A historical lesson, drawn from Heroku's loss to Kubernetes despite Kubernetes having markedly worse UX: Kubernetes won because it preserved customer choice and avoided vendor lock-in, while Heroku's superior experience came bundled with platform lock-in. The claim is that avoiding lock-in can beat a better user experience once the stakes — infrastructure dependence — are high enough.
Used as a template for predicting how the current multi-lab AI competitive landscape will play out if the current fast-moving race stalls: expect the ecosystem that best avoids lock-in to win out over the one with the best immediate experience, echoing Owning Core Infrastructure as Competitive Moat concerns from the buyer's side rather than the builder's.