Hacker News Reader

They would display numbers less than 1000 as-is, and only start adding the "k" and "m" prefix after the 4-digit and 7-digit threshold was crossed.

But how could they maintain an accurate count? Maybe they were just persisting the user-friendly format alongside the actual count...

The endpoint of the API is probably just rounding the accurate number and returning a friendly number... or it's all bullshit anyway.