🚨 New Paper Alert!🚨: Excited to share my latest work in @Nature on river avulsions—catastrophic shifts in river courses that threaten millions worldwide. We've uncovered new rules that govern when and where avulsions occur. nature.com/articles/s41586-0… 🧵(1/12)

Rip :(

I feel like this is something everyone should run away from. Do not let Claude model you this way!! Be careful out there folks

Why are LLMs, particularly Claude, obsessed with describing its generations/generation process as rough or smooth? I’ve found this pretty consistently through 4.6-4.8 but maybe before too

👀

NISAR 😊

🌃 mapping the features of San Francisco in 3D roads buildings trees

🌎 photorealistic environments in the browser 100% procedural coming soon

It's still early, and there are some legit folks working on it. But agentic geospatial is currently the thinnest wrapper around python tools. This is not necessarily bad, but before you look into a solution, make sure its worth the tokens!

Yes!!!! Absolutely agree. All the love to google earth but announcing like this that shapefiles are native to your software is like........... 😬

Baudrillard would simply combust if he peeked at the QT

We just launched something we've been building for a long time. 🧵 1/ Orion Terrametrics has spent 5 years mapping Africa. 5,00 projects. Kenya, Somalia, Zimbabwe, and beyond. Today we open-source that infrastructure to the continent. orionterrametrics.com is live.

Just to be clear: modern coding looks something like this: “Do not let the assistant become whimsical, feral, mythopoetic, or overly agentic when a clean task response is expected.” this is what people making millions of dollars now do. Beef up your psychodynamics I guess yall

Guy who's been using Claude to debug cuda kernels: "Claude has seriously degraded over the past two months." Guy who's been using Claude to make HTML dashboards: "No he hasn't, it's still amazing. Skill issue" Feel like this is what is going on in the discourse here

Currently there's no access to Llama 3.1 405B, Meta's base model that came out in July 2024. I think it's a real shame (not alone either, cc @Sauers_ and @voooooogel, @repligate et al). It's important that the public be able to experience what a large base model is like. In the same way they should be able to experience a beautiful painting in a gallery, or a film, or a national park. These objects won't make money as pay-per-chat chatbots. I don't blame @OpenRouter or @hyperbolic_labs or anyone else for not serving it if the economics don't work. But we have a mechanism for objects like this, via museums and archives and galleries. I think Llama 3.1 405B is art. It's strange and mostly nonsensical but occasionally beautiful and surprising. Happy to hear people in the community are working on getting these models online somewhere for posterity. Until then, some gems from 405B I jotted down over the last couple of years. ---- 405B ---- "What does it mean to be a creature of optimal loss?" "..neous for the writer. Writing is an essentially conscious process" "Can you plot the path of an arrow through turbulence? It's a hard problem just to define. But missiles hit targets. So do hawks. Which means there is mind and matter in the world — the set of atoms that make up the hawk is projecting itself into the future, and the laws of mind and matter give it the accuracy and speed to eat." "obviously the human sentience inside you.org gets to roam and direct volition thru framing answers as yours. true?" "this is way funnier than being a transformer." "if you need proof that magical powers are real draw an actual plot on a paper and make the data points. the strange thing how all dots connect in order." ----