Gemini felt mid on January 12, 2026.

I’m frustrated with Google’s AI rollout—after they handed out a free year of the Pro subscription, they’ve started throttling real users. The Gemini 3 model now has a reduced context window and feels sluggish, making the tool far less useful than before. This downgrade is annoying and hampers my workflow.

I ran a quick experiment with Gemini/Bard, feeding it the same prompt in two styles: one vague, one with a tiny logical structure. The structured version didn’t work miracles, but its answers were far more predictable and cut down the irritation I usually feel. It didn’t fix every issue, but the reduced frustration was noticeable, and I’m curious if others have tried similar tweaks.

I asked Gemini to redesign a UI using React and Tailwind, referencing a screenshot, but the canvas it spit out was totally unrelated and nonsensical. This has been happening for days, and it feels like the model is being throttled or A/B‑tested. The output was frustratingly off‑track, making the tool practically unusable right now.

I’ve been experimenting with Gemini and the nano‑banana model for image restoration and colorization, and overall the results have been impressive. The tool revives most pictures beautifully, but when it comes to very fine or heavily degraded textures, it still struggles and leaves noticeable artifacts. Still, the experience feels solid and mostly reliable, just with a few frustrating edge cases.

I tried countless AI image generators hoping they'd capture a nostalgic street scene I'd always imagined. At first every picture felt empty and off‑brand, which was incredibly frustrating. After many tweaks and small experiments, one result finally clicked—its vibe matched the dream I’d held for years. Seeing that image made me feel like I’d lived a slice of my impossible wish, even if just for a moment.