Are new anthropic rate limits due to a new training run launching?

I don’t think Chrome being sold would be bad for Google. It’s clear that it’s an outdated product but still produces so much money they don’t feel the need to really iterate over it. Perplexity and SearchGPT are dominating this knowledge search space at the moment and given the

Pretty good deal for $130 lifetime access to Gemini and 4o, wonder where your data’s doing tho. Why are there so many simple wrappers now?

Autoregressive model autoregresses.

Straight Lines on a graph => The cost of intelligence and autonomy continue to decline… Happy there are some new players like DeepSeek willing to disrupt the flow, I think 2025 could be a pretty crazy year

A flavour of what’s to come from their next gen more capable reasoning model with some emergent capabilities? Pre-release safety considerations from Anthropic is a good sign.

This is maybe the worst take I’ve ever seen - Someone who learnt to code in Python and had no idea what memory or pointers were until i took DSA in C

Not enough.

Oh no Dario, can you handle the recursive self improvement just to win the race that cannot be won?

I am disappointed.

LLMs will NEVER truly reason, because reasoning must be self-consistent, yet LLMs are made out of MATH and thus CANNOT be self-consistent

Think Grok can really crack continual learning if you just retrain every night. It wont truly be what Ilya wants of course, but it will be continual none the same. Elon Musk would love to get tailored news from things I select thats accurate, fix the notication 🙏

chris Each century, we are turning Earth more and more into a computer

A man cannot prompt the same model twice, for it is not the same model, and he is not the same man

NEVER DOOM.