Knowledge catalogue
this would be more credible if Sam had left OpenAI when the board fired him. 🚨 Sam Altman fires shots at Anthropic >“ai becomes too powerful we can’t control it” >“power gets too centralized… one comp
arXiv:2510.01192v2 Announce Type: replace-cross Abstract: Emulating empathy in human-robot interaction is a key component for achieving satisfying social, trustworthy, and ethical robot interaction wi
In any organizational behavior class, students will learn that effective delegation is among the most important skills for a seasoned leader. Getting meaningful work done involves careful coordination
So what’s actually the best local AI harness rn? I’ve read a TON about this already and somehow ended up more confused than when I started so I figured screw it, let the community decide. Right now I
🦔OpenAI's CFO held an all-hands meeting today to tell employees the company will go public by 2027 and not to worry about Anthropic beating them to it. This came days after the CRO, COO, and product c
arXiv:2608.16893v1 Announce Type: cross Abstract: Expert surveys are widely used in security research to study practitioner workows and decision-making, yet recruiting domain experts - especially in S
Today we release my favorite episode of Training Data yet: the great Rich Sutton. @RichardSSutton wrote the textbook, wrote The Bitter Lesson (and many other on-point essays like 'Self-Verification, T
Hi everyone, I use AI agents to code and manage my personal files heavily. However, I am worried that the agents may accidentally delete some important files on my machine, outside of the defined scop
arXiv:2608.14391v1 Announce Type: cross Abstract: Recent video generators can fabricate realistic depictions of wars, disasters, public emergencies, and other real-world crises, creating substantial r
the crux of why AI safety in systems that are built around LLMs is never going to work: they just can’t be trusted to follow instructions. Claude Code w/ Fable 5 has strict instructions in my project
Here are the gems the AI picked: 1. Sperm whales may have collectively learned how to evade whalers. Within a few years, the success rate of whalers’ harpoons reportedly fell by 58%, suggesting that i
I do agree that the public has a negative view of AI (and that this is a big problem), but I don’t think it is primarily caused by me or any other AI leader warning about AI’s risks. I think it is fun
Quick update on the RTX 5060 Ti local LLM repo. It has changed quite a bit since my previous posts. The project started as a collection of practical notes and benchmark results. That was useful, but a
Ollama makes it easy to run a model locally, but the hard part for me starts after the model responds: what keeps an agent from having unclear tool permissions, unbounded loops, silent partial failure
Qwen 3.8 27B is now available on Ollama. It's one of the best open models at this size, and made for agentic tasks and professional work. Try it directly with the apps & harnesses you use: Claude Code
I've been an Ollama Cloud subscriber for ~6 months. I generally use the the latest GLM models available for coding as well as a personal instance of Open WebUI. I have tried to get answers directly vi
arXiv:2608.11645v1 Announce Type: new Abstract: Volumetric video streaming turns privacy into a 3D, multi-view problem. Unlike ordinary video, where sensitive content can often be redacted frame by fr
I’m genuinely pretty disappointed with how my Ollama Pro annual subscription turned out. For context, I’m a technology editor and developer, and I had actually used Ollama Pro before. My previous expe
arXiv:2608.10773v1 Announce Type: cross Abstract: The increasing use of Generative Artificial Intelligence (GenAI) in journalism raises concerns about possible detrimental effects both on journalism a
Open source is so back. Zuck just announced Meta is opening the weights for Muse Glimmer, with Muse Spark 1.2 coming soon But a year ago, everyone doubted Meta's position in the AI race In an intervie
We converted the model from the original safetensors and found two issues. The first one made our quantization fail several times, the second one does not fail at all, it just quietly ruins the base 1
Wowee!! Just when you thought it couldn't get better for open weight models, we probably have had our best period yet!?!?! Models that rival the closed frontier, Opus level models on non-insane hardwa
MiniMax H3 loads a 32B text encoder, 15.7 GB, just to turn your prompt into a conditioning tensor. I replaced it with a Qwen3-VL 4B or 8B plus a learned map into the same space. Same DiT, same VAEs, s
I discovered recently that Google uses their own TPUs, like tiny ASIC cards like the toy ones that existed for bitcoin. And while it sounds inefficient the fact they use thousands of them because...th
anyone confused about neurosymbolic AI—and its recent enormous victory—should read this. complete and total vindication for what I have been saying here all along. (see also my essays on Claude Code a
So I can either pull the trigger on a 128gb AI max+ 395 laptop or wait for RTX Spark for LLMs. Maybe I get it now and the price of the spark is super high so it's a good purchase or maybe the Spark sh
OMG i wrote some of the original work on what is to be neurosymbolic in 2001 and dude who probably hasn’t read that work is trying to school me on the definition 🤦♂️ coding harness and tools calls ar
interesting (and completely opposed to @haider1’s take on the same graph) *the labs themselves* have only modestly changed their predictions on AGI timelines over the last decade. note also that (on a
Hey everyone! Scenema Audio is now a native ComfyUI custom node. Same model that powers scenema.ai now quantized so it fits on 8GB VRAM. When we first released it a few months ago as an API and Docker
Hear me out before you roll your eyes. We tend to only recognize turning points in hindsight. Nobody in 1993 thought the Mosaic browser would be a history book moment, but the web is. I think Stable D
The bottleneck for AI progress was never compute, it was always the verifier. Recursive self-improvement is limited by verification, not computation. Compute buys proposals - verifiers buy knowledge.
I've been out of these forums for awhile but I figured I would provide a formal update on how this has been going now that it has some operation time under its belt, just to put the information out th
According to rumors (spread by myself) Le Chaton FAT will be 26T-a3b and I AM READY for it. Let's be real, I can't afford that many 5060Ti, so I got 12x Gen 4 3.2 TB (two per card). This gives me abou
Checkmate: you can’t take the harness (which is typically in large part symbolic) away from the neural model without giving up performance. HUGE victory for neurosymbolic AI, straight from @AnthropicA
The egg test for AI 'I hereby offer @elonmusk a million dollar bet against his prediction that Optimus will be better than the best humans in surgery by the end of the decade.' ~ @GaryMarcus Haha. I'm
TLDR: There is no one prompting trick that will result in Krea 2 Turbo giving you exactly the style you want and across the whole image. Instead, if you are trying to achieve styles without the use of
Been trying to find something that actually handles my workload well instead of just being 'fine.' Started on Qwen 2.5 7B, moved to Qwen 3 8B, and right now I'm using Nemotron 3 Ultra (the big 550B on
arXiv:2607.26899v1 Announce Type: cross Abstract: Diverse human groups produce diverse ideas, the raw material of innovation. Generative AI challenges this engine twice over: everyday AI assistance ma
After trying several AI headshot apps, I decided that they all suck. I've always liked Marco Grob's TIME cover portraits so I've been refining some prompts to generate these types of headshots. Both b
I've spent the last couple of days trying different setups (Ollama, Continue, Claude Code, Gemini CLI, OpenRouter...) and at this point I feel like I've spent more time configuring tools than actually
Hello, I have a friend who's using various local LLM's like qwen3.6 27B, 35b-a3b, North Mini Code, and qwen2.5-vl-7b (just for vision). They have a use case where they're trying to have an LLM drive a
arXiv:2607.26348v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly used as synthetic users, stand-ins for human respondents whose simulated answers feed product, policy, and
Been running a GPU price tracker (https://www.pricesquirrel.com) since March, covering 20+ EU stores, recently added RAM, SSDs and CPUs too. Every GPU tier has gotten cheaper since launch. The RTX 509
arXiv:2607.22702v1 Announce Type: new Abstract: Text-motion representation learning has advanced rapidly, with growing interest in multi person interactions for animation, AR/VR, and embodied AI. Thes
arXiv:2607.21632v1 Announce Type: new Abstract: Traditional benchmarks for LLMs primarily rely on static datasets and objective scoring metrics, which often fail to capture differences in response qua
arXiv:2607.22241v1 Announce Type: new Abstract: Recent advances in video diffusion models have spurred interest in human-object interaction (HOI) video generation, which demands fine-grained control o
Recent discussions about open-sourcing make me feel that I should go back and revisit these important open-source works in representation learning that pushed the field forward and eventually made vis
everyone is talking about open-weight models. at @huggingface we're invested in open source AI. like, a lot. and more every year. i pulled the public github numbers for our orgs. good excuse to introd
arXiv:2607.21368v1 Announce Type: new Abstract: Multi-object tracking in dense crowds requires solving a bipartite assignment problem between detections and trajectories at every video frame. The clas
This is rather counterintuitive, since banning Chinese models would benefit them the most. Jensen Huang is also against the ban, although the interests here are more obvious. Do you think that if Arce
On July 23, 2026, a Polymarket tweet announced that Anthropic donated $20 million to a political nonprofit calling for stricter AI regulation ahead of the U.S. midterm elections. The post, posted at 4
This story is wild. The short version: OpenAI were running a cybersecurity test against an unreleased model, with the model's guardrail features turned off. Rather than solve the test, the model broke
AI models pushing the frontier are a growing challenge for cybersecurity. A few weeks ago, I asked Demis what's underhyped in AI right now and on his mind: 'I'm very excited about this new agentic era
Disclosure: I work at Pluralis Research, the lab that built this. Code is open, and I'm happy to answer questions. TL;DR: As far as we can tell, this is the first RL post-training run whose entire rol
Something I have been thinking about: in the past, the best engineers I knew spent a lot of time automating their work in various ways. Better vim/emacs automations, writing lint rules to catch repeat
Thinking Machines Lab: Thinking Machines Lab debuts Inkling, an open-weight MoE model with 975B total and 41B active parameters, trained to be broad rather than optimized for one area — Try on Tinker
This AI recursive self improvement (RSI) paper shows no sign of fast takeoff. The AI model advances at about [Intelligence]^0.075, or the the 13th root of input intelligence [1,2]. That means the inte
On July 14, 2026, OpenAI CEO Sam Altman announced a 2.5‑fold increase in usage of the company’s agentic products—Codex and ChatGPT‑based tools—within the preceding week. The tweet, which received 703.
hello! Hello. We have reached 8M active users across Codex and ChatGPT Work. We are once again resetting the usage limits for all. And we continue to not have the 5h rate limit as well, allowing every
Creating my own interfaces in real-time to do the stuff I want to do is ridiculously empowering. Here's an example: - took ~10 app screenshots of things I wanted to fix - asked Claude to make a feedba