How far can reasoning models scale? | Dark Hacker News