Robotics IoT AI

built explained shared

Hands-on work in robotics, IoT and AI, documented step by step. Projects ship with a GitHub repo and an install script, so the same setup runs on your own hardware.

Browse

Project code updated as on Sep 2026

ROBOTICS

A four-level Raspberry Pi robot

Updated for Raspberry Pi OS 13 (Trixie)

One rover, built up in four levels — a step-by-step build for students and hobbyists. Each level is its own repository and its own article, so you can stop at any of them and still have a robot that drives.

people running this build

1,296 installs
83 countries
Raspberry Pi token •••• ESP32 token •••• REST API this server Water tank live level Pi remote gpio · camera ESP32 remote 8 gpio buttons reading command token on the device · no port forwarding
IoT

Browser widgets and controls for SBCs and MCUs

Widgets that pair with a board over the internet: a Raspberry Pi remote for GPIO and camera, an ESP32 remote, and a water tank reading a live ultrasonic sensor. No port forwarding and no server code — the device holds an access token, the REST API on this server does the rest.

Create an account, download the script to your board, and it appears in your browser.

boards talking to this site

247 devices
2 in the last 24h

Boards that have actually paired and sent data — devices, not sign-ups.

device reporting now· 25.6 cm· just now
EDGE AI

On-device machine learning

Updated for Raspberry Pi OS 13 (Trixie)

The camera feed is read by a TensorFlow Lite model running on the Pi itself. Model Garden benchmarks 20+ models on the same board so you can see what actually fits, and a Coral USB Accelerator makes them 10–30× faster when one is plugged in.

people running model garden

90 installs
25 countries

The same install telemetry as the robot, counted for the Model Garden benchmark alone.

one topic, one level deep IN PLAIN WORDS Maximum Calibration Error Imagine a weather app that says “80% chance of rain.” If, over many days with that same forecast, it actually rains about 80% of the time, the app is well calibrated. If it only rains 50% of the time, the app is overconfident. MCE measures the worst mismatch like this.
NOTES

A curated knowledge base

A topic is broken down into the terms that actually carry it, and each term is explained from more than one angle: plain words first, then an analogy, then the precise technical definition — so you can stop at the depth you need. Quizzes and a concept map are generated from the same structure.

AI writes the first draft; a human verifies, corrects and curates every note before it is published.

WHAT'S NEXT

Upcoming ideas — experiments in progress on the bench. Each one that works becomes a repository and an article here.

  • RobotNewer models on board, and Gen AI — so you can ask it things instead of steering it.
  • IoTMore boards to plug in — the Arduino Uno Q, and other new microcontrollers.
  • Edge AIMore models on the test bench, each one measured on the same Pi.
  • NotesNew topics being written: robotics, and today's small boards.

BLOG

The write-ups: build guides, experiments and notes from the bench.

Scroll to Top