Comment by randomizedalgs

After skimming some of the writeups, I'm surprised that the frontier internal model still writes just as poorly as Sol.

Maybe good AI paper writing is further away than I thought...