My subscription just ended. Until the very end, it kept lying, which is really annoying.
It messed everything up so creatively that I started to doubt whether the results of the project I had been working on for the past three months were even reliable.
So far, using it in detail,
I mainly use Antigravity CLI. I create scripts in the terminal, run them, monitor them, analyze the results and organize them.
For the first two months or so, I used Pro and it was okay.
But suddenly, it said my usage limit had been reached. When I checked, there was a sudden 5-hour policy? Shouldn't they have told me about this change?
Anyway, my token supply for a week ran out in just a day and a half, so I panicked. I couldn't afford to waste the remaining three days, so I upgraded to Ultra.
The reason I subscribed to Gemma was that if you register for the developer program, it gives you API usage credits. For example, if Pro is £18.99, it gives you £7 worth of credits, and if Ultra is £79.99, it gives you £30 worth of credits, making it cheaper than Claude. I needed to process large amounts of text.
Originally, I subscribed with a personal account and only used it for company work when needed, so this condition was good for me.
Ah, that's when the misfortune began. As the project progressed, incomprehensible bugs started to appear. It's too complicated to describe in detail here on this forum, but I'll try to summarize the core issues:
“The results should be the same every time I run this script. I fixed the random seed.”
“Yes, I fixed the random seed to 42. This will solve the problem and here's a simple test result showing that the results are the same even with repeated execution.”
A few weeks later, something seemed wrong with the results, so I checked again and suddenly it said “The random seed is None”. Or,
I reduced the items from 7 to 3 and experimented for a few weeks, but suddenly it started using the remaining 4 items without any explanation. Or, I added a new parameter value to the existing 5 values, and it changed the existing values without any reason or even removed them.
It goes on and on.
Finally, I also asked the Gemma chatbot about the subscription end time. “I received an email saying my subscription ends on the 17th, but what time exactly does it end? Is it 00:00 on the 17th or 23:59?” When I asked that, it said midnight, and even after asking several times to confirm, it insisted that it would end at 23:59 PST on the 17th.
However, the subscription cancellation ended earlier than that. So I asked again, “I just received a notification that my subscription was canceled? Are you lying?” And it said, “Actually, Google calculates the subscription start time.” (Omitted harsh words)
Moreover, there are quite a few posts on Reddit about Gemma Shield.
People say that Gemma is actually very good, but when it comes to providing services to users, the limitations imposed result in lower performance. Or that people don't understand it properly and complain about it. Or that AI is like that anyway, so why don't you know? They also question how much money they are receiving for posting these negative comments.
Judging by the tone and content of the messages, they seem like bots. When I confidently said "Gemma isn't that great" in Korean, a reply appeared instantly saying “ㅋㅋ ㅈ* 하네” (a derogatory Korean expression).
It was fun and dirty to be with you for so long. Goodbye forever, Gemma.
