Just as I previously came here to criticize( The 3.6 Flash model... it's weak at coding - #8 by ExuCaveira_666 ) model 3.6 for being weak at programming, I now want to offer my congratulations. This model is on par with DeepSeek V4 Flash 0731; it has stopped making basic errors and shows improved task completion. The reasoning capabilities are excellent—this is the kind of commitment developers look for, regardless of the broader AI race. Quality is crucial in coding, and the training team deserves praise; model 3.7 is truly an evolution of 3.5. Thank you to DeepMind and Alphabet for listening to us and improving your products—it is great to see this capacity to engage rather than ignore developers, as other labs often do. In my view, ignoring feedback drives users away—whether the product is free or paid—because a lack of visible evolution leads to disengagement. You are that trend, and rest assured, this will drive increased usage of your excellent products.