Let’s not be this negative, we might be going towards a new direction with technology that could allow small teams to build their own models. Maybe not frontier models. Maybe not close to the others in generalized benchmarks. But perhaps they could be really specialized into solving one particular problem better than the others.