[Paper Notes] PointVLA: Injecting the 3D World into Vision-Language-Action Models
Published:
Posts by Tags
3D Point Cloud
3D Scene Generation
3D Vision
AI
Bootstrapping and the Game of Entrepreneurship
Published:
Bootstrapping is the same pattern across compilers, AI, and startups: borrow external structure to cold-start, then recursively replace dependencies until the system sustains itself.
Michael Polanyi: Tacit Knowledge, Personal Knowing, and What AI Still Cannot Tell
Published:
Polanyi’s framework of tacit knowledge and personal knowing offers a lens for understanding what AI still cannot do — and why embodied, context-dependent skill remains hard to formalize.
The Singularity is Near
Published:
Autonomous agents are now cheap and networked enough to spread faster than any institution can contain — the question is what order emerges after control becomes partial.
Intelligence Is Not Only Reasoning
Published:
Intelligence depends on retrieval and communication as much as reasoning — a powerful model limited by information gaps is still limited.
Cognitive Bandwidth in the AI Agent Era
Published:
As AI tools get more capable, the bottleneck shifts from tool friction to human cognitive bandwidth — compressing intent and steering effectively becomes the key skill.
AI Agents
Distilling Persons into Agents: A Survey of Recent ‘Person-as-Skill’ Projects
Published:
A survey of projects that compress human personas into AI agent skills — from departed colleagues to public figures — and what this trend reveals about AI, memory, and identity.
The Singularity is Near
Published:
Autonomous agents are now cheap and networked enough to spread faster than any institution can contain — the question is what order emerges after control becomes partial.
Intelligence Is Not Only Reasoning
Published:
Intelligence depends on retrieval and communication as much as reasoning — a powerful model limited by information gaps is still limited.
Cognitive Bandwidth in the AI Agent Era
Published:
As AI tools get more capable, the bottleneck shifts from tool friction to human cognitive bandwidth — compressing intent and steering effectively becomes the key skill.
AI Ethics
Distilling Persons into Agents: A Survey of Recent ‘Person-as-Skill’ Projects
Published:
A survey of projects that compress human personas into AI agent skills — from departed colleagues to public figures — and what this trend reveals about AI, memory, and identity.
Action Coherence
Admittance Control
Force Control and the Missing Layer in Embodied Intelligence
Published:
A late-night conversation over grilled skewers revealed a gap: most embodied AI researchers don’t think about what the robot arm is actually doing at the control level — and closing that gap might reshape how we think about action spaces.
Agentic AI
[Paper Notes] CUDA Agent: Large-Scale Agentic RL for High-Performance CUDA Kernel Generation
Published:
Alignment
Automated Discovery
Benchmarking
Bimanual Interaction
Bimanual Manipulation
Blog Notes
Bootstrapping
Bootstrapping and the Game of Entrepreneurship
Published:
Bootstrapping is the same pattern across compilers, AI, and startups: borrow external structure to cold-start, then recursively replace dependencies until the system sustains itself.
CUDA
[Paper Notes] CUDA Agent: Large-Scale Agentic RL for High-Performance CUDA Kernel Generation
Published:
Claude Code
Claude Skills
Distilling Persons into Agents: A Survey of Recent ‘Person-as-Skill’ Projects
Published:
A survey of projects that compress human personas into AI agent skills — from departed colleagues to public figures — and what this trend reveals about AI, memory, and identity.
CoRL
[Paper Notes] DayDreamer: World Models for Physical Robot Learning - CoRL 2022
Published:
Key information
- This paper learns two models: a world model trained on off-policy sequences through supervised learning, and an actor-critic model to learn behaviors from trajectories predicted by the learned model.
- The data collection and learning updates are decoupled, enabling fast training without waiting for the environment. A learner thread continuously trains the world model and actor-critic behavior, while an actor thread in parallel computes actions for environment interaction.
Code Generation
[Paper Notes] CUDA Agent: Large-Scale Agentic RL for High-Performance CUDA Kernel Generation
Published:
Codex
I Vibe Coded an Operating System
Published:
I built a scripting-first OS prototype in a weekend with Codex, where the system language gradually takes over its own environment from the host.
Cognition
Cognitive Bandwidth in the AI Agent Era
Published:
As AI tools get more capable, the bottleneck shifts from tool friction to human cognitive bandwidth — compressing intent and steering effectively becomes the key skill.
Communication
Intelligence Is Not Only Reasoning
Published:
Intelligence depends on retrieval and communication as much as reasoning — a powerful model limited by information gaps is still limited.
Coordinate Frames
How to Align Teleoperation Devices with Robot End Effectors
Published:
The hardest part of teleoperation is aligning coordinate frames between input devices and robot end effectors — a practical walkthrough.
Cross-Embodiment
Cross-Embodiment Learning
Data-Centric AI
Dataset
[Paper Notes] EgoVerse: An Egocentric Human Dataset for Robot Learning from Around the World
Published:
Dexterous Grasping
[Paper Notes] D-REX: Differentiable Real-to-Sim-to-Real Engine for Learning Dexterous Grasping
Published:
Dexterous Manipulation
[Paper Notes] SaTA: Spatially-anchored Tactile Awareness for Robust Dexterous Manipulation
Published:
Differentiable Simulation
[Paper Notes] D-REX: Differentiable Real-to-Sim-to-Real Engine for Learning Dexterous Grasping
Published:
Diffusion Models
Diffusion Policy
Digital Twins
Distilling Persons into Agents: A Survey of Recent ‘Person-as-Skill’ Projects
Published:
A survey of projects that compress human personas into AI agent skills — from departed colleagues to public figures — and what this trend reveals about AI, memory, and identity.
DiscoRL
Discrete Diffusion
Distance Fields
Dynamics Models
EMG
Economics
The Singularity is Near
Published:
Autonomous agents are now cheap and networked enough to spread faster than any institution can contain — the question is what order emerges after control becomes partial.
Egocentric Data
Egocentric Vision
[Paper Notes] EgoVerse: An Egocentric Human Dataset for Robot Learning from Around the World
Published:
Embodied AI
[Paper Notes] RLinf → RLinf-USER → RL-Co: A Full-Stack RL Pipeline for Embodied VLA Training
Published:
Embodied Intelligence
Force Control and the Missing Layer in Embodied Intelligence
Published:
A late-night conversation over grilled skewers revealed a gap: most embodied AI researchers don’t think about what the robot arm is actually doing at the control level — and closing that gap might reshape how we think about action spaces.
End Effectors
How to Align Teleoperation Devices with Robot End Effectors
Published:
The hardest part of teleoperation is aligning coordinate frames between input devices and robot end effectors — a practical walkthrough.
Entrepreneurship
Bootstrapping and the Game of Entrepreneurship
Published:
Bootstrapping is the same pattern across compilers, AI, and startups: borrow external structure to cold-start, then recursively replace dependencies until the system sustains itself.
Epistemology
Michael Polanyi: Tacit Knowledge, Personal Knowing, and What AI Still Cannot Tell
Published:
Polanyi’s framework of tacit knowledge and personal knowing offers a lens for understanding what AI still cannot do — and why embodied, context-dependent skill remains hard to formalize.
Flow Matching
Force Control
Force Control and the Missing Layer in Embodied Intelligence
Published:
A late-night conversation over grilled skewers revealed a gap: most embodied AI researchers don’t think about what the robot arm is actually doing at the control level — and closing that gap might reshape how we think about action spaces.
Foundation Model
Gaussian Splatting
[Paper Notes] D-REX: Differentiable Real-to-Sim-to-Real Engine for Learning Dexterous Grasping
Published:
Generative Models
Goal-Conditioned RL
Goal-conditioned
Grasping
Hand Morphology
Hand Motion
Human Video Learning
Human Videos
Human-to-Robot Transfer
[Paper Notes] EgoVerse: An Egocentric Human Dataset for Robot Learning from Around the World
Published:
Humanoid
Humanoid Robotics
[Paper Notes] DiT4DiT: Jointly Modeling Video Dynamics and Actions for Generalizable Robot Control
Published:
Imitation Learning
[Paper Notes] EgoVerse: An Egocentric Human Dataset for Robot Learning from Around the World
Published:
[Paper Notes] SaTA: Spatially-anchored Tactile Awareness for Robust Dexterous Manipulation
Published:
[Paper Notes] D-REX: Differentiable Real-to-Sim-to-Real Engine for Learning Dexterous Grasping
Published:
Impedance Control
Force Control and the Missing Layer in Embodied Intelligence
Published:
A late-night conversation over grilled skewers revealed a gap: most embodied AI researchers don’t think about what the robot arm is actually doing at the control level — and closing that gap might reshape how we think about action spaces.
In-Hand Manipulation
Inference Acceleration
Information Retrieval
Intelligence Is Not Only Reasoning
Published:
Intelligence depends on retrieval and communication as much as reasoning — a powerful model limited by information gaps is still limited.
Information Theory
Intelligence
Intelligence Is Not Only Reasoning
Published:
Intelligence depends on retrieval and communication as much as reasoning — a powerful model limited by information gaps is still limited.
Internet
The Singularity is Near
Published:
Autonomous agents are now cheap and networked enough to spread faster than any institution can contain — the question is what order emerges after control becomes partial.
Inverse Dynamics
Inverse Kinematics
How to Align Teleoperation Devices with Robot End Effectors
Published:
The hardest part of teleoperation is aligning coordinate frames between input devices and robot end effectors — a practical walkthrough.
Inverse Kinematics for Any URDF: Tasks, Weights, and Solver Design
Published:
A practical guide to formulating IK as weighted nonlinear least-squares, with task definition, damping, and solver design that works for any URDF.
Knowledge Transfer
LLM
[Paper Notes] CUDA Agent: Large-Scale Agentic RL for High-Performance CUDA Kernel Generation
Published:
LLM Agents
LLM Workflows
Latent Actions
Legged Locomotion
[Project Notes] Robotic World Model Lite
Published:
Long-Horizon Control
Machine Learning Theory
Manipulation
Memory
Meta-Learning
Model-Based Control
Model-Based RL
[Project Notes] Robotic World Model Lite
Published:
Motion Generation
Motion Imitation
[Project Notes] PAL Robotics in mjlab
Published:
MuJoCo
[Project Notes] PAL Robotics in mjlab
Published:
Multitask Learning
Novel View Synthesis
Offline RL
[Project Notes] Robotic World Model Lite
Published:
Online RL
Operating Systems
Bootstrapping and the Game of Entrepreneurship
Published:
Bootstrapping is the same pattern across compilers, AI, and startups: borrow external structure to cold-start, then recursively replace dependencies until the system sustains itself.
I Vibe Coded an Operating System
Published:
I built a scripting-first OS prototype in a weekend with Codex, where the system language gradually takes over its own environment from the host.
Optimization
Inverse Kinematics for Any URDF: Tasks, Weights, and Solver Design
Published:
A practical guide to formulating IK as weighted nonlinear least-squares, with task definition, damping, and solver design that works for any URDF.
PAL Robotics
[Project Notes] PAL Robotics in mjlab
Published:
Paper Notes
[Paper Notes] EgoVerse: An Egocentric Human Dataset for Robot Learning from Around the World
Published:
[Paper Notes] CUDA Agent: Large-Scale Agentic RL for High-Performance CUDA Kernel Generation
Published:
[Paper Notes] SaTA: Spatially-anchored Tactile Awareness for Robust Dexterous Manipulation
Published:
[Paper Notes] DiT4DiT: Jointly Modeling Video Dynamics and Actions for Generalizable Robot Control
Published:
[Paper Notes] D-REX: Differentiable Real-to-Sim-to-Real Engine for Learning Dexterous Grasping
Published:
[Paper Notes] RLinf → RLinf-USER → RL-Co: A Full-Stack RL Pipeline for Embodied VLA Training
Published:
Persona
Distilling Persons into Agents: A Survey of Recent ‘Person-as-Skill’ Projects
Published:
A survey of projects that compress human personas into AI agent skills — from departed colleagues to public figures — and what this trend reveals about AI, memory, and identity.
Personal Thoughts
Force Control and the Missing Layer in Embodied Intelligence
Published:
A late-night conversation over grilled skewers revealed a gap: most embodied AI researchers don’t think about what the robot arm is actually doing at the control level — and closing that gap might reshape how we think about action spaces.
The Singularity is Near
Published:
Autonomous agents are now cheap and networked enough to spread faster than any institution can contain — the question is what order emerges after control becomes partial.
Intelligence Is Not Only Reasoning
Published:
Intelligence depends on retrieval and communication as much as reasoning — a powerful model limited by information gaps is still limited.
Cognitive Bandwidth in the AI Agent Era
Published:
As AI tools get more capable, the bottleneck shifts from tool friction to human cognitive bandwidth — compressing intent and steering effectively becomes the key skill.
Philosophy
Michael Polanyi: Tacit Knowledge, Personal Knowing, and What AI Still Cannot Tell
Published:
Polanyi’s framework of tacit knowledge and personal knowing offers a lens for understanding what AI still cannot do — and why embodied, context-dependent skill remains hard to formalize.
Pinocchio
Inverse Kinematics for Any URDF: Tasks, Weights, and Solver Design
Published:
A practical guide to formulating IK as weighted nonlinear least-squares, with task definition, damping, and solver design that works for any URDF.
Planning
Point Clouds
Policy Improvement
Policy Learning
[Paper Notes] DiT4DiT: Jointly Modeling Video Dynamics and Actions for Generalizable Robot Control
Published:
Pretraining
Productivity
Cognitive Bandwidth in the AI Agent Era
Published:
As AI tools get more capable, the bottleneck shifts from tool friction to human cognitive bandwidth — compressing intent and steering effectively becomes the key skill.
Programming Languages
I Vibe Coded an Operating System
Published:
I built a scripting-first OS prototype in a weekend with Codex, where the system language gradually takes over its own environment from the host.
Project Notes
[Project Notes] Robotic World Model Lite
Published:
[Project Notes] PAL Robotics in mjlab
Published:
I Vibe Coded an Operating System
Published:
I built a scripting-first OS prototype in a weekend with Codex, where the system language gradually takes over its own environment from the host.
Prompt Engineering
RL Algorithms
RSS
[Paper Notes] SAM-RL: Sensing-Aware Model-Based Reinforcement Learning via Differentiable Physics-Based Simulation and Rendering - RSS 2023
Published:
Key information
- Model-based reinforcement learning (MBRL) is potentially more sample-efficient than model-free RL.
- Integration of differentiable physics-based simulation and rendering facilitates the learning process.
- Pipeline: Real2Sim -> Learn@Sim -> Sim2Real to produce efficient policies.
Reinforcement Learning
[Project Notes] PAL Robotics in mjlab
Published:
[Paper Notes] CUDA Agent: Large-Scale Agentic RL for High-Performance CUDA Kernel Generation
Published:
[Paper Notes] RLinf → RLinf-USER → RL-Co: A Full-Stack RL Pipeline for Embodied VLA Training
Published:
[Paper Notes] SAM-RL: Sensing-Aware Model-Based Reinforcement Learning via Differentiable Physics-Based Simulation and Rendering - RSS 2023
Published:
Key information
- Model-based reinforcement learning (MBRL) is potentially more sample-efficient than model-free RL.
- Integration of differentiable physics-based simulation and rendering facilitates the learning process.
- Pipeline: Real2Sim -> Learn@Sim -> Sim2Real to produce efficient policies.
[Paper Notes] DayDreamer: World Models for Physical Robot Learning - CoRL 2022
Published:
Key information
- This paper learns two models: a world model trained on off-policy sequences through supervised learning, and an actor-critic model to learn behaviors from trajectories predicted by the learned model.
- The data collection and learning updates are decoupled, enabling fast training without waiting for the environment. A learner thread continuously trains the world model and actor-critic behavior, while an actor thread in parallel computes actions for environment interaction.
Research Automation
Reward Modeling
Robot Foundation Models
Robot Learning
Robotics
Force Control and the Missing Layer in Embodied Intelligence
Published:
A late-night conversation over grilled skewers revealed a gap: most embodied AI researchers don’t think about what the robot arm is actually doing at the control level — and closing that gap might reshape how we think about action spaces.
[Paper Notes] EgoVerse: An Egocentric Human Dataset for Robot Learning from Around the World
Published:
Michael Polanyi: Tacit Knowledge, Personal Knowing, and What AI Still Cannot Tell
Published:
Polanyi’s framework of tacit knowledge and personal knowing offers a lens for understanding what AI still cannot do — and why embodied, context-dependent skill remains hard to formalize.
[Project Notes] Robotic World Model Lite
Published:
[Project Notes] PAL Robotics in mjlab
Published:
[Paper Notes] SaTA: Spatially-anchored Tactile Awareness for Robust Dexterous Manipulation
Published:
[Paper Notes] DiT4DiT: Jointly Modeling Video Dynamics and Actions for Generalizable Robot Control
Published:
How to Align Teleoperation Devices with Robot End Effectors
Published:
The hardest part of teleoperation is aligning coordinate frames between input devices and robot end effectors — a practical walkthrough.
Inverse Kinematics for Any URDF: Tasks, Weights, and Solver Design
Published:
A practical guide to formulating IK as weighted nonlinear least-squares, with task definition, damping, and solver design that works for any URDF.
[Paper Notes] D-REX: Differentiable Real-to-Sim-to-Real Engine for Learning Dexterous Grasping
Published:
[Paper Notes] RLinf → RLinf-USER → RL-Co: A Full-Stack RL Pipeline for Embodied VLA Training
Published:
[Paper Notes] SAM-RL: Sensing-Aware Model-Based Reinforcement Learning via Differentiable Physics-Based Simulation and Rendering - RSS 2023
Published:
Key information
- Model-based reinforcement learning (MBRL) is potentially more sample-efficient than model-free RL.
- Integration of differentiable physics-based simulation and rendering facilitates the learning process.
- Pipeline: Real2Sim -> Learn@Sim -> Sim2Real to produce efficient policies.
[Paper Notes] DayDreamer: World Models for Physical Robot Learning - CoRL 2022
Published:
Key information
- This paper learns two models: a world model trained on off-policy sequences through supervised learning, and an actor-critic model to learn behaviors from trajectories predicted by the learned model.
- The data collection and learning updates are decoupled, enabling fast training without waiting for the environment. A learner thread continuously trains the world model and actor-critic behavior, while an actor thread in parallel computes actions for environment interaction.
Robustness
Runtime Systems
I Vibe Coded an Operating System
Published:
I built a scripting-first OS prototype in a weekend with Codex, where the system language gradually takes over its own environment from the host.
Sample Efficiency
Scaling Laws
[Paper Notes] EgoVerse: An Egocentric Human Dataset for Robot Learning from Around the World
Published:
Science Robotics
Self-Hosting
Bootstrapping and the Game of Entrepreneurship
Published:
Bootstrapping is the same pattern across compilers, AI, and startups: borrow external structure to cold-start, then recursively replace dependencies until the system sustains itself.
Sim-to-Real
Sim2Real
[Paper Notes] D-REX: Differentiable Real-to-Sim-to-Real Engine for Learning Dexterous Grasping
Published:
Simulation
Single-Image Reconstruction
Spatial Grounding
State Representation
Synthetic Data
Systems
[Paper Notes] RLinf → RLinf-USER → RL-Co: A Full-Stack RL Pipeline for Embodied VLA Training
Published:
Tabletop Scenes
Tacit Knowledge
Michael Polanyi: Tacit Knowledge, Personal Knowing, and What AI Still Cannot Tell
Published:
Polanyi’s framework of tacit knowledge and personal knowing offers a lens for understanding what AI still cannot do — and why embodied, context-dependent skill remains hard to formalize.
Tactile Sensing
[Paper Notes] SaTA: Spatially-anchored Tactile Awareness for Robust Dexterous Manipulation
Published:
Task and Motion Planning
Technical Notes
How to Align Teleoperation Devices with Robot End Effectors
Published:
The hardest part of teleoperation is aligning coordinate frames between input devices and robot end effectors — a practical walkthrough.
Inverse Kinematics for Any URDF: Tasks, Weights, and Solver Design
Published:
A practical guide to formulating IK as weighted nonlinear least-squares, with task definition, damping, and solver design that works for any URDF.
Teleoperation
How to Align Teleoperation Devices with Robot End Effectors
Published:
The hardest part of teleoperation is aligning coordinate frames between input devices and robot end effectors — a practical walkthrough.
Test-time Guidance
URDF
Inverse Kinematics for Any URDF: Tasks, Weights, and Solver Design
Published:
A practical guide to formulating IK as weighted nonlinear least-squares, with task definition, damping, and solver design that works for any URDF.
Unsupervised RL
VLA
[Paper Notes] RLinf → RLinf-USER → RL-Co: A Full-Stack RL Pipeline for Embodied VLA Training
Published:
VLM Fine-tuning
Video Diffusion
[Paper Notes] DiT4DiT: Jointly Modeling Video Dynamics and Actions for Generalizable Robot Control
Published:
Video Generation
Video World Models
Vision-Language Models
Vision-Language-Action
Vision-Language-Action Models
[Paper Notes] DiT4DiT: Jointly Modeling Video Dynamics and Actions for Generalizable Robot Control
Published:
Vision-Tactile Learning
World Model
World Models
[Project Notes] Robotic World Model Lite
Published:
Zero-shot Generalization
Zero-shot Manipulation
update
Updating website
Published:
A major site update done mostly by Codex — page refactoring, content restructuring, and bilingual support.
website
Updating website
Published:
A major site update done mostly by Codex — page refactoring, content restructuring, and bilingual support.