Gemini felt dumb on December 31, 2025.

I keep hitting Gemini 3.0 and it randomly caps words and wraps half the output in quotes, even when I explicitly tell it not to. About 40‑60% of the text looks like a broken human writer, forcing me to waste time fixing the glitches. It feels like the model ignores my instructions, making the tool more of a hassle than a help.

I tried using Gemini 3 and was constantly bombarded with unnecessary quotation marks, which made the output look cluttered and hard to read. The tool’s behavior was irritating, turning simple responses into a mess of stray quotes, and it felt like the model wasn’t handling punctuation properly, which broke my flow and trust.

I tried two variations of the AI’s reply and instantly felt a difference. The “A” version said “absolutely right,” which annoyed me, while the “B” version started differently and felt much more natural. That subtle change made the response far more pleasant, and I’m optimistic the upcoming release will keep improving this tone.

I tried Gemini for the first time after getting fed up with GPT’s overly formal, robot‑like answers. The voice toggle that lets me keep the convo flowing without constantly hitting the mic felt refreshing, and the replies were concise and “chill,” just what I wanted. It’s still early days and I notice a few rough edges, but overall the experience was pleasant and surprisingly helpful.

I tried using Gemini Pro to summarize my lecture slides, make flashcards, and answer course questions, but the tool kept hallucinating massively. When I uploaded a PPT, it produced notes about completely unrelated topics I'd asked about before, even pulling from old chats. Asking why only got a generic apology. It behaved like it couldn't focus on the new image, repeatedly spitting out past answers. The experience was frustrating and felt very dumb.