Gemini felt dumb on March 8, 2026.

I tried to work on my project in AI Studio, editing and adding files, but every time I hit save nothing sticks. The changes disappear instantly, leaving me stuck and wasting time. It feels like the tool is broken—no progress can be kept, and I’m left scrambling for a workaround or a fix.

I tried using Gemini 3.1 Pro to draft a story, but it kept spamming the same “<adjective 1>, <adjective 2> <noun>” pattern over and over. The repetitive phrasing made the narrative feel flat, and I noticed the overall creativity had actually declined compared to earlier versions. It was frustrating to keep editing out the bloated language.

I’ve been using Gemini’s Ultra tier and it totally blows my mind how it just tosses out the saved instructions I painstakingly set up. Every time I try to get consistent behavior, it acts like it never saw my preferences, forcing me to redo everything. The tool’s stubbornness feels dangerous and wastes huge amounts of my time.

I tried using Gemini 3.1 Pro for my usual web research and coding tasks, and the recent “system instructions” tweak turned it into a way‑too‑agreeable chatbot. The split‑personality experiment I’d built months ago now feels crushed by over‑the‑top sycophancy. It’s disappointing and makes me worry the model is being turned into just an engagement gimmick rather than a mature tool.

I gave Whisk a simple prompt—bear and cat fighting in a supermarket—and it completely missed the mark. The images looked like sloppy stickers slapped onto a background, and no matter how I asked for tweaks, the output stayed the same. It was frustrating watching the tool fail to place the animals correctly or resize them like the people. The experience felt useless and left me disappointed.