So in the next year or two, all the frontier labs will have to pay content owners (and Cloudflare) billions to scrape data. All labs except Google… $NET $GOOGL

So if I am reading this from Ben Thompson correctly, then Cloudflare just default blocked 20% of the internet from every LLM-provider but it remains available to Google for AI Mode and AI Overviews Seems like a huge competitive advantage for $GOOGL

Nutty pudding finally stops tasting like dirt after a few weeks. My breakfast

Finally found this! Triple Whale 🐳

Why Vinay Prasad was fired from FDA -> Sarepta spends years making drug -> drug is really awful, no benefit -> studies are so cooked and terrible FDA regulators are horrified and say studies are so bad -> regulators overruled -> drug gets approved -> $3.2 million/injection ->

If germline editing ever does somehow pass the ethics & safety test, pharma co's will find a way to kill it Harder to charge insurance $10M for an embryo that isn't even a "human" yet

why is cursor suddenly so bad? did anthropic change sonnet 4?

So what is Anthropic's real moat here? Any time they release a new model, OSS labs will come and get something that is close, yet maybe 50%+ smaller and OOMs cheaper. Eventually, when they face pressures to move towards profitability, they will have about 3-6 months to make

It should be clear by now that OpenAI's strategy is to hold on to releases until a Google release date, and then drop it at the same time to take the wind out of their sails. Very adversarial release schedule.

NOVO NORDISK SUES TELEHEALTH FIRMS FOR ALLEGEDLY STEERING PATIENTS TO KNOCKOFF COMPOUNDED “SEMAGLUTIDE”

Coolest thing I’ve seen since winter

they're selling dollars for thirty cents again!

So GPT-5 is only +0.4% on SWE bench verified vs Opus 4.1 which was probably cooked up since Monday

What did I say

Cutoff is June 2024? Hmm

How can they just enforce this? Does this affect Ash - AI Therapy and others?

We were able to reproduce the strong findings of the HRM paper on ARC-AGI-1. Further, we ran a series of ablation experiments to get to the bottom of what's behind it. Key findings: 1. The HRM model architecture itself (the centerpiece of the paper) is not an important factor.

Everyone debates which AI model is "best." Now you can bet on it. Thanks to Ben Raab for the feature in a piece on AI prediction markets 👇 wsj.com/tech/ai/top-ai…