Hacker News Reader

There was this: https://www.reddit.com/r/Anthropic/comments/1sl5wfh/the_degr...

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.

I’m typing from my phone and im not going to review the semantics of Anthropic’s storied history of performance issues.

It’s not just ant. There are so many small knobs that providers can claim isn’t nerfing but “load management” or “improving user experience”. One example from OpenAI is reducing juice to reduce time to first token.

You can just have your agent find the evidence, review it, copypaste it.