Gemini felt smart on February 16, 2026.

I ran side‑by‑side tests comparing Nano Banana Pro to GPT 1.5 on a list of well‑known supercars. Using the prompt “Generate an image of [Car Model]”, I took the first front‑facing result and measured how accurately each model captured the cars. Nano Banana Pro hit near‑perfect scores (95‑100 %), while GPT 1.5 lingered around 80 %, making the difference feel stark and a bit frustrating for the older model.

I’m frustrated that the Gemini app, which used to have a thoughtful, context‑aware filter, now abruptly cuts off its streaming replies whenever it spots something it deems inappropriate. The new filter feels like a trigger‑happy political watchdog, stripping away content in a shallow, over‑zealous way that makes the tool feel unusable and overly censored.

I tried Nano Banana Pro on a bunch of old family photos and vintage album covers, and the results blew me away. The upscaled images kept the original texture while adding crisp detail, making the pictures look like they were taken yesterday. I was thrilled seeing the tool bring forgotten memories back to life without the usual artifacts or over‑sharpening.

I’ve been using Gemini Pro for RPG writing and recently noticed a drop in quality. The NPC dialogue now just spouts questions and fails to move the plot forward, leaving me stuck in circles. I keep going back to AI Studio for better output, but the limits are frustrating. I’ve tried other AIs, yet Gemini used to be the best, so this regression feels disappointing.