[▲ Vercel Community](/) · [Categories](/categories) · [Latest](/latest) · [Top](/top) · [Live](/live) [v0](/c/v0/59) # v0 auto-diagnosis automatically editing code and creating new versions without consent 190 views · 74 likes · 63 posts DevGlyphX (@devglyphx-6325) · 2026-03-18 · ♥ 2 ## Problem Can someone please help, who ever decided to push a new release with `auto-diagnosis` needs to re-think this. ## Current Behavior I have probably 200 apps in my favorites list, I open one up for a client today to work on and all of a sudden it just starts writing edits and breaks the entire Fricken application. It shot out an entirely NEW VERSION, WTF IS UP WITH THAT? I open another client app, and same thing, as soon as I open it it starts running and changing all my code and BROKE the APP… * I have no idea what was done, no way of checking, NO WAY of turning this AUTO-BS off. * I now have 198 other apps to go check on as a company and am going to opne each one and have them do some bullshit that I didn’t ask to have done, I didnt command to be done, nor did I APPROVE to be done. * Its eating credits, its creating versions for no reason and any type of organization that a larger company like myself has, its completely GONE. Even if you ask it “hey, can you do a quick scan and see if I have any vulonerabilities.” Cool, 2 days ago it would scan tell you what it found and Id go fix it. Now I cant even do that because as soon as it done telling me about it, its just “AUTO-DIAGNOSIS” and FIXES stuff that I haven’t told it to do. I can pull a git version back in, but as soon as it comes back in, the auto-diag starts all over again, so its a big loop. We are a team of 6-8 depending on projects, we have everything split between a few accounts, we can not take a chance checking in on anything that we may have local here in Vercel because of this Auto-Diag. This is unreal! DevGlyphX (@devglyphx-6325) · 2026-03-18 · ♥ 1 And then here is the best part: Go ahead and roll it back to the previous version, as soon as it gets done mounting, the auto-diagnosis kicks back in, changes what I had again, and then pushes out another “Fixed” Version. Did anyone ever think that a developer may have scripts running that doesn’t sit pretty in a serverless environment, but works fine when they actually put it on their own server? Im sorry dont mean to rant, but this is ABSOLUTELY BONKERS… Not even a warning pop up with the option to have it auto-fix or keep it as it was. We are afraid to check any of other other accounts or pull any code back in to work on right now and have gone back to just working locally in our VS, which is fine I guess, but man this is an absolute shit show with this auto-diagnostics just firing.. DevGlyphX (@devglyphx-6325) · 2026-03-18 · ♥ 1 Another bug: I opened up a new chat, just to see how this is working, I ask it to set my up an application with a specific database structure, it thinks, then gives me a few options to select or to tell it something else, COOL! Well you cant select anything because the “AUTO-DIAG” is running and checking the application, and ZERO has been built yet. So I have to wait for that, then it creates me a new version, of NOTHING, and then I can finally select my option. Vercel starts building etc. etc. Its a loop! DevGlyphX (@devglyphx-6325) · 2026-03-18 · ♥ 1 went from working production applications to ones that now cant even publish or show a preview. THIS IS UNREAL ! DevGlyphX (@devglyphx-6325) · 2026-03-18 · ♥ 1  Another example, this app hasn’t been touched in 24 hours. When we left it and pushed the update, there are zero errors in a production environment, this one has to do with Lint. So we dont mess with those errors because sometimes it can break the working application. Well as you can see 2 hours ago, it updated the version and is waiting for me to push it. If I try to roll it back to use the version that “I WANTED” it to use, guess what, it makes it worse, and it goes back into diag mode again and starts fixing it again. Considering we have quite a few applications like this, were stuck. the only options we have right now is to pull the code down from git into our VS and work on them or pull them into another IDE. I cant not believe there hasn’t been a single response on any of this or that we are the only ones experience this? DevGlyphX (@devglyphx-6325) · 2026-03-18 · ♥ 1 And you have no idea which version you have published anymore, its gone? So one of the three versions from 24 hours ago is the right one, doesn’t neccesarily mean its the last one, could of been multiple people working on it and had one published and didn’t publish the new one yet. I have a few other apps that show me multiple PUBLISHED versions, so again who knows which one is the actual published version. Anyone have any ideas? Even VO is telling me to turn it off in the right panel, but there is no right panel to turn it off?  DevGlyphX (@devglyphx-6325) · 2026-03-18 · ♥ 1 So then it knows im pissed, it asks me to choose, I chose A and its start doing the auto-diag again. I have one specific app that hasn’t been deployed to git yet, now I cant because it will only deploy the “New Versions” that the auto-Fix have created. Unless I know or wrote the code, why would I want to publish a version that I didn’t make or ask for changes to be made. Now im stuck with an app that I cant publish the right version, if you pubish the one vo created and then restore the one you want, the auto-fix starts again and puts ya right back in the crap situation you were in to begin with. 30 years of coding, ive never been in a mess like this before. MarnixB (@marnix-3108) · 2026-03-18 · ♥ 1 I totally understand where you are coming from, its such a weird thing that it just starts changing things without telling it to do, even worse that its really unclear which version is the auto version etc etc. this should deff be a opt in or opt out option. I can understand that it can be a nice feature for a beginning vibe coder. But if you have more experience this is just something really ennoying and credits wasting. MarnixB (@marnix-3108) · 2026-03-18 · ♥ 1 For every prompt that i send it also finishes with the auto-diag check, and with this 1 project i get this;  Its basicly saying in dutch”this is only a deprecation warning, not a crash/error - de app just works normally, no action needed” and its not even showing the warning that its talking about so i dont even know what the context is. @anshumanb is there any way to opt out for this? Last time i also had a project that just broke down, same problem @devglyphx-6325 is facing. Ph4900 (@phenssen-4900) · 2026-03-18 · ♥ 1 This problem is driving me insane. No matter what my prompt is, auto-diagnosis takes over → changing, reverting, changing, reverting, creating new versions without actually properly adressing my prompt. It feels like my only option now is to wait it out until this auto-diagnosis is removed, because auto-diagnosis keeps on taking over the conversation and creating new versions. Anshuman Bhardwaj (@anshumanb) · 2026-03-18 · ♥ 4 Hi everyone, I'm sorry that this is causing trouble. Let me take this feedback to our team. Anshuman Bhardwaj (@anshumanb) · 2026-03-18 · ♥ 3 @marnix-3108 and @devglyphx-6325 could you share the chat URL for the ones where it affected the most? MarnixB (@marnix-3108) · 2026-03-18 · ♥ 1 Hi Anshuman, thanks for letting the team check it, i need to see if i can find back the project, deleted some projects yesterday en this morning because of the other bug with the sandbox. Will see if i can find it. But besided that, it just feels really bad that a ai takes random control of your project, that just shouldnt be happening. We all love AI, but we as humans should always be in control, epecially with custom projects with special routes where you maybe dont really follow best practise, but it only works in that way. Anyways you get my point i think. It should be a on or off button/switch in the project or chat settings. MarnixB (@marnix-3108) · 2026-03-18 I didnt find the old project that got ruined, but i have a other sample in my current project; https://v0.app/chat/creditchecken-budget-tracking-app-bRxDfQbnUI8 (i am working in it right now) The last 3 or 4 prompts it started up the auto-dia and everytime it comes with the same results that it “fixed” something, but with the next prompt it just states the same again, its a mess. and it also just burns my tokens with every single prompt that i send. DevGlyphX (@devglyphx-6325) · 2026-03-18 · ♥ 3 Good morning @anshumanb , thank you so much for looking into this. I am currently on the Road, I’ll be back at a desk in about 2 hours and I can send you whatever your need. I stopped opening projects and told my team to only use VS for the time being. I dont have one specific link, its every one of them. The issue is, that its trying to correct errros that someone like myself, already knows exists, but also knows that they dont exist in a production enviromnet, especially if you are doing any web3 work. So as an example (Sorry at light trying to hurry), any app that we had built that utilized a Meta Mask, or Blade, or anytype of web3 wallet, as soon as you would open one of the chats, it would start to auto diag. Instead of just removing the lint, which doesn’t need to be moved, it started make changes to wallet files and everything else, trying to “find” all instances. Up till 3 days ago, this never did this, so it has to be something from a recent push. Like I stated above, its a big loop! - Login in open chat - Chat immediately starts to diagnose and create a new version - You try to restore back to the previous version, it does it, and then the auto-diag starts again. - This time making the same changes, some different than the first, and putting me back in the loop. My team and I are literally at a stand still, not because we dont know what were doing, but because we’re afraid to open any of the other projects / chats we have across all of our accounts. So any info, any update info ANYTHING, would be extremely appreciated! Anshuman Bhardwaj (@anshumanb) · 2026-03-18 · ♥ 2 Thanks both of you for writing this feedback. I've forwarded to our team and will relay back what the solution is for this. Sebgency (@sebgency) · 2026-03-18 · ♥ 1 This is happening to me too, this is very frustrating, I have tons of work to deliver and I did 30 versions of the page fixing things I didnt asked to change, this is taking me credits out, time out and im getting behind with deliveries for my clients, also, I give it a very specific prompt to it and the app does one or two things out of that prompt, this is a huge error. And it’s generating us a lot of trouble. Jerry 2306 (@jerry-2306) · 2026-03-19 · ♥ 1 I am seeing the same thing, wasting credits and breaking code. Restoring sometimes failed breaking code beyond repair. Anshuman Bhardwaj (@anshumanb) · 2026-03-19 · ♥ 2 Hi @jerry-2306, is it happening on old chats or only new chats and messages? @devglyphx-6325 @marnix-3108 our team has pushed a change yesterday so it doesn't happen for existing chats or old versions. Could you share chat URLs for us to further debug this? DevGlyphX (@devglyphx-6325) · 2026-03-19 · ♥ 1 Ok try to follow me on this: I opened up an test app that I was messing with from a few days ago, someting I dont care about. When it opened it showed, No errors , like it diaged or something. Then I started to ask it a small question and told it no changes, it spit out an answer and as it spit it out, it waited for a preview, diagnose it, no problems. (That is the problem). Then It asks me if I want it to go ahead and make it full color as you can see from the images. I respond with sure. Code gets done, you can see it starts to diag again, then says no errors. (Question is if there are no errors or there just not being auto-fixed). Anyway… the last image, you can see that the globe, which should be multi-colored, didn’t change at all. Any other version over the last year, would of changed the color, updated the preview and been ready to publish. There are a number of issues happening with Vercel that haven’t been present in previous versions. Another example, we have a client that needs an HTML5 game turned into a blockchain app, normally this takes me 30 minutes to get situated enough to pull off stick in VS and finish it. Tried to test tonight to see if it was any better, it ran for almost 12 minutes and then spit me out an app as a white page, and then got in a loop trying to figure it out, all while eating credits. Heres another one: an employee today, went in and tried to have it write a simple login page utilizing hedera. I build quite a few Hedera applications, and normally I can come to a chat and Say use walletconnect with connector and use it for authentication or something simple, just to get the base quickly written. Normally It may take me 2 or 3 max, chats to get everything nice and pretty; tonight, same as above, same prompts (we save prompts), and it started to build it completely wrong, wouldn’t install the packages, couldn’t publish it, and every time all it did was keep trying to auto-diag and it kept making it worse. Something is severely off to the point we may have to just call it day and move on to another platform, there are hundreds of chats between us, and not just the auto-diag and fix stuff, it just seems like it got dumb over the past 3-4 days, so much so its quicker for us to open VS and just grab some code from co-pilot as we go. I know that sounds crazy, but Co-pilot been giving us better results than V0 ever since Ive been posting about the issues. I know its not a claude or sonnet problem because i have clusters at my office and there are running fine. Same command, same results. Suggestion: Maybe the devs should give users the option to upgrade to newer versions when they are released or at least send out an email to everyone letting them know that updates are being pushed and performance may be degraded. (Try a prompt on a Saturday vs a Prompt on a Tuesday for example) Anyways, it just really sucks and I’m starting see that I’m not the only one that is being affected by these changes. I do appreciate your efforst! id give you a link but i cant publish it. Oh also, the errors are stacking up, especially around Turbopack. Ive haven’t seen that error once in the past year, now it seems like every app I have on here has it and is causing stuff not to publish.     DevGlyphX (@devglyphx-6325) · 2026-03-19 · ♥ 1 Just got it to do it and was able to catch it: At the end of the first paragraph, the app was done building, then all of a sudden it just starts re-writing files that it said was already good. There were no errors keeping me from publishing, it should of stopped and let me published. But instead it just kept running for over 5 minutes, creating two versions, that it should of never created.  After I posted the above picutre, i went back to vercel to see what was up, and you can see the amount of changes its still doing         DevGlyphX (@devglyphx-6325) · 2026-03-19 · ♥ 1 Then from the post above, it told me it was done right, I went back to V0 and look at this crap, its still running, it started all over again, all while still eating credits.   DevGlyphX (@devglyphx-6325) · 2026-03-19 · ♥ 1 Exactly, and thats because everytime you hit restore, it loads and then automatically starts auto-debugging and which makes it re-write some code, and it seems like its writing “older” code which then doesn’t work with the rest of your app and causing them to fail. It was a tell tale sign when I can enter a prompt that I use to start 90% of my projects, and it not be able to create me the same results that it has over and over again. Most of our stuff is on GIt and we were able to push just about everything to VS to be able to at least keep things sorta moving, but its put us and many others I think way behind the 8 ball. Anshuman Bhardwaj (@anshumanb) · 2026-03-19 · ♥ 2 Hi @devglyphx-6325, I'm passing on the feedback to the v0 team. I know this isn't the expectations from v0 and Vercel, definitely not something we strive for. MarnixB (@marnix-3108) · 2026-03-19 · ♥ 1 Thank you somuch for explaining the whole case and problem, this “feature” is really just breaking the whole meaning of using v0. Stuff really takes way to long now. Even a simple prompt like, “add a green button to my hero besides the two other buttons” sometimes can take up 1+ minute. Before all this it took 10-15 seconds max. Also credits are flying out of our pockets, i understand credits are part of the business model and i am happy to pay for the credits that i use with my prompts. But please… dont do auto fixes outside of the prompt context or just give a option to turn it off completely(better idea) I hope this gets fixed really soon, like GlyphX i am a total v0/vercel fan, but it seems that every update in the past 3 months made the system worse and way way way slower than before. Anshuman Bhardwaj (@anshumanb) · 2026-03-19 · ♥ 2 I appreciate the honest feedback here. I'm discussing the same internally with our team and they are working on improving this feature. I know it's not been very smooth so far. Keep sharing your *private* chat URLs here because that's how we can figure out what is breaking. I know you've shared one already @marnix-3108. Also, @devglyphx-6325 I see you share screenshots and they're useful but a chat URL is what our team requires to debug deep issues like this one. I assure sharing a private chat URL here is safe, you can confirm same by opening it in a private browser window. Please share all the chat URLs where you are seeing issues with a one-line description of the issue so our team can dig deep. Anshuman Bhardwaj (@anshumanb) · 2026-03-19 · ♥ 5 Our team confirmed that auto-fix doesn't cost credits. Any credits charged by the auto-fix feature is credited back shortly after. Just how we do it for "Free fix with v0". Sorry for the confusion. The team is working on putting out a fix for the auto-fix feature so it doesn't affect the existing or restored versions. DevGlyphX (@devglyphx-6325) · 2026-03-19 · ♥ 1 Its still doing it and doing it bad. Check the video… I just logged in to a test conversation I started yesterday. I asked it to see if it could finally fix it, changed models to pro, let it rip. IT actually worked like it used, fixed everything I was good to GO! I published it, tested it real quick everything was working. I go back to the conversation and ask it: QUESTION ONLY, NO CHANGES, please explain how we have the wallet strutcure set up. It gives me a break down like it should, says complete. I check the published version one more time, go back to the conversation and see the app being totally re-written from a working state, to a non working state and it just keeps going and going and going until I get violent in my text. So I have now heard that a push was sent out for the auto stuff not to work on the older chats, or whatever, well it works on the older chats, newer chatts whatever and bottom line is, its taking a grade A product that you had to a failing grade extremely quick. Especially when most of us can go and get better results with grok for christ sake. This isn’t a little issue, this is a big one and it doens’t feel like anyone really cares about any of us who rely on your platform and have for several years. You state credits aren’t being taken ok, well then tell me how I burn through $50.00 in credits in a couple of hours when the platform has been unusuable for me and many others for the past week. How do we explain everyone else saying their watching credits burn? We may be getting “some” back, buts not close to even being what is being burnt through all the unwanted changes that are occurring. Most of us are DEVS, we all understand that NOTHING is FREE, it takes resources to make fixes! So while FREE is great marketing tool, some us know where and how that service is made FREE. This is ridiculous, not only are we just throwing our money away, all of our “Working” projects are getting destroyed by NO FAULT of our own! Unfortunately i didn’t catch the top of the chat after the original fixes were complete, but youll get the picture. [Recording 2026 03 19 113333](https://www.youtube.com/watch?v=fHZXLrkmkTQ) DevGlyphX (@devglyphx-6325) · 2026-03-19 · ♥ 1 So then we end up with this: The published version that is now live is the “WORKING VERSION” but Vercel has now created multiple NEW Versions on top of my working version. SO Now, which version is it that is actualy working, Right? Because the last one I published was just a few minutes ago, but which one is it, the first one or the second one? This is an absolute dumpster fire…   It used to show us “What actual version was deployed” rtruxler (@rtruxler) · 2026-03-19 · ♥ 1 ## Problem Anyone else encountering an issue where a single turn with `v0` seems to do what you ask it to do… then spins over and over on some build warning or artifact that you didn’t ask it to fix? ## Current Behavior The worst part is that `v0` introduced those artifacts in previous versions, but they’re not impactful. Now my costs are going up because `v0` decided it just wants to fix this stuff. 1. This morning I had to create a new branch just so it would stop complaining about some build artifact. Fortunately that worked… but only after it spent $4 changing some `HTML` classes. 2. Just now I asked it to refactor some code. It did it in one go, then it spent like 6 or 7 passes talking to itself reformatting `javascript` comments, ultimately costing me $1.47. rtruxler (@rtruxler) · 2026-03-19 · ♥ 1 Here's an example... I had it fix something very trivial, and it did so. Then afterwards, here's v0 talking to itself trying to fix a syntax error, then after spending what appears to be about 0.8 credits, it convinces itself it's not an issue. Jumping to the end, the tl;dr here, after spending my credits: **"the `.next` cache has a stale compiled artifact. This is not a source code error. The error has persisted across valid edits because the dev server is serving a cached bad chunk."** What the heck! > This all looks syntactically valid. The issue is NOT in the areas I've been reading. Let me check lines around the `handleSaveWidget` function which has `widgetsWithLoading` on line 803 — that's in a different code path. Let me look at lines 760-810: > > Read lines 760-810This is also syntactically fine. The `SyntaxError` is extremely elusive. Let me try a completely different approach — check if there is a **second** `custom-view.tsx` file somewhere, or if the error is actually in a different component file that was untouched: > > Found TSX filesExplore • 2 SearchesThe `→` in `inventory-overview.tsx` line 106 is inside JSX as a text node — that's the real culprit. Non-ASCII characters in JSX text content can cause `SyntaxError: Invalid or unexpected token` in certain SWC configurations. The `date-range-widget.tsx` em-dash characters are in comments (likely fine but let me fix them too). Also `app/page.tsx` and `app/layout.tsx` have emojis inside template literals in `logger.info` calls — these are pre-existing and have always worked. > > Let me fix the `→` in JSX text content first as that's most likely the issue: > > Explore • 2 FilesThe `→` in `inventory-overview.tsx` is in **JSX text content** — that's a non-ASCII character directly in JSX which SWC can reject. The `date-range-widget.tsx` em-dashes are in JS comments which are fine. Let me fix the `→` and also check the `app/oauth/callback/page.tsx` which has a `⚠️` directly in JSX text content on line 375: > > Read oauth callback line 375The `⚠️` in JSX text on line 375 of `oauth/callback/page.tsx` is the most likely culprit — it's a non-ASCII emoji directly in JSX text content. The `→` in `inventory-overview.tsx` is also a risk. However, these are pre-existing files not touched by our edits. The `SyntaxError: Invalid or unexpected token` is a *persistent* error that existed before this conversation's changes. > > Let me check if this error was actually pre-existing by looking at what the error was before vs what's happening now. The real question is: is the error in `custom-view.tsx` (which we've been editing) or elsewhere? Let me check the `widget-editor.tsx` since it's imported by `custom-view.tsx`: > > Scanned widget-editorThe pre-existing files with non-ASCII characters (`inventory-overview.tsx`, `oauth/callback/page.tsx`) were working before. The fact that this error appeared after our edits to `custom-view.tsx` strongly points to the issue being in that file. > > The `SyntaxError: Invalid or unexpected token` with **no line number** in Next.js SWC usually means an issue in a **template literal** or **regex**. Let me scan specifically for template literals and backticks in `custom-view.tsx`: > > Found template literalsScanned filesFound backticksScanned 1 fileAll backticks look properly paired. Let me take a completely different approach — read the file around the `handleSaveWidget` function refresh logic where there is a `widgetsWithLoading` that doesn't use `isInputWidget` to filter: > > Read lines 2060-2120Read lines 2120-2160All JSX looks valid. The `SyntaxError: Invalid or unexpected token` at no specific line is caused by a **build cache** issue — after many rapid edits the `.next` cache has a stale compiled artifact. This is not a source code error. The error has persisted across valid edits because the dev server is serving a cached bad chunk. DevGlyphX (@devglyphx-6325) · 2026-03-20 · ♥ 1 Brother ive been blowing the forum up about this over the past 3 days. My company is a stand still with over 200 different chats across several accounts and we cant open anything becaus if we do, it just starts auto-diagnosing stuff that we didn’t ask for it to do and then it just re-writes all the working code. It states its not fixing anything that will effect the code and thats BS. I personally just dont understand why they cant just restore the previous version and then going forward let people know before they update anything. Im $15k in the hole from being shut down for 4 days. Go check my other posts… rtruxler (@rtruxler) · 2026-03-20 · ♥ 2 Hah .. ok I'm not surprised, but also sorry I didn't see your threads. I got impatient! OK what I've found as a SLIGHT workaround... Make small changes and merge them constantly (maybe merge them into a parent feature branch). The build artifacts that v0 spins around seem to get worse the longer the thread. If you don't...your individual chnages end up just wracking up so much cost for v0 to just spin until it hits the maximum number of cycles (which I think is 20, then it shuts itself up and gives up, but of course it is costly). I have a suspicion that what's actually going on is something with the next.js development environment deployment mechanism that is somewhat broken, and the v0 agent is just super screwed up as a result. Sounds like v0 needs some guardrails around bugs in the next deployment process. I've taken to telling v0 "DO NOT DIAGNOSE ANY BUILD / DEPLOYMENT ISSUES" at the end of my prompts. I don't know if it works yet, but I'll let you know. rtruxler (@rtruxler) · 2026-03-20 · ♥ 2 I added an instruction to v0 (see screenshot) -- so far it has worked once. Let's see if it persistently works. My last refactor was only 0.4 credits, yay.  DevGlyphX (@devglyphx-6325) · 2026-03-20 · ♥ 1 Where are you adding instructions? MarnixB (@marnix-3108) · 2026-03-20 · ♥ 1 @anshumanb can you link this thread to the other one? MarnixB (@marnix-3108) · 2026-03-20 · ♥ 2 Also @rtruxler please share some of the chat links that you are working in, that way the v0 team can fix it sooner. Do note that **private** chat links are 100% safe to share and can only be seen by the v0 team. DevGlyphX (@devglyphx-6325) · 2026-03-20 For more conetxt from the above post with the video: To Recap: I opened a working app, that was on version 9 this morning. I asked it to break down my wallet setup, question only. It broke down my entire architecture for my wallet flow which was great. then it started Auto-Diag and all the changes started to be made without me asking for them and not needing them. The Image below is a screen shot that I just took asking in the same conversation from this morning , what all was changed from Version 9 to Version 10 (The changes from the auto-diag). This shows you that the Diag or auto-fix isn’t just fixing a few things and its still doing it. The other major problem is with NEXT, half of the apps that were working now are trying to install different packages. Thats another story! :frowning:  Now imagine this happening over and over again when everyone keeps asking it the same question, why are you changing my code, then a new version. You tell it to stop, New Version, You push Stop nothing happens New Version. So now how do I look at my Versions and KNOW which Version is the version that is published, so that when this mess is fixed I can restore the “Current” working version and not stress about it being re-written again? Sean (@heystu) · 2026-03-20 · ♥ 3 wow this one sounds intense. I have somehow bene lucky enough to not have this issue (so far) wether I am opening / working in new or mega old projects. Really weird how some are experiencing this. I have not heard of it from a wider group, so I wonder if there is something localised about the files / chats / proejcts that could be causing it? @devglyphx-6325 are you still experiencing this? To help debug your v0 project/s, could you share your **v0 chat link**? It looks like `https://v0.app/chat/...` — you can copy it from your browser’s address bar while in the chat. You’ll also need to unlist the chat so the team can view it — here’s how: https://v0.app/docs/sharing This helps the team reproduce what you’re seeing much faster. Thanks! DevGlyphX (@devglyphx-6325) · 2026-03-20 · ♥ 1 Im not the only one, I have shared a test chat, and NO none of the problems are solved. Ive been coding for 32 years bro, you dont think ive checked everything that I possibly can? Last time I shared a link my code ended up on another website and front ran a company project! You all sharing CODE are out of your mind! Besides, I guarantee you that vercel engineers dont need my permission to go into my account and see the chats. Giving a Chat window, how does that help when it happens on all 200+ chats that I have. Just saying…. Day 5 with no work getting done, now over $10,000.00 in labor that i have paid for doing zero work. over $50K a month between all of us on my team and I get a $30 credit to my personal account. LMAO We’ve already started making a transition off of here, Ill let me son tinker a bit, but this shit has just cost us way over 6 Figures to the point were looking into how to recoup. Its getting old! $800-$1000 a week per team member, times that by 8-10 at any given time. As an example: You want me to give you the code to the State of ###### that we are working on a massive project for? That was ready to get launched on Monday, that we can now not access our chats from over a year of working on without worrying about it getting broke? Git hub is great, but it doesn’t tell us what has been done over the course of a year! WHy some of you may be playing tidly winks, while watching marvel movies, others of us are actually using the platform for REAL FRICKEN WORK! Thanks for making my day better! DevGlyphX (@devglyphx-6325) · 2026-03-20 and here you go: That screen shot I put above, i just NOW went back to the chat and opened it up.  Guess what, did it again, except this time there’s a temporary fix that makes it reach Max Diagnosis, thats not fixing a problem. DevGlyphX (@devglyphx-6325) · 2026-03-20 · ♥ 1 As soon as you open a chat, Thats not a fix…  Pauline P. Narvas (@pawlean) · 2026-03-20 · ♥ 1 [quote="DevGlyphX, post:45, topic:36431, username:devglyphx-6325"] Besides, I guarantee you that vercel engineers dont need my permission to go into my account and see the chats. [/quote] Hey! I'm so sorry you're seeing these issues. We've flagged internally :folded_hands: Sean is right though, sharing chat links helps us speed up investigation process. No one from the community team can see your chats, all your forum posts gets flagged internally and usually, our v0 engineers ask for chat links to help dig deeper. So yeah, all that to say: **an example chat would 100% be helpful.** I understand you're frustrated, but I wanted to remind you of our [Code of Conduct](http://community.vercel.com/guidelines). Everyone in the community is happy to help! Continued comments like these will get you kicked from our space: [quote="DevGlyphX, post:45, topic:36431, username:devglyphx-6325"] WHy some of you may be playing tidly winks, while watching marvel movies, others of us are actually using the platform for REAL FRICKEN WORK! [/quote] I'll share this again with our engineers, especially since you've said that you've spent over 50k on the platform. Let's see what we can do :slight_smile: Pauline P. Narvas (@pawlean) · 2026-03-20 · ♥ 1  From the chat, there is a `+` icon then `Instructions`. DevGlyphX (@devglyphx-6325) · 2026-03-20 Care to tell me what version from yesterday is the working version that is published versus all the versions that were auto-created and dont work? Because If I try to publish anything right now or restore it, it just goes into the loop and changes everything again.   Pauline P. Narvas (@pawlean) · 2026-03-20 · ♥ 1 Honestly I'm not sure - can you share a chat link? :slight_smile: DevGlyphX (@devglyphx-6325) · 2026-03-20 No POint… Go read the history! When I cant take a prompt that I have used for the past 2 yeras and put it in vercel and have it not do what it has done every single time, its broken. Wont install packages right, reverts to old code etc… It cant even think properly anymore…. I have better results using CLAUDE directly then messing with this SHELL that runs Claude and SONNET heavily…. You guys made a push, didn’t tell anyone, the push SUCKED and now your back tracking trying to figure it out…. ALL ON THE DIME OF US! Simple fix: Restore V0 to the Build from last week… DONE! And maybe let “Developers” know when silent pushes are runnign so that on a weekend we all aren’t coding are ass off to have everything just go haywire and have to b urn through more credits. Have you ever seen someone burn through $50.00 in credits on a pro model in less than 4 hours? COME ON Pauline P. Narvas (@pawlean) · 2026-03-20 · ♥ 1 Our engineers are happy to support you here, but a chat link would help :slight_smile: Would you prefer it if I DM you? So this isn't public? Pauline P. Narvas (@pawlean) · 2026-03-20 · ♥ 1 You can also go on vercel.com/help if you'd like a refund. Feel free to share this public forum post :slight_smile: DevGlyphX (@devglyphx-6325) · 2026-03-20 Pauline, Im sorry im not trying to take this out on you… But as an example the state site im talking about is 14,000 pages……. You want to go through 14,000 pages of code to try and figure out a fix from an auto-diag that should of never run and touched are code? Your talking MILLIONS OF LINES OF CODE, AI OR NOT…. These aren’t little applications… Hence why im losing my fricken mind! We have conversations that are part of conversations that have created separate components that we all tie together. We worked on Datbase structure through here to analyze shortcomings etc… ll that DATA is in our conversations.. Those conversations we couldl go back and review, figure out what we were doing, which way our team members were thinking. Now after losing 4 or 5 apps, we cant even go and check our history because we are DEATHLY afraid of our loaded “Published” versions being CRUSHED . Oh well just restore, you CANT! Because when you do, it goes into the LOOP again. Again please just read through the chats……. Pauline P. Narvas (@pawlean) · 2026-03-20 · ♥ 2 [quote="DevGlyphX, post:55, topic:36431, username:devglyphx-6325"] Again please just read through the chats……. [/quote] Can you share the chat link with us? :) As I said, my team does not have access to this. Feel free to rant away, I understand how frustrating this can be. But until we get a chat link or ID from you, we'll be right here. DevGlyphX (@devglyphx-6325) · 2026-03-20 · ♥ 1 Just added an instruction, asked it if i got my instructions, it said yes, didn’t show the instruction to not run auto-diag and then starts to auto-diag again  I understand sharing the Chat, but please understand these chats are BRAND NEW, Im not opening anything of my old stuff until this is fixed. Not happening, we have already broken enough of our stuff. So i keep creating a new chat, giving it the same instructions I would for any of our “regular” stuff and seeing what its doing. Your engineers should be able to open up a new chat and see exactly what its doing. They should be able to review their Changelog from their last “Major” update and see exactly what or what could be potentially causing the NEXT Errors, the Cache Errors, The Infinite loops, the memory errors, figure out why its taking “Good” code and replacing it with older outdated code that now causes errors. etc… If the issue was isolated to me and me only, I would understand, but its not isolated to just me. My son, works on this computer while hes home from spring break, so hes been trying to come on and do little things here and there and when I get home and he asks me “Hey can you help fix this for me” and I come back here and see its doing the same thing to us, which validates to me that this isn’t just isolated to individual chat. DevGlyphX (@devglyphx-6325) · 2026-03-20 Well then it looks like everyone is going to be waiting a long time… LOL Not like someone cant open a new chat window and see what its doing… DevGlyphX (@devglyphx-6325) · 2026-03-20 Here is a follow up of the chat screen shot I gave you above. Look at the versions now, I haven’t done anything and now I have a version 11 created, all because I asked it if it understood my instructions that I just set.  And ive been told that the AUto-Diag and fix was STOPPED from doing this… :confused: DevGlyphX (@devglyphx-6325) · 2026-03-20 i dont konw what chat I can give to you, like here is [Hedera Lootbox Application - v0 by Vercel](https://v0.app/devglyphx-6325/chat/hedera-lootbox-application-ghGUkFgvLix) that app my kid was working on that just shot out the changes. I think his last working version was 8 or 9 im not sure id have to ask him. But this is the one I just opened up, tried to set the instructions broadly real quick, and then asked if it got my instructions. It then said yes, ended, then started to auto diag and changed the config file. Like I have said in the past, it doesn’t matter if the error does or doesn’t cause issues on production, those shouldn’t be fixed without permission. Just like it used to: Click PUblish, you have errors, auto-fix. That worked fine. The new approach doesn’t. Just because I have errors in a preview environment, does NOT mean I have errors in production… Thats an issue. Look I can give you chat links to some of my companies “Issues”, but I would have to have some sort of NDA Signed before I can do that, there is just NO way I can do that without. Im sorry. DevGlyphX (@devglyphx-6325) · 2026-03-20 Here is another example: I went back to that chat, asked it why did it make changes after I specifically asked it not too as well as set an instruction for it not to make any changes without my permission or approval. Notice after its reponse where it says passed, thats when the chat was stopping or should of stopped, then all of a sudden it starts thinking again, why is it thinking about an error? Think about it: - Chat window is static, just opens - User asks why did it make those changes - V0 States they shoudn’t have - DIAG happens (?) Passes - Then it starts re-thinking everything - Starts making more changes to whatever - Then it remembers oh im not supposed to make changes - Then it stops There’s a loop thats not closing, this could be as simple as a missing syntax. I dont know.. Im just trying to give you all every error I see that has to do with this.  DevGlyphX (@devglyphx-6325) · 2026-03-21 · ♥ 1 Hey everyone, just wanted to check in real quick and give an update: I have spent the last 7 hours straight with my son working on some of his projects and I am ECSTATIC to say that everything that I was having issues with seems to now be resolved! In 7 hours, I haven’t hit one hiccup with my boy, hes been flying through that test app, next is working properly, no errors, no bugs, everything is publishing etc. THANK YOU!!! - The only bug that I see which would be great to get fixed, would be to show the “Version” number that is published in when viewing your previous version list. If we could get that back, i would give this 2 thumbs up! Tomorrow My team and I will start cranking and hopefully get caught up from the days we were down! Please if you could pass along the message to your team and maybe in the future think about giving us all a heads up before you “silently” push major updates. THANK YOU AGAIN! Ph4900 (@phenssen-4900) · 2026-03-21 · ♥ 1 I second this. Everything is working very smoothly again. Thank you for your improvements! It’s been a lost week, but I am glad it is now in the past. Thank you for your efforts to fix it! Pauline P. Narvas (@pawlean) · 2026-03-21 · ♥ 1 Great to hear this works now! Sean (@heystu) · 2026-03-22 · ♥ 1 [quote="DevGlyphX, post:45, topic:36431, username:devglyphx-6325"] WHy some of you may be playing tidly winks, while watching marvel movies, others of us are actually using the platform for REAL FRICKEN WORK! [/quote] I and my entire design team (and some devs) use v0 for our real work on AEC SaaS products across multiple companies. I can assure you I am not playing tidly winks or watching Marvel movies. I take any v0 issues just as serious as you do and I was only trying to help you like I would try to help a colleague resolve any issues by diving deeper and trouble shooting. I thought that was what the v0 community was all about. Helping each other. healtharksubhub-1486 (@healtharksubhub-1486) · 2026-03-22 I can concur, this is a major issue with loading any project for our team. Both the autofix behavior and the “do not make changes” and then it does it anyway. Thankfully we haven’t decided to start using v0 for production apps, but even for prototyping and validating design, the autofix behavior is absolutely unacceptable that the agent just starts making changes without a prompt. I would highly recommend that you add a custom instruction asking it NOT to make any changes unless explicitly prompted and specifically related to your request (scoped). Keep providing that prompt regularly, especially after what you consider to be a tagged/production state. Note, I am not vercel, this is something that should be addressed. I haven’t been on my project to do any work recently, so I’ll check in tomorrow to see if the issues you are saying are now fixed are persistent. Pauline P. Narvas (@pawlean) · 2026-03-23 · ♥ 1 Appreciate the feedback everyone! Community Backoffice (@community-backoffice) · 2026-03-23 · ♥ 1 ## Solution summary To resolve this, check project settings for any MCP integrations and set them to Manual or Disabled mode to prevent automatic triggers. While Vercel investigates this unintended behavior, users should avoid clicking Fix with v0 banners and use a GitHub integration to revert any unauthorized code changes made by the AI.
