Hacker News Reader

Thank you!

I’m incredibly skeptical that OpenAI is spinning up custom ASICs for improved inference performance, but they never thought of optimizing KV cache until a tiny Chinese lab did it? Give me a break.

View on HN

They certainly thought. But were they able to do it now without DeepSeek papers?

timeline suggests not.

So DS comes up with 437x improvement, and the evidence that O/A copied them is that they dropped caching prices by like 50%? Really?

What evidence evidence would actually change your stance on this? a formal admission from OpenAI and Anthropic. Yet that's unlikely

Any actual evidence would be worth considering. This is a pretty weak correlation and an assumption, nothing more.

80%. You can't do that unless you significantly cut costs somewhere. The timing is certainly indicative.

Reply on news.ycombinator.com