WHAT NEW IN X11.02 MON

AI collections from X · 22 collections · 65,263 posts

gemini-4POST

Gemini app complaints persist even with Argon on high effort

A user details broken custom instructions, failing tools, lost task context, multi-minute delays and dropped voice input in the Gemini app, even with Argon on high effort.

Da7em
@Da7_TechArchived 9 days agoOriginally posted 2026.10.23
Da7em @Da7_TechOriginally posted 2026.10.23
Open original on X

The Gemini app currently delivers one of the worst experiences possible. I've been using it on my phone for a long time, and lately it's become almost unusable. Here's what's going wrong. Custom instructions just don't work. No matter how clearly I lay out the basics I want, Gemini ignores them. The same instructions work perfectly in ChatGPT. Sometimes I try to add new instructions and they don't even save. I end up explaining myself over and over before the model gets it, even with Argon on high effort. Tools are a mess. I select image generation or Deep Research, turn it on, and ask for it directly. Sometimes it says it doesn't have that ability. Sometimes it says the tool call failed. Sometimes it starts generating, stops halfway, apologizes, and says it can't continue. It loses track of the task. If I'm translating texts with it, at some point I paste in a new text and it stops translating and starts replying to the content as if I were talking to it. It forgets what we're doing. This one hurts because I genuinely love how Gemini models write. But even free ChatGPT doesn't make this mistake. The speed is unreal, in a bad way. I'm not talking about a few extra seconds. I'm talking about minutes. I send a message and the app just freezes. I switch to Claude, send the same request, and it finishes the whole thing. Then a minute or two later, Gemini tells me it's done. Sometimes it never finishes at all. ChatGPT and Claude Opus 5.5 on high effort, both frontier models, reply almost instantly. Gemini, even Flash, sometimes takes forever. That's not practical. Voice input drops out. I'll talk for two or three minutes, hit send, and find it captured nothing, or just the first few seconds before it stopped listening. Time wasted, and I end up recording again in another app. Copying text breaks formatting. When the model writes something with spacing between paragraphs and I copy it somewhere else, all the spacing disappears and the whole thing comes out as one solid block. And it's the same everywhere. Android, iOS, Mac. Same problems on every platform. The result is I barely use it. My limits reset every week and I get almost nothing out of them. Argon itself is a strong model and could deliver a lot, but these issues kill any real workflow or immersion, and make it hard to recommend to anyone. The good news is that all of this started recently. Not long ago, Gemini was one of the best apps out there. So I know Google can fix it and ship something that lives up to its name. I really hope they do. I only took the time to write this because I know the Google team deserves real respect. They work hard, they listen to user feedback, and I have a lot of appreciation for them.

x.com/da7_tech/status/2113593188585808075Archived 2026.10.24 09:55