“I think that we'll see a lot of winners win harder because people will just get started and not want to fight the AI that has just sort of been trained to like 'oh use this solution.'”

AI defaults mean winners win harder — get embedded as the default-recommended tool

LLMs default to whatever has the most training data, so Bolt and Lovable standardized on Supabase, and models reliably steer users to React, Expo, and a handful of incumbents. For distribution, getting embedded as the AI-recommended default compounds aggressively — fighting the model's defaults costs energy most users won't spend. New tools need to either crack the default or be unmissably superior.