Thanks for the suggestion! We’re still sticking with tab-based autocomplete rather than always-on - while developing the feature, I did field research and found that it’s nearly an even split for auto-vs-trigger completion, and our personal preferences at the office were for triggered completion, because it puts less pressure on the performance of the method that gets hints, and it poses less of a usability issue if you have thousands of completions.
But the new tutorial series mentions autocompletion in the first tutorial, which I hope helps to draw more attention to it.