It would also be useful for funding R&D into benchmarking models, which is currently mostly done by the labs themselves right now.

→ View original post on X — @emollick,