Gemini felt dumb on May 16, 2026.

I tried out the Gemini 3.0 fast model and man, the results were just straight-up unhinged! It’s so weird and nonsensical that it honestly felt like my brain was being fried by a chaotic brainrot edit. While it's hilarious to see the tool act this way, the experience was just pure, overwhelming absurdity. It’s definitely not what I expected, but it's incredibly wild!

I'm feeling so let down. It is honestly so frustrating to realize that their SOTA model can only output half the tokens that the competition can. It feels like a massive step backward whenever I try to use it for complex, long-form tasks. I really expected much more, and it's just sad to see such a significant limitation holding it back from being truly great.

I'm convinced the Pro model got a quiet update because it is absolutely brutal! I don't even use it for programming, but the way it performs is just mind-blowing. It totally outclasses GPT 5.5 and Claude 4.7. Honestly, I'm terrified; it's going to hurt like hell when Google eventually nerfs this beast. I'm just riding this wave of incredible performance while I can!