Gemini 2.0 may be the biggest, strongest model yet, and we need it to prove that even with orders of magnitude of scaling, the returns are significant enough to allocate sufficient compute. Will it be able to do that?