Gemini felt dumb on January 9, 2026.

I was surprised when I asked Gemini for Arabic/social studies revision links and it spouted a list of URLs instead of the custom text it used to generate. I discovered this today and can’t recall seeing the same behavior back in May 2025. It felt like the tool had changed, and the shift left me a bit frustrated because I rely on its original output style.

I tried to create a detailed Japanese city image using Gemini, but one of my accounts kept saying there was heavy traffic and wouldn’t produce anything. The other account worked fine with the exact same prompt, so it wasn’t my wording. It was irritating to hit a wall for no reason, and I’m left searching for a fix or workaround.

製作程式錯誤百出

I tried using Gemini 3 Pro for a deep‑dive tutorial on setting up a new laptop—from debloating to optimal settings and FOSS tools. I asked for a step‑by‑step manual, but the reply was only a fraction of the length I got from ChatGPT. The ideas were solid, yet the instructions were vague and far from the granular walkthrough I needed, leaving me frustrated that Gemini seems to hold back on detail.

I use Gemini every day, but the default UI was a nightmare—cluttered chats and lost images. I built Toolbox for Gemini, adding unlimited smart folders, export options, a prompt library, and a one‑click AI Prompt Enhancer. Now I can organize projects, pin key messages, and chain prompts automatically. It’s transformed my workflow and even 2,000 other users are benefitting.

I tried to design a costume‑swapping magic trick with Gemini, making sure it stayed modest and stage‑like. The model suddenly blocked me for using rope, yet let handcuffs and drawstrings slide through, and even flagged a female magician in mesh pantyhose as porn. The inconsistency left me irritated and confused, questioning whether the tool’s moderation rules are even fair.

I've been using Gemini 3.0 lately and it’s been a nightmare—responses are slow, often garbled, and the web app crashes on simple prompts. I’ve lost time trying to fix its output, and the slowdown feels like a step backward. I’m urging the Google team to notice this regression and act, because the current state is seriously hampering my workflow.