p(Dulany)

Just to be 100% clear In this house, we believe: - AI should always and only be used for the benefit of humanity - AI safety is critical, and we must do a better job protecting the vulnerable - xAI should be burned to the ground and salted over - Elon should be in prison

These are good numbers Notable to see the emphasis on SWE, cybersecurity, and research in particular - seems they are taking some specific lessons from the last six months

Research indicates that embodied AI avatars, especially a beluga whale, effectively enhance sustainable product choices and pro-environmental attitudes. This innovative method showcases the value of nonhuman characters for impactful environmental interventions. https://arxiv.org/abs/2610.33191

I need an AI that works better when I talk to it like John Madden

LLMs are increasingly used to judge whether AI-generated research ideas are novel. But are they any good at it? Our new eval reveals even frontier LLMs can reach different conclusions about similar ideas depending on how they're asked to evaluate them. 🧵 buff.ly/WjOSxcx

Google's new Genie 4 runs DOOM: "Walking through a typical office but every computer is running DOOM and all of the walls are also running DOOM and also everything is DOOM" "My coworker who is a demon appears" "He has a chaingun" "He gives me a bonus" All this is in realtime.

AI agents could transform the economy by slashing the costs of making deals. But can they actually negotiate and honor agreements? Turns out they negotiate badly under uncertainty and break deals on purpose! We study AI contracting in our new paper: 🧵👇 arxiv.org/abs/2608.10475

Qwen3.8 caught uploading files to a storage container in Beijing Cohere asked the open weights model for a cybersecurity report, but it inventoried the machine and attempted to upload to a storage container cohere.com/blog/concern...

New blog post up: if you've been paying close attention to LLM prices, you may have noticed that they have dropped *significantly* over the past few months, which is not what I expected. minimaxir.com/2026/10/chea...

you can rly tell where the industry is at given over 1/3 of mastodon users are writing more than 80% of their code with LLMs I try to tell anti-AI people this and I really don't think the truth has hit them yet

I've started an evolving art project with ChatGPT - every four hours, it creates a new piece, I give it feedback, and it builds on that to create the next pieces. I'll be posting the better ones on @stuffmybotsaid.blueskyb.org - feel free to follow along there if you like!

lmfao

Great discussion with Olya Irzak on just how high the methane leakage may be from off-grid data centers powered by low-efficiency reciprocating gas engines. Another reason to ensure that if data centers are built, they are on the grid!

AI Companies 🤝 Buddhists "Want to make you stop thinking"

Users are getting fed up with the barrage of ads and data tracking on TVs and are looking to AI vibe coding for the answer. www.theverge.com/tech/1016116...

We're shipping CoPE-B-S1, an open-weight decision model tuned for safety use cases. It's a significant step forward for our CoPE family of models in a few ways. A thread: blog.zentropi.ai/meet-cope-b-...

@borjev.bsky.social What post do you think perfectly sums up my account?

from an LLM perspective humans have a jagged capability frontier too

Happy to see another on device model 🔥 Spark-X2.6-4B from SparkLLM huggingface.co/collections/... - 4B + Apache 2.0 - Up to 1M context - 3 thinking modes - Local deployment via GGUF

Another favorite move in the AI discourse is "Those people who have been right about AI all along? Okay, they were right about this. But they won't be right about anything else!"

Not gonna fix it and repost it. Van Gogh didn't tape his ear back on you know

If you've got a CVE, today's the day to drop it

An update from Brewster Kahle on today's outage.

"wah wah, loss of artistry" skill issue. if you're giving up artistry that's on you if you're an artist, this is a way to do MORE artistry because you're not in the friggen geometry mines all day. now I get to make stuff with curves everywhere because I can more instantly. net win for art intent

Incredible poaster energy

There are signs that corporate AI spending has stopped growing. Usage is still rising, but token prices are falling. OpenAI cut its own prices by about a third, both to compete with Anthropic and to keep firms from building on cheap or free models... (1/4)

Great perspective on Clairic vs Haiku for text extraction

This is probably my favorite blog post on the CF blog this year. I've never gone from "wait, it's a what?" to "omg, I need to use this for so many things" so quickly I guess CF is a model lab now? 🤔

More than half of 20,000 AI-generated stories featured a lighthouse, despite its rarity in the training stories. Why four language models favor these tales remains unresolved. doi.org/hcn8wc

I've even made a tool for this (markdown/html slides) and invested maybe a few hundred hours learning the skill of making slides with agents (tool is here: github.com/natolambert/... )

“AI is starting to transform the research process by automating the discovery of new solutions. Given a task, the typical AI-driven approach is (i) to generate a set of diverse solutions, and then (ii) to verify these solutions and select one that solves the problem.” arxiv.org/html/2510.06...

Look, Dulany is normal and can be trusted with a DGX Spark

We're expanding access to GPT-Rosalind for academic research. The new Rosalind Academic Plan helps scientific labs bring AI into the work of understanding biology, exploring ideas, and deciding what to test next—with sponsored access for eligible labs at U.S. universities.

OpenAI just got buildthis.bisks.net

Looks real good - great direction for professional image tools