Gemini felt smart on February 12, 2026.

I keep getting slammed by Gemini’s memory in the app—it hoards my job info and forces it into every conversation, even when it makes zero sense. The weighting is so aggressive that irrelevant context hijacks the chat, making the tool frustratingly noisy. I’m that close to turning memory off because it’s ruining the experience, especially compared to Claude or OpenAI, which don’t have this problem.

I spent months treating Gemini like a simple search engine and got bland, generic answers. When I switched to chatting with it as a partner—telling it about my project, letting it interview me, and asking it to probe my goals—the responses transformed. The tool began acting like a high‑level consultant, asking insightful questions and offering tailored advice, which felt like a real “superpower.”

I was trying to run a long prompt in Google AI Studio’s Gemini 3 Pro preview, but halfway through the response I hit an error screen. The screenshot shows the cutoff, and I’ve also hit a 50 k token ceiling that now throws other errors. It feels like the tool’s generosity has vanished, leaving me frustrated and unable to get full answers.