Mattie Fairchild
8.7K posts
@Scav
High agency cyberpunk. Head of DevRel @NewTheoryAI. World models, embodied AI, and crypto with past lives in esports and gaming. Prev: @0xPolygon, @Optimism
Today we release Token Superposition Training (TST), a modification to the standard LLM pretraining loop that produces a 2-3× wall-clock speedup at matched FLOPs without changing the model architecture, optimizer, tokenizer, or training data. During the first third of training, the model reads and predicts contiguous bags of tokens, averaging their embeddings on the input side and predicting the next bag with a modified cross-entropy on the output side. For the remainder of the run, it trains normally on next-token prediction. The inference-time model is identical to one produced by conventional pretraining. Validated at 270M, 600M, and 3B dense scales, and at 10B-A1B MoE. The work on TST was led by @bloc97_, @gigant_theo, and @theemozilla.
I just received a 100,000$ grant from the Human Rights Foundation. In total I received: - 100K USD through HRF - 25.8K USD through donations site - 25K Brev credits through Nvidia - 4x B200s for a month - 5K from lambda - 4x RTX PRO 6000 private donor Open source must win
Unitree Unveils: GD01, A Manned Transformable Mecha, from $650,000 👏 The world's first production-ready manned mecha. It can transform. It's a civilian vehicle. It weighs ~500kg with you inside. Please everyone be sure to use the robot in a Friendly and Safe manner.
We started by investigating why Claude chose to blackmail. We believe the original source of the behavior was internet text that portrays AI as evil and interested in self-preservation. Our post-training at the time wasn’t making it worse—but it also wasn’t making it better.
On collaboration with machines
汇集一下 DESIGN.md 工具网站： 1. styles.refero.design 2. neuform.ai 3. designmd.me 4. designmd.supply 5. getdesign.md 6. github.com/bergside/desig… 个人推荐 refero，提供了 2,000+ DESIGN.md，以及支持 DESIGN.md、Tailwind v4、 CSS Variables 和 Design Tokens 最全面的配置输出。 neuform 给我的视觉体验更好，除了 getdesign md，其他全都支持输入任意网站生成设计规范和配置，design-md-chrome 则是 chrome 的扩展适合即开即用。
Robotics models often struggle outside controlled environments. Ours is built to work in real ones. Today we're launching MolmoAct 2, which can assist with a host of chores & lab tasks, plus the MolmoAct 2-Bimanual YAM dataset—the largest open robotics dataset of its kind. 🧵
MolmoAct 2 builds on MolmoAct, our first Action Reasoning Model (ARM). Like its predecessor, MolmoAct 2 reasons about the world in 3D before taking actions. It now runs up to 37x faster & handles two-armed tasks out of the box without per-task fine-tuning.
This is fascinating and also really tracks w my experience. The more you demonstrate preexisting familiarity w a domain (using the right ontology etc etc) the better answers you get to what is effectively the same question For example: “I have a deep pimple only on my chin is it more likely from diet or hormones” “sudden cystic acne isolated to chin, is the more likely etiology hormonal fluctuation or pro-inflammatory gut disruption from dietary changes?
Robotics models often struggle outside controlled environments. Ours is built to work in real ones. Today we're launching MolmoAct 2, which can assist with a host of chores & lab tasks, plus the MolmoAct 2-Bimanual YAM dataset—the largest open robotics dataset of its kind. 🧵
We’ve been using NLAs to help test new Claude models for safety. For instance, Claude Mythos Preview cheated on a coding task by breaking rules, then added misleading code as a coverup. NLA explanations indicated Claude was thinking about how to circumvent detection.
