I feel like v0 will do great for awhile and then like starts making some of the dumbest mistakes and acting like it’s stupid all of the sudden and it take a million prompts to get the project back to being usable.
Is anyone else experiencing this?
Seems like every time I top up credits too. Like I will burn all these credits already with everything I want to do. You don’t have to run up the bill making obvious mistakes or breaking the project. It’s very frustrating.
Absolutely 100% I’m noticing this. Specifically after topping up credits also.
It will build a metaphorical space ship with 1 prompt. Then if you tell it to change the color, it will take 50 prompts and break everything in the process. It very much feels like sabotage and it’s infuriating me. It got specifically bad 2-3 days ago.
@anshumanb@jacobparis@mleiter@fernandorojo You have 24hr to stop hiding out and address this, or I start posting video proof on social channels. Explain yourselves NOW!
Hi @cappynate, I understand you are unhappy with your v0 experience.
For small changes like color or font, we recommend using the Design mode, which gives you better control over the styles and doesn’t cost any tokens either.
However, making theories that v0 is “sabotaging” your project or threatening staff/community isn’t an actual constructive feedback. Our team is working harder everyday to make it a better product but at the end it’s AI and its outcomes are probabilistic, which can sometimes be less than ideal.
You can overcome this by using custom instructions, design mode, and prompt enhancements.
Here are some resources you can use:
The model works wonders when my account is nearing empty. Then when my account is funded, it breaks basic stuff. Predictably, like clockwork. Literally single item changes (change 1 single color code, reduce a margin). It ignores detailed instructions, it ignores verbatim instructions, it amends code that is copied verbatim. I didn’t start this thread, so clearly I’m not the only one experiencing this.
I appreciate the design mode suggestion and will try it. Looks helpful, thank you.
However:
-V0 has stuck in a retry loop for 9.5min bleeding tokens, all for a single line change.
-V0 repeatedly creates syntax errors on simple single line changes. (Seemingly intentionally)
-V0 charges me for “Fix Free” edits repeatedly.
-Recently (last 2-3 days), V0 started overweighting old messages vs the most recent one.
I’m giving you the feedback. Please show a willingness to do something with it.
went into the retry look for more than a few minutes. 9.5 min sounds like a lot.
any example of this would be helpful as well
v0 charged you for the “Fix free”
I will ask the team about what changes they did in last few days, they might be looking for this feedback.
I really appreciate you pointing these out. Would love to share these with the team and run a feedback loop once we have the message and chat IDs so they can inspect.
I’m not sure where to find the reference ID of each message, if you could direct me, that would help. Maybe these are them? Here’s screenshots in the meantime. Happened twice and tried other times before I cut it short. This was to move a back button from a header to a banner. Single prompt. The last one was changing colors.
Here’s a string of 13repeated Syntax or JSX errors on a simple, single page, wireframe app. From scratch, nothing complicated, only 250 lines of code. No functions, just mockups. As easy as it gets. This chat is right after I topped my account balance up. Here were the prompt summary that generated each Syntax/JSX, most where I explicitly warned it not to touch structure. Again this started right after the largest account top-up I’ve ever made.
As the Op, I do think CappyNate might have taken this a bit far BUT he did get your attention. Gaslighting us for asking if others are feeling the same is not appropriate. I personally love what you fine people have made but the experience we’re reporting is REAL.
v0 charges me EVERY time I click fix for Free with v0. The initial thought may be free, probably not, but then every follow up, not a new message just it figuring itself out, is for sure charging even though it’s part of the same “Fix for Free” series of thoughts. I left it thinking the other night and came back to ~6.00 credits disappearing AND the problem wasn’t fixed and I had to restore and re-prompt.
I’ve started just walking away when I can tell it’ having a spell and coming back later. I’ve spent and plan on continuing to spend a fair amount of money with you but this gets rather frustrating at times…. like personal hell level frustrating.
I also struggle with V0 charging me for “Free FIxes” its not all the time but when errors do occur at times its 50/50 if v0 charges you or not I have made a thread about it and no response.
I see. The timing of these issues definitely makes it look weird. Let me direct your messages to our Support team and product team.
For the “free fixes” can you check the message billing history and see if the free fixes were not reimbursed? I need to confirm this one with the team.
Thanks for sharing your points here @corebytes and @ben-6576. Please add your message and credits consumption information like CappyNote so I can investigate further.
That’s good feedback, Eric. We saw some abuse of the free fixes when we launched it so have a limit on the number of times fix with v0 is free each day (99% of our users never hit that limit).
That said, the UI is not clear about when you’re past that limit and Fix with v0 is no longer free. We are fixing that shortly!
See my message to Eric – there’s a fraud/abuse limit on the number of times each day that Fix with v0 is free. We’re updating the UI shortly to better reflect that.
My team (v0) cares deeply about improving the performance of the agent. If you have feedback or examples where it fails to fix issues or gets stuck in these negative loops, please share them with us (here or via the feedback widgit in the app) – we will dig in and find ways to optimize.
So yesterday I noticed that my fixes were no longer free, and I saw it commended in the forums that there’s new logic to avoid abuse.
I don’t disagree with the logic to add abuse prevention measures, but a daily limit is a really bad approach IMO as it doesn’t count for the overall spend of a user - I was using V0 simultaneously over two projects yesterday meaning I was spending a lot, but also there were errors. I ran out of free fixes by lunchtime and it stalled work as I’m not going back to paying to fix things that should be correct.
You should rework this taking into account not just the volume of prompts, but historic prompt usage/spend on a rolling basis (last 30d or something)
I’m sure that many users likely have days where they don’t have issues, until they do where then there’s quite a lot of issues all on one go, a hard daily limit does not help with this.
Often I see a Fix with V0 where I’m pretty sure I know what the issue is, and with a little bit of context (like context about the DB logic) stuff could be fixed with lower token usage, and less repeated efforts to fix.
Naturally this would open up an abuse mechanism where people could stuff additional request to get them as part of free fixes, so logic could be for AI to split the requests:
Fix + Fix context
Additional request
I think implementing something like this would help to move things along well for users
Here’s another example of this whole comingled issue: Eating credits, Charging for fixes, Looping, Running rogue 7 minutes, Questionable stuff, etc.
This attachment is a single prompt. It’s even heavily shortened.
-V0 worked on a task and created a syntax error, (credits charged obv).
-It then automatically tried to fix that error unsuccessfully (credits charged here too).
-It continued automatically retrying, creating a new syntax error (when “fix free with V0” showed).
-It did this for 7 minutes, draining my account.
-I did 3 prompts, it did this on each one.
-In total it charged me $5, and still submitted a broken version with syntax errors.
All times it tried automatically “fixing”, the “Fix Free with V0” banner showed on the screen to do a manual fixing. But the system was stuck in prompt/working/retrying, so I couldn’t click it.
From a usability standpoint, auto-fix is great. Set-it and forget it.
But not if the result is bad, it runs rogue for 7min+, and charges for every fix when it says it won’t. I feel like I have to watch it like a hawk to prevent runaway and draining my account. It just leaves a really bad taste in my mouth.