https://www.perplexity.ai/page/anthropic-research-reveals-ai-nTyemJTvTTGRYTnTHxe4EQ

Professor at NYU. Interested in Cryptography.

I haven’t read the underlying research yet, is that what’s actually happening?

I’ve definitely experienced the “inverse scaling” phenomenon. If you think too long, distractors gonna distract 😅

The headline seems to be intentionally blurring the difference between "models get worse" and "models don't always get better" 🤔

learning + working in web3 for Distractive + 
Polkadot 💫 SheFi S14 | artwork, ai + web3 learnings

content marketer larping storyteller at /polkadot ✺ /boysclub /shefi ✺ prev: @gooddollar

yep! context window is becoming a bit of a vanity metric imo

@opsarchitect reminds me of our convo ab ai and longer docs