I am using Hermes from @NousResearch and GPT-6-Sol to vibe-train a @pollenrobotics Microduck to Moonwalk in sim. It isn’t there yet… but the process has been incredible. AI is about to give us a lot of much more general robots I think.
Earlier this year, i wrote about startups including @Letta_AI and @LangChain exploring new ways for AI models to consolidate memories through something reminiscent of dreaming. Anthropic now says that Claude does something similar wired.com/story/sleeptime-co…
Memory is what allows Claude to learn from experience. In a new Engineering Blog post, we share how “dreaming” helps Claude organize what it’s learned, correct its mistakes, and find what it needs more easily the next time. anthropic.com/research/compo…
Will Knight retweeted
AI agents can now break into realistic enterprise networks on their own. A capability once limited to trusted-access programs is available to anyone.
Today we’re releasing Vals CyberOffense.
@willknight from @WIRED wrote an article about our work, link in the thread.
I'm building my own GTA-style game featuring babies and robots. the roofs are upside down but you can't have it all.
Will Knight retweeted
> be AI company
> scrape the entire internet
> ignore licenses and copyright
> download millions of pirated books
> use them to train frontier models
> spend tens of millions buying physical books
> cut off their spines
> scan every page
> destroy the originals
> lecture everyone else about intellectual property
> we are the good guys
We added two sentences to Claude Code's prompt on Opus 5.5 to steer it away from recreating someone else's specific game because using Claude to infringe IP is against our usage policy.
Original games, genre games, and games built from your own designs and assets are meant to keep working.
We'll continue to iterate to ensure we're striking the right balance. In the meantime, if Claude turns down something that's yours, send it my way.
this work seems incredible
Back in ~Dec 2025, I really wanted to work on mobile manipulation, but robots were either too expensive or not performant enough.
So we built our own!
Excited to finally share something I’ve had so much fun working on:
1. Pixel: a co-designed full stack system for learning mobile manipulation.
2. Pixel-UMI: environment aware whole body manipulation without robot data.
w/ @JasonJZLiu
(1/N)
Someone got the fly connectome to control a Microduck's perception in sim. fun idea, and would be fascinating to see how one navigates using it.
huggingface.co/spaces/pollen…
Incredible flex from the OG
This is an old hat. That's exactly what the neural fast weight programmers did in the 1990s (including the 1991 unnormalised linear Transformer). Overview with links to the original papers: people.idsia.ch/~juergen/fas…
I'm headed to the Wired World Fair in Miami this week. Hit me up if you are there too!
wired.com/wiredworldfair/
In this week's AI Lab: To explore AI’s predictive skills, I let Codex loose on a fantasy team. It tried to bench Erling Haaland at first, but it soon learned the error of its ways.
Also: Rogue crypto agents, shutdown requests, and voice-based agent discrimination.
wired.com/tag/ai-lab/
Will Knight retweeted
The chances of being killed by a landmine delivered by a bear are low, but they are never zero.
this is wild.
I've always believed that intelligence should come from simple, general building blocks that learn their own abstractions directly from raw data; in other words, the bitter lesson. I think we're finally seeing a real step towards AI that can learn from all data.
Language modeling on raw bytes is a microcosm of this ideal, and to me it's a central problem in AI: if we can't build LLMs that learn end-to-end from bytes, how will we handle the rest of the world's data that we understand even less?
Last year we introduced hierarchical networks (H-Nets) as a first step. Our new version finds up to 3-5x efficiency gains in controlled scaling law experiments, and at larger scale matches an open-source model trained on 11T tokens with ~10x less data and compute. The posts below walk through what we found and why we think it works.
Will Knight retweeted
me: kinda fucked up how many people are offloading their whole lives to muse or instinct or grokbot or whatever
claude: yeah
have we truly explored the combat potential of cybernetic pugs? their low center of gravity would surely be an advantage in paw-to-paw combat.
oh hell yes. this is EXACTLY what I want my agentic mega-swarm to look like.
1/14) Virtual animals are a long-standing goal in neuro-AI, but they should actually behave like real animals.
We present SCAMPER: a virtual rat for RL studies that moves naturally during both exploration, and goal-directed behavior.
Project page: aidansirbu.com/SCAMPER-site/
I initially thought this was AI-generated (there's A LOT of fake humanoid stuff out there) but it seems not. It's incredible how powerful the robot seems to be. It's also wild how rapidly @UnitreeRobotics is developing these systems.
