sentdex
Automation, LLM APIs, and Coding Tools with a focus on applied robotics and engineering.
Nutrition Label
Sentdex delivers hands-on engineering content that bridges the gap between AI software and physical robotics. Viewers can expect raw, code-heavy demonstrations where he builds, breaks, and troubleshoots complex systems in real-time rather than polished theoretical overviews.
Strengths
- +
- +
- +
Notes
- !Content ranges from intensive coding sessions to lighter updates or unboxings.
- !He frequently purchases expensive hardware himself, often clarifying lack of brand affiliation.
Rating Breakdown
Breakdown across the key dimensions we rate. Methodology →
Recent Videos
Training a Unitree G1 to Walk w/ Reinforcement Learning
Unitree G1 Security Disaster
Testing VLMs and LLMs for robotics w/ the Jetson Thor devkit
Reinforcement learning with Unitree G1 humanoid - Dev w/ G1 P.5
A bigger brain for the Unitree G1- Dev w/ G1 Humanoid P.4
Unitree G1 - Moving the arms/hands - Dev w/ G1 Humanoid P.3
Unitree G1 LiDAR, SLAM, navigation and control. Dev w/ G1 Humanoid P.2
Unboxing the Unitree G1 Edu Humanoid
Vibe Coding a Robotic Hand to Crawl (Inspire RH56DFQ)
Vibe Coding Robot Hands w/ Cursor (Inspire RH56DFQ-2L/R)
Programming with LLM Agents in 2025
What's going on everybody?
Building an LLM fine-tuning Dataset
Visualizing Neural Network Internals
Getting Back on Grid
Why this rating
Evidence receipts showing why each dimension is rated the way it is.
“We're just going to run it and see what happens... [Robot arm moves physically in the background]”[04:15] →
Demonstrates 'show, don't tell' by running code on physical hardware with visible, real-time results.
“It's a little bit jittery... we might need to tune the damping.”[06:30] →
Openly acknowledges performance imperfections and friction points during the development process, rather than presenting a polished fake result.
“The biggest thing that is missing... is domain randomization. That is huge for sim-to-real transfer.”[04:12] →
Identifies a specific, high-level technical limitation in the software stack (mjlab) rather than just glossing over why the training is difficult.
“Loss is going up. So that's why I stopped it... maybe I just need to change learning rate.”[11:25] →
Shows actual training data (TensorBoard) including failure metrics, rather than just showing the successful final result.