Christian Shewmake

🎉Our NeurReps 2023 Call for Papers just dropped! 🎉 Come share your work with the community at NeurIPS this year! If you're curious about math, minds, and machines and interested in getting involved, let's chat! 🧠

🦾🔥Super exciting event set up in our sibling-space in the mission, realllllly looking forward to seeing what we can get up to with this hardware and compute 🚬🤖

@MissionRobotBay let's see if @charles_irl makes an appearance 🧠 💚

Announcing: The Embodied Metal Hackathon! Happening Jul 17-19 in San Francisco's Mission. Three days in a factory loft with real robots, hacking to make them do something wild. We're packing the floor with the best founders, developers, and mad scientists across Embodied AI. If you'd rather make your AI *do* things – in the real world! – instead of just write about them, this event is for you. Hosts and sponsors: @MissionRobotBay, @savant_vc, @newtheoryai, and @NorthstarRobots, with compute from @modal. Space very limited, so apply early: luma.com/embodied-metal

San Francisco! I'm hosting a robotics hackathon for the most badass physical AI founders in the bay area. Want to hack a metal arm to do something weird? Every team will have one to jam with. Fancy a humanoid? We've got one. There's probably some awesome stuff you can do with the hands we have lying around. - ML/AI specialist and wish you could stick your neural nets inside a body? Come hack. - Stablemaster of the largest herd of coding agents this side of the Mississippi? Point your harness at our robots. - Can what you make be qualified as 'mad science'? I want to see it. - Physical AI founder who wants to put your product to the test on a new embodiment? I promise to help break it. We're putting special work into onboarding, a bootcamp, day-of support, and more than a few surprises. Most importantly, you'll be hacking alongside leading builders in Embodied AI. Special thanks to @savant_vc for manifesting the idea, to @modal for providing compute, to @NorthstarRobots for their wells of data, and @newtheoryai for the space, chalkboards, and stack. Space is super limited. My attention can be caught by engaging with this here tweet. I can't guarantee you'll get in if you comment something brilliant, but it will help your chances.

Today @cusp_ai and @KemiraGroup announce a milestone in AI-driven materials discovery. We have used generative AI to design new materials targeting PFAS removal from drinking and process water at trace concentrations.

let's bring it back!

My app that helps students think more carefully through their projects now looks really pretty by 2026 standards (imho). Try it out at planyourscience.com

CIMC.ai is having an A-Life Hackathon next week Wednesday afternoon, in San Francisco. Join us if you are working on emergent behavior and A-Life and would like to show off your work and ideas! luma.com/b0zikaeq

@wellingmax @mmbronstein @AlexanderTong7 @erikjbekkers A memorable final quote from @AlexanderTong7 (echoed by several of the panelists): "geometry is about understanding the world, and it is here to stay."

These Dover math book covers are so beautiful.

The PhD urge to renounce to money, youth, friends and socially isolate yourself to become cracked at an esoteric subject

Modern deep learning is very complex. We are hopeful that many lines of research will improve our scientific understanding of DL through the lens of learning mechanics. Hope to find more solveable models & limits, empirical laws, scaling arguments, universal phenomena, etc

full speed ahead @learning_mech and @blake__bordelon!

ICLR 2026 is almost there! We have 6 exciting keynotes covering a range of areas from machine learning to robotics, neuroscience and AI for science: Maja Matarić, Max Welling, Percy Liang, Katie Bouman, Karen Adolph, Pablo Arbeláez See you all soon! #ICLR2026

@getreddy4this "Frontier" is largely a misnomer for what's happening now. There is a misalignment of strengths. Great scientists are great because they can ask the important questions that no one else is asking. There are many many more people who can proficiently execute an incremental agenda.

There's a broadly held misconception in AI that methods that scale well are simple methods -- even, that simple methods usually scale. This is completely wrong. Pretty much none of the truly simple methods in ML scale well. SVM, kNN, random forests are some of the simplest methods out there, and they don't scale at all. Meanwhile "train a transformer via backprop and gradient descent" is a very high-entropy method, easily 10x more complex than random forest fitting. But it scales very well. Further, given a simple method that doesn't scale, it is usually the case that you alter it to make it scale by adding a lot of complication. For instance, take a simple a simple combinatorial search-based method (not scalable at all) -- you can make it scale by adding deep learning guidance (which blows up complexity). Scalability usually belongs to high-entropy, complex systems.

Job news! And heck, totally new adventure news. I've joined @newtheoryai as Head of Developer Relations to bring 'world models' into being. Literally, into beings. World models are an emerging flavor of AI based on state, physics, and causality instead of just words and pixels. This is important for anything that depends on moving around in and interacting with the real world. Which means ROBOTS. Lots of robots. But also "embodiment": any being, process, or robot through which AI pushes into the real, physical world. A great model means simulations that can run the next five seconds ten thousand times, and pick the best option. It means being able to run reality in headless mode. It means being able to pair a fast, intuitive brain with the logical language-based brains LLMs offer. And holy smokes, I'm excited. I met @JonathanHillis and @cashewmake2 at the same time I was already plunging down the world model rabbit hole, and knew from the first meeting I had to join. The rate of nerd snipes per minute in this office is incredible. My fundamental thesis is we're in a period of exponential growth in AI and robotics. This sounds obvious, but most people aren't making decisions as though that's true. Things that seem five years away are actually one year away. Things that seem six months away might suddenly be walking down your street tomorrow. You can build a robot *right now* for less than $500, and that number's still dropping. With Claude Code, anyone remotely technical is about to discover they can manufacture and program their own mechanical familiars. The ecosystem isn't ready for that flood of people yet. 'Robotics' still sounds academic, or like you need an electrical engineering degree. 'Embodied AI' is better, but esoteric as hell. My job is to get the ecosystem ready and push these new abilities into as many hands as possible. That includes New Theory's world model we're building at New Theory, but also the tools and actual robots across the ecosystem. If you're in robotics, embodied AI, and world models, I want to talk to you -- especially if you're putting real robots to work. DMs are open 🦾💃

I wrote a parasocial love letter to Nara Smith, the model-tradwife-influencer extraordinaire. My feed has become dominated by her videos: the elaborate domestic rituals where she dons couture gowns and cooks everything from scratch, from bread to cheese to facial moisturizer. open.substack.com/pub/everydayth…