Scaling linear algebra in the end is probably all we’ll need in the end. Only missing data and compute to get there