Maybe it is my way of using Gemini 3 Pro but I don't see the benchmark gains of 3 Pro over 2.5 Pro reflected in my daily use at all. There's definitely none of that "wow-factor" when Gemini 2.5 was released.

Is anyone seeing significant improvements in day-to-day coding use? If so, it would be great to get a glimpse into the use-case (language, application area, complexity) and how the model is "driven"/used.