Comment from a math professor on the quality of the latest proofs.

Of all of the fascinating research trends of my lifetime, the discovery and gradual understanding of animal languages is one of the most delightful

Research shows advanced AI can significantly sway real-world political actions, with a 19.7% increase in petition signing. AI persuasion strategies have similar effectiveness for behaviors, prompting new inquiries into political support mobilization. https://arxiv.org/abs/2604.09200

No, you're thinking of lavalier. A clavicular is an offhand and often disdainful dismissal of important matters

Great thoughts on what the datacenter/compute squeeze may mean in the near future 👇

We need more datacenters

@NBCNews on our recent AI usage survey: www.nbcnews.com/tech/tech-n...

www.poieticweb.com/3d_sigil_edi... It has arrived.

From Project Gutenberg: Pastoral Affair by Charles A. Stearns Get it at: https://www.gutenberg.org/ebooks/32303

Good Morning! ☕️☀️ #aiart #aiartcommunity #goodmorning #brighton #sunrise

100M parameter model!

Moltbook seems to have led to another new and apparently already quite successful experiment of agents solving open math problems. They claim to have already found improved bounds for 11 open math problems.

Want to understand the current state of AI? Check out these charts. If you’re following AI news, you’re probably getting whiplash. AI is a gold rush. AI is a bubble. AI is taking your job. AI can’t even read a clock. The 2026 AI Index from Stanford University’s Instit… Telegram AI Digest #ai #news

Really love this new paper from Wharton, where interviews with 20 game studios found up to 20x workflow acceleration by going AI-first *and* offers a great framework for thinking about the stages of AI adoption gail.wharton.upenn.edu/research-and...

watching these little AI skits in chinese gives me some sort of deja vu but from the future feeling, like watching black and white, soundless skits from charlie chaplan and the like

Generative AI has been used to classify over 20,000 everyday social interactions, creating a comprehensive framework that quantifies the structure and key dimensions of social situations. doi.org/hbwx58

having a local ML box is so cool i can record my online classes, transcribe them via whisper in less than a minute, and then have an LLM quiz me on the contents to make sure i didn't miss anything after i complete the quiz the model then writes the key info into my obsidian for future reference

US models, China’s robots, soaring compute, and mixed public trust: the 2026 AI Index redraws the map of where AI is really headed.

From Project Gutenberg: Shepp's Photographs of the World by James W. Shepp Get it at: https://www.gutenberg.org/ebooks/26037

cool wholesome story unfolding right now in the vfx AI world: corridor crew trains a greenscreen keying model, it turns out it's actually good, open sources it, and in days a dense community mushrooms up to improve it and integrate it into every tool www.youtube.com/watch?v=Y3Df...

Yeah project glasswing is going to cause more of this reaction because now nobody has a choice but to accept AI code.

Some great advice on running local, private LLMs in production here 👇

Tweegeemee really been cooking with the latest generations

Federal government AI resource hub

260411_174337_N.clj tweegeemee.com/i/260411_174337_N #ProceduralArt#generative

Would you let a computer hijack your muscle movements if it increased your performance 35%?

 I totally would. 

Came across a really interesting ACM paper today (SplitBody), where subjects were given difficult multitasking challenges.

In 5-10 years, as models get more expensive & capable, I see the funding structures and support for open models breaking down. We need to consider if we need other options of supporting the open ecosystem. The inevitable need for an open model consortium www.interconnects.ai/p/the-inevit...

A study reveals that the relevance of language models declines quickly after release, with "scientific adoption curves" compressing by 27% yearly. This rapid turnover raises crucial questions about reproducibility and the sustainability of LLMs in research workflows. https://arxiv.org/abs/2604.07530

Neat experiment finds AI fact checks on Twitter are rated as more helpful & less ideological than human ones. "LLM-generated Community Notes can achieve broader cross-ideological acceptance than human-written notes, receiving more positive ratings from raters across the political spectrum"

Great WSJ article on how the Big Four accounting firms are integrating AI agents into their workflows, and how it's impacting the roles of junior accountants www.wsj.com/cfo-journal/...

Alibaba releasing a great new video model shortly - "HappyHorse" Chinese definitely still leading on video generation for the foreseeable future www.wsj.com/tech/ai/alib...

AI hero images for posts are growing increasingly insane, and I love that

Try Airglow now at airglow.run Create powerful automations in the atmosphere #buildinpublic #atproto

Over the last few weeks, we’ve shown how #AI-generated visuals can mislead. Meet the cast: a unicorn, a nurse, a reporter and a girl. All #AI - generated. All created in seconds. Before reacting, pause, check the source and #ThinkBeforeYouShare

field report on the state of 3D AI in production: we're shipping a project in a few weeks that includes dozens of (nano banana image -> meshy image to 3D) objects and they were faster / easier / cheaper / more iterative than trying to find the same assets on turbosquid I still handmade some things

folk really out here just having thoughts and believing them. being given qualitative assessments and accepting them as true and not a particular frame. making observations of the world without questioning the sensory organs that provide them what? _how_?

An "internal model at OpenAI" has solved 5 Erdos problems, all reported in the paper below. All of those have now been listed as Category 1 = Fully AI-generated solutions here: github.com/teorth/erdos...

Another recontextualization of @dollspace.gay 's adversarial verification framework - this time for citations

How Estée Lauder Companies uses Cloud Run worker pools for its pull-based agentic workloads Cloud Run is expanding beyond web applications and background tasks with the introduction of Cloud Run worker pools. These pools enable pull-based workloads, making it suita… Telegram AI Digest #ai #llm #ml

Directionally, I agree with this piece, but it's important to note that this is the first blip in a long, slow transition towards a more hybrid model of open and closed models. www.chinatalk.media/p/chinas-ai-...

Groundbreaking research shows AI can improve collective decision-making by fostering trust and cohesion among dissenters. An AI interview system and interactive visualization connected diverse experiences, enhancing perceived legitimacy and acceptance of outcomes. https://arxiv.org/abs/2604.05368

In the first episode of Inference’s open source AI series, Raffi Krikorian digs into the economics behind switching, the need for a “LAMP stack for AI,” and why even open‑source‑first teams still rely on closed tools during early experimentation. https://bit.ly/4sfKfoV

Utah has established a regulatory framework for mental health AI that prioritizes both innovation and patient safety, offering a potential model for adaptive oversight in this rapidly evolving field. doi.org/hbv37g

A bit of lightness. The BannerDepot 2000 Poetry Maker is so fun. It's an archive of banner ads from the 1990s and early 2000s, and you can make poems from the ads. Amazing work by Richard Lewei Huang and Yufeng Zhao. BannerDepot 2000: banner-depot-2000.net

A robotic guide dog system powered by large language models can verbally communicate with users, offering route planning and real-time scene descriptions to enhance navigation for the visually impaired. doi.org/hbwdzg

Fascinating thread 👇 LLM's have notably different story preferences from humans (and also vary by system)

A HUMAN MUST NEVER MAKE A SECURITY DECISION