Hacker News Reader

I mean there is a direct link two comments down from here from 30 minutes before your comment: https://news.ycombinator.com/item?id=49902477

That’s NOT Anthropic admitting to “nerfing” their model as claimed above (which implies intent), that’s a regression which they quickly fixed.

Christ this forum has become intellectually dishonest.

There has always been plenty of intellectual dishonesty here, but remember Hanlon's Razor.

Er, perhaps take a look after the edit:

> There are recorded cases of real regressions, but they're better characterised as incidents, not nerfs

I dont really characterize a nerf as always by intent.

So I found this incident, as they called it, to still be relevant and why benchmarks such as the OP are useful.