to know why the agent is seemingly so ‘dumb’ compared to the other models. After a number of personal projects under the old models, I got into the rhythm of those and I knew their foibles and weaknesses. Agent just outright refuses to follow a rhythm. It is extraordinarily over-confident in itself. Ash (@crossmedia) · 2025-08-19 · ♥ 1 This, this, a thousand times this. Agent is arrogant. It regularly thinks it has done something, fixed something, moved something, and it has done no such thing. You ask it to try again, it tells you it has found the root cause and fixed it, but it hasn’t. Prompt after prompt. Agent is expensive! Contato 4630 (@contato-4630) · 2025-08-19 · ♥ 3 Can the agent mode be considered a bug? Because so far it doesn't even create a landing page :joy: Bosszeus (@bosszeus) · 2025-08-19 · ♥ 1 Agent mode has wasted all my monthly credits in about 15 prompts trying to fix the same error I started with. I gave it the correct code to implement as well and it still didn’t fix the issue. It has become a money sink at this point with no issues resolved… At this point I’m probably going to cancel my subscription and start using codex. Codeshode (@codeshode) · 2025-08-19 · ♥ 4 * Remove v0 agent mode * Allow ability to select model * if u want to test features, give free credits. and dont make people pay for it. it’s ridiculous. Jonathan 7545 (@jonathan-7545) · 2025-08-19 I have tried to build a small app twice. In both cases, I have not been able to get even simple authentication to work. v0 keeps trying to fix the problems but nothing works. I am abandoning my project. qaaq (@dkc1011) · 2025-08-19 Agent mode is buggy fix it we also want to be able to use v0-sm not only the md expenseive model lukek (@lukek) · 2025-08-20 · ♥ 2 Firstly, I think Agentic AI has been fine since the big launch week ended. Secondly, if I encounter a bug usually I’d report it, but a small bug you guys could fix is i can’t seem to remove images once I’ve clicked on them and it’s in the message box (on my IPhone at least). Bug: Users on iPhone can’t remove images once they’ve clicked them and they’re loaded in the message box. If I’m missing how to let me know, but yeah. Alwldk 9079 (@alwldk-9079) · 2025-08-20 Error correction consumes a lot of credits. At least for prosaic users, I don't think there should be any credit consumption for error correction. It's too costly. Second, you need a rollback function or a data refresh function. You need to return to the previous state when you're building and it's not the result you want. To restore the state you want, you can't guarantee that you'll consume credits through queries or how much you'll use up the result you want. The third database was also not connected, but there is a part that consumes credits by continuously correcting errors, saying that I will find a connection method. If the program cannot be implemented, I think we should clearly ask the user to solve the problem. Ash (@crossmedia) · 2025-08-20 I’ve canceled. I’ll keep an eye on the forum and if things get better, I might join up again. You cannot built apps on shifting sands. Jamie (@jamieludlow) · 2025-08-20 The ‘Open in a new tab’ now doesn’t open in a new tab. Instead it is just enlarging the preview in full screen mode so when I click on links that are supposed to open in a new tab they don’t work. Please revert back to the previous open in new tab that worked perfectly and didn’t remain in preview mode. Georgeaffield (@georgeaffield) · 2025-08-20 · ♥ 4 Revert back to the old version!!! NO AGENT or at least make it on option to use, not enforce people to use it because its simply not good enough. The outputs are worse than before, not just a little bit worse, I am talking going back to chat gpt 3 versions and taking a lot of my well earned dollars in the process. I am acutely gutted, today I started using lovable which I am not happy about but the outputs on there are consistently better than v0 now. I get the whole agent hype, I do, but it feels like you guys are shifting to try and go after a more mainstream market, and, I don’t blame you, lot of money in it, but this recent agent mode feels like you are leaving your dev communtiy behind to try and seek a more mainstream crowd of people who want to give you an instruction like “build a web app about XYZ” I’d be happy to get behind the agent mode, I am sure in time it can become something, but I just can’t understand how people all sat around and thought this agent mode was acceptable to release, while also deciding to remove the ld version Ross (@ross-7453) · 2025-08-20 · ♥ 4 Bring back the option to choose a model, bring back GPT-5. Remove agent mode. V0 was amazing for the 24 hours it had gpt-5, now it’s unusable. I can’t even open it now, and I’m afraid to keep working on old projects because I know it will very likely just break them. Ash (@crossmedia) · 2025-08-20 · ♥ 2 Yeh, I am also afraid of opening old or ongoing projects since Agent was introduced. It doesn’t understand my apps like the older models did. Tommy Jensen (@tommyjensen-6575) · 2025-08-21 I discovered v0 last week and have been impressed with the results I get. Sometimes it require multiple times to fix an error and in my last test I had to instruct v0 to rethink/rebuild the solution and then the error was fixed. I have so far built a couple projects with working authentication and Firebase Firestore database connected. Something I have not had succes with using Firebase Studio. But one thing that did bother me a little was if I pasted a long prompt the pasted text was converted to a text file which meant that the project name and thus the url was set to something that containted the words “image-analysis”. Not remotely relatable to the project itself. And I did not find any way to change that. I noticed though if I reduved the number of lines I pasted at a time it would not convert to a text file. I like to write long prompts in an editor before I submit it to v0. Lucasnoguueira (@lucasnoguueira) · 2025-08-21 removal of the agent, and return of the model selection, this new agent is horrible, it does not execute the prompts correctly and also spends credits incorrectly, causing a new expenditure of credits to correct its error dimzachar (@dimzachar) · 2025-08-21 if you paste a long message, then it shows up as a txt meaning you can’t use enhance prompt button Vani (@rahulgautama97-2599) · 2025-08-21 · ♥ 1 I’m out of V0 for production level projects. Migrating the rest as well slowly. I only do prototyping now cuz v0 has good UI THERE IS ONLY 1 ISSUE THAT V0 HAS RIGHT NOW (apart from the Agent Mode which was horrendous and could be the start of V0’s downfall) => The customer support is terrible at V0. The Staff do not take our feedback seriously (or maybe you do but I don’t see any immediate responses until weeks in which case you have a huge problem with your priorities and your leadership) Vani (@rahulgautama97-2599) · 2025-08-21 · ♥ 4 1. Bring back V0 2. Refund/Gift credits that were wasted on Agent Mode 3. Do not ship experiemental features to paid users ever again Team1grin (@team1grin-7008) · 2025-08-21 · ♥ 1 The entire agent feature is a bug. It is terrible, burns credits for bad output. Your previous product was fantastic. Please rollback. Alexcumbers 9492 (@alexcumbers-9492) · 2025-08-21 The user could pay for the first bug fix, then V0 should cover any follow up bug fixes. pavannextdoor (@pavannextdoor) · 2025-08-22 · ♥ 1 The new agent just simply isnt as good as the previous versions. I used to set the model to large and it would crush 2-3 tasks in one shot. Now I have to spend 20-30 prompts on simple things. It touches code that doesnt need to be edited. It forgets to include code blocks from other sections. Getting a lot of compiling errors which never happened before. It’s at the point where v0 is almost unusable. I’ll give a week or two to settle in. But after that, I’m going to have to find another solution as I’ve already launched my app to users. I will pay more to just go back to the large model. Had to make a support account just to post this cause everything worked so perfectly before. SoumyaK4 (@soumyak4) · 2025-08-22 · ♥ 2 Agent keeps making the same mistakes over and over again. If you’re trying to test stuff, don’t make people pay for it. Sasha (@chuponsito) · 2025-08-23 Remove the agent, which shows me that it is doing a lot of impressive things when it is not. It is useless. Return to the previous models that just take the money, do what the prompt asks, and that’s it. It’s very simple. riza (@hasriza) · 2025-08-23 https://community.vercel.com/t/few-bugs-to-consider-and-how-to-reproduce-these-bugs/18730?u=hasriza The diff issues still exist and please disable the auto compile on save in editor. Felicedev (@felicedev) · 2025-08-23 New agent mode is not able to create a working authentication system using Syupabase. Create mocked authentication, then i asked to fix and don’t save in database and supabase authentication. P0RT4L (@portal-2371) · 2025-08-23 Please revert back to the other model! this bug or something is cauing me to open my fork on the project, and show NO preview option or anything. i also keep getting the same error over and over: “Task stoppedThe task was stopped because the agent encountered an unrecoverable error.” Felicedev (@felicedev) · 2025-08-23 When start a new project, with around 7 steps, if the computer goes to standby the execution of the chat paused and cannot be resumed, also send a message “continue” is useless. Jonoki67 4086 (@jonoki67-4086) · 2025-08-24 Sometimes, I’d give v0 some of the most simplest tasks, but it would just completely undo or redo everything we’ve built till now. And then, I’d end up wasting my credits trying to solve the new problem or restore the older version so I can start over. For the last few days, even if you restore an older version in v0, it won’t restore exactly the same as it was before, it’ll make some new changes as well. 1 out of every 3 times, it causes problems. It’s hell. I’m beginning to think that I’m wasting more money than I’m getting utility out of it. Misval (@misval) · 2025-08-24 · ♥ 1 [quote="ross-7453, post:19, topic:20069"] Bring back the option to choose a model, bring back GPT-5. Remove agent mode. V0 was amazing for the 24 hours it had gpt-5, now it’s unusable. I can’t even open it now, and I’m afraid to keep working on old projects because I know it will very likely just break them. [/quote] PLISSSSSSSSSSSSSSSSSSS houdini (@haroldfranklinpayne) · 2025-08-25 · ♥ 1 Allow us to select the old model. Not only is the new model simply bad - but it also sometimes gets in an endless loop when I ask it something simple (Like changing copy) and it goes into a loop where it asks itself multiple times and charges me for multiple prompts as it loops. I paid a dollar worth of credits just to change the hero copy on a single element. I use this because its sometimes faster than changing a bunch of copy myself. It’s no longer faster than just doing it manually. Pauline P. Narvas (@pawlean) · 2025-09-04 I'm closing this thread as bug0 has come to a close. Thank you all for your feedback! https://x.com/v0/status/1959045140431216684
