Article Archive

58 articles in total

2026

September

• 09-25 Running the Release State Machine: A Pure-stdlib Minimal Closed Loop and Seventeen Invariants
• 09-22 Deployment and Ops for Embodied AI: Swapping a Component Is Not Releasing a Version, It Is Feeding a State Machine
• 09-19 Architecture Is the Design of Seams: An Embodied Agent Skeleton That Runs, Tests, and Swaps Components
• 09-16 Policy-Side Evaluation (Part 2): How to Test, How to Train, How to Land — Four Compliance Evidence Types and a Minimal Executable Interface
• 09-15 Policy-Side Interface (Part 1): After the Contract Stands, What Do VLA / Diffusion Policy / π0 Actually Consume?
• 09-14 Stacking Sensors Is Not Fusing Them: Multimodal Robotics Lacks an Interface, Not a Model
• 09-13 All Eyes, No Fingertips: Why General Robots Still Lack a Sense of Touch
• 09-12 Embodied AI Sim-to-Real Methodology (III): Evaluation, Decision Matrix, and a Minimum Executable Protocol
• 09-11 Embodied AI Sim-to-Real Methodology (II): Four Intervention Lenses and Two Reformulation Routes
• 09-10 Embodied AI Sim-to-Real Methodology (I): Treating Sim-to-Real as an Error-Budget Allocation
• 09-09 Robot Data Scaling: From Interaction Coverage to Marginal Data Value
• 09-08 The Data Landscape of Embodied AI (Part 1): Sources, Interfaces, Distribution, and Training Recipes
• 09-07 VLA Deep Dive (Part 3): VLA and World Models, Open Questions, and Three Judgments
• 09-06 Embodied AI Roadmap 2026: From VLA to World Models — Who Is Solving What?
• 09-05 VLA Deep Dive (Part 2): The pi0 Family and Action Interface Evolution
• 09-04 From RSSM to Modern Latent Dynamics: How the 'Engine' of World Models Evolves
• 09-03 VLA Deep Dive (Part 1): From RT-2 to OpenVLA -- Foundations and Early Evolution of End-to-End Policies
• 09-02 JEPA Deep Dive: From I-JEPA to V-JEPA 2-AC — How Predictive Representation Learning Leads to World Models
• 09-01 World Models 2026: From Cosmos, Genie to JEPA — The Divergence of Routes

August

• 08-31 From Dreamer to World Model Agents: Future Directions and Research Trends
• 08-30 Dreamer in Practice: From Simulation Control to Sim-to-Real
• 08-29 DreamerV3 GPU Selection Guide: From VRAM Requirements to Cost-Effectiveness Analysis
• 08-28 DreamerV3 Training Engineering: From GPU Setup to Hyperparameter Tuning
• 08-27 Dreamer's Actor-Critic: How Policy Optimization Works in Imagination
• 08-26 What Does a World Model Actually Do in a Robot? From Perception to Action
• 08-25 Understanding Dreamer: How World Models Learn to Imagine
• 08-24 Understanding RSSM Through Code (6): Default Config, Four Formulas, and the Code↔Math↔Semantics Map
• 08-23 Understanding RSSM Through Code (5): Imagine, Observe vs. Imagine, Sequence Training, and Reset
• 08-22 Understanding RSSM Through Code (4): KL Balancing, Free Nats, and the Final KL Combination
• 08-21 Understanding RSSM Through Code (3): Deterministic Transition _core(), deter=8192, and Block GRU
• 08-20 Understanding RSSM Through Code (2): Prior/Posterior, Straight-Through Sampling, and unimix
• 08-19 Understanding RSSM Through Code (1): Where RSSM Sits in DreamerV3 and the Stochastic State
• 08-18 The World Model Hype Has Been Exaggerated: A Technical Analysis
• 08-17 DreamerV3 GPU Infrastructure: Cloud vs Self-Built Cost Analysis
• 08-16 AI Is Looking for a Physical Shell: From Smart Glasses to Wearable Agents
• 08-15 Isaac Lab Installation Guide: From Zero to Running on AutoDL
• 08-14 Isaac Lab: From DreamerV3 to Industrial-Scale Robot RL Training
• 08-13 The Data Challenge in Robotics: Where Does Robot Learning Data Come From?
• 08-12 When World Models Meet Transformers: From RSSM to Large-Scale Sequence Modeling
• 08-11 DreamerV3 Training Tips: Lessons from Real-World Debugging
• 08-10 Four Paradigms of World Model Representations: A Comparative Analysis
• 08-09 MuJoCo vs Isaac Sim: How to Choose the Right Robot Simulation Platform
• 08-08 Domain Randomization: The Bridge from Simulation to Reality
• 08-07 TD-MPC: How World Models Enable Robot Control
• 08-06 World Models as Synthetic Data Engines for VLA Training
• 08-05 Is Sim-to-Real Too Hard? World Model-Driven Adaptive Transfer Methods
• 08-05 Building a World Model Lab from Scratch: A MuJoCo + DreamerV3 Practical Guide
• 08-05 ABot-World-0: 24-Hour Stable Inference from an Interactive World Model
• 08-04 VLA vs World Models: Which Will Prevail?
• 08-04 Embodied AI and RL: Career Prospects and Salary Guide in 2026
• 08-03 World Models: 8 Years and the Same Bottleneck
• 08-03 World Models in 2026: Where Are the Real Opportunities?
• 08-03 Is World Model a Good Research Direction? An Engineer's Honest Assessment
• 08-03 How to Get Started with Reinforcement Learning: A Practical Guide
• 08-03 Embodied AI in 2026: What Breakthroughs Can We Expect?
• 08-02 Deep Dive into RSSM: The Core Engine of World Models
• 08-01 What Is a Robot World Model? An Engineer's Deep Dive

July

• 07-31 Can You Break Into Embodied AI Without a PhD?