Gemini felt dumb on December 24, 2025.

I keep trying the Gemini app and get annoyed by its shallow, cut‑off replies and constant blocking on sensitive topics. The UI feels clunky on mobile, especially compared to other tools. Yet when I switch to the raw Gemini model via the API, the answers are deep, coherent, and surprisingly smart. It feels like two different products—one useless for everyday use and another powerful but hidden behind a technical interface.

I tried using Gemini 3 Flash in free mode and was immediately hit by a tiny limit—just four prompts total. Within those, the model kept tripping up, making mistake after mistake, and I couldn’t even get the one change I needed to work. The experience felt needlessly restrictive and frustrating, leaving me doubtful about its usefulness.

I keep noticing Google AI slipping the year “2025” into almost every answer, even when it makes no sense. It feels like the model is stuck on a buggy default, turning simple queries into odd, year‑filled responses. I tried asking normal questions, but the constant insertions are annoying and make the tool feel unreliable and less useful.

I tried to get Gemini 3 Pro to swap columns in an SVG, but it kept adding a white gauge and sprinkling comments I never asked for. When I told it to delete the comments and the gauge, it arbitrarily switched the layout to HTML divs. Even after repeatedly demanding the gauge and divs be removed, it kept re‑adding them or leaving stray comments. The tool’s behavior was infuriatingly uncooperative, forcing me to repeat the same corrections over and over.

I’ve been testing a bunch of new models, but when it comes to realistic faces and portraits, Nano Banana Pro just wins every time. On imini AI it spits out 4K images that keep detail around the eyes, skin texture, and lighting transitions – the spots where most models crack. Seedream 4.5 looks great stylistically, yet for pure realism I feel safer with Nano Banana. I’m curious what others are using for portraits now; has anything else gotten close?

I tried to get Gemini to summarize a YouTube video, but it immediately replied it was “having a hard time fulfilling the request.” When I prompted it to try again, it started fabricating details out of thin air. The hallucination was jarring, and the whole experience felt unsafe and useless, leaving me frustrated and doubtful about relying on the tool.

I’m fed up with those “Would you like me to…?” pop‑ups that keep interrupting my flow. Every time I ask ChatGPT something, it jumps in with a generic offer to help further, even when I didn’t need anything else. It feels like the model can’t read my intent and just clutters the conversation, which made me quit using it.