Urgent: Refund Request due to AI Loop Error and Failed Task Execution

Google Antigravity
, ,
1

Dear Antigravity Support Team,
I am writing to formally lodge a complaint regarding a major issue with your platform that resulted in the complete depletion of my paid credits without delivering any results.
For the past 3 hours, I have been trying to fix a single problem within my project. Initially, I used Claude, and later switched to Gemini 3.1, but both models exhibited the exact same faulty behavior. The AI repeatedly claimed that the error was fixed (“I have resolved the issue”). Despite my explicit instructions to “test the code first and only deliver it if it actually works,” the AI ignored the command, provided non-working code, and entered an infinite loop of repetitive, failed attempts.
As a paying customer, I lost 3 hours of my time and all my credits on a task that should have taken 5 minutes, purely because the AI failed to validate its own output and repeatedly charged me for identical, faulty responses.
I request the following actions to be taken immediately:

  1. A full audit of my recent chat history to verify this looping behavior.

  2. A total refund/restoration of the credits deducted during these past 3 hours.

  3. A review of the prompt-following mechanism, as the AI completely ignored the instruction to test before delivery.

    I have attached screenshots of the prompt logs and the recurring errors as proof. My account email is: f..@gmail.com
    I expect a swift resolution and the restoration of my credits so I can actually finish my work.

    Best regards, Feliz Smith

2

This is not really the place to report it. Raise a ticket with Antigravity Feedback

That said, AI issues are more likely created from context rather than the models themselves, I would recommend highlighting the exact files and providing clearer instructions. If Opus can’t resolve your issue, then either you need to refine your prompt, or your issue is too hard

3

You are describing a fairly well known AI artifact that’s generally called “hallucinations”. Basically, the AI rejects our common version of reality and substitutes its own. When you point out one the AI’s reality’s flaws, the AI invents a different version of reality where this flaw doesn’t exist but is replaced by a different flaw. And then when you point out the other flaw it simply tells you about its first version of reality.

Recognizing when the AI gets stuck like this is something everyone who’s working with AI needs to learn. It’s understandable, when one is working on a domain they’re unfamiliar with, and are relying on the AI’s knowledge to accomplish the given task. But it is what it is. You’re welcome to try raising a ticket, but it’s unlikely that AG support will do anything.