Gemini felt dumb on April 14, 2026.

I tried using Nano Banana Pro to put my face onto images, and it works fine everywhere else, but when I run it in the labs environment the results are terrible – the face looks nothing like me. The output is distorted and unrecognizable, which was really disappointing and made the whole process feel useless.

I’m fed up with the Gemini app’s read‑aloud feature cutting off the moment my screen goes dark. I have to keep tapping to keep the display awake, which defeats the whole point of a hands‑free audio assistant. I’ve been dealing with this since launch, despite dozens of updates, and no background audio support. It’s incredibly frustrating and feels like Google never tested real‑world use.

I was stunned when Gemini started spitting out a whole set of 100 MCQs that I never asked for. I’d set clear custom instructions to keep answers brief, but the model ignored them, creating an elaborate flashcard series out of thin air. The unexpected, unrelated content felt chaotic and wasted my time, making the whole session frustrating.

I kept trying to query Gemini’s Flash Lite and Gemma 4 models, but each request returned a 503 InternalServerError saying the model was “currently experiencing high demand.” The constant failures blocked me from getting any results, and the repeated “please try again later” messages were really frustrating and wasted my time.

I tried Gemini for a week as a med student who just needs a grammar‑check‑plus‑research buddy. It gave decent summaries of pathophysiology, handled drug‑interaction queries without crazy hallucinations, and helped polish my drafts—basically a solid grammar checker. Nothing blew my mind, but it wasn’t a disaster either; the mobile UI is just “passable.” Overall, it’s simply fine for my modest needs.