The Internet of Things,
wired up from first principles.

What is IoT, really?

The Internet of Things is a network of physical objects — sensors, appliances, vehicles, machines — embedded with electronics and software that let them collect data and exchange it with each other or the cloud, usually with little to no human in the loop. It's not one technology; it's four layers stacked on top of each other, each doing one job:

A single project rarely needs custom work at every layer — most tutorials start at Perception and stop at Network. The Processing and Application layers are where a toy project turns into a product.

Why IoT matters now — and what the world looks like without it

Sensors got cheap, Wi-Fi chips got tiny, and cloud storage got nearly free — all in the same decade. That combination is why IoT went from research labs to every home in about ten years. Here's the difference it actually makes:

How an IoT system actually works

Strip away the branding and almost every IoT product is the same five-node pipeline. Data flows right; decisions and commands can flow back left.

The gateway step is easy to skip in a tutorial (your ESP32 can talk straight to the cloud) but matters at scale — a real deployment often has dozens of cheap sensors reporting to one gateway that handles the expensive internet connection.

The full tech stack

Raspberry Pi vs Arduino

The single most common beginner confusion. The short version: Arduino is a microcontroller — it runs one program, close to the hardware, forever, with no operating system. Raspberry Pi is a full computer — it boots Linux, runs multiple programs, and is far more powerful but also far less predictable in timing.

Sensors you'll actually use

How to learn IoT with AI as your lab partner

AI won't solder your board, but it collapses the two slowest parts of learning embedded systems: reading dense datasheets and debugging a circuit you can't see. Use it as a second pair of eyes at every step, not a replacement for actually wiring things up.

1. 01

   Get one board blinking

   Arduino Uno or ESP32 + the classic blink sketch. This proves your toolchain, drivers, and upload process all work.

   Paste any upload error verbatim into Claude — 90% of first-week bugs are a missing board driver or wrong COM port.

2. 02

   Read one sensor

   Wire a DHT22 or an LDR and print readings to the Serial Monitor. Learn what 'noisy data' actually looks like.

   Ask AI to explain your sensor's datasheet in plain English — datasheets are dense on purpose, AI compresses them fast.

3. 03

   Simulate before you solder

   Use Wokwi or Tinkercad Circuits to test wiring and code virtually — no risk of frying a board while learning.

   Describe your circuit in words and ask AI to generate a starting Wokwi diagram.json to skip the blank-canvas problem.

4. 04

   Talk to the internet

   Publish sensor data over MQTT to a free broker (e.g. HiveMQ public broker) and view it on a simple dashboard.

   Ask AI to write both the publisher firmware AND a matching subscriber script — mismatched topics are the #1 silent failure.

5. 05

   Add a brain

   Move from 'if sensor > threshold' to a small on-device model (TinyML) or a cloud rule engine for real decisions.

   Ask AI to explain the tradeoff between edge inference (fast, private) and cloud inference (powerful, needs connectivity) for your exact project.

6. 06

   Build a full loop

   Sensor → microcontroller → broker → database → dashboard → alert. One real end-to-end project teaches more than ten tutorials.

   Have AI review your architecture diagram before you build — catching a design flaw on paper is free; catching it after wiring 12 nodes isn't.

Cheat sheets

Protocol comparison

Arduino sketch skeleton

void setup() {
  Serial.begin(115200);      // debug output
  pinMode(2, INPUT);         // sensor pin
  pinMode(13, OUTPUT);       // actuator / LED
}

void loop() {
  int reading = digitalRead(2);
  if (reading == HIGH) {
    digitalWrite(13, HIGH);  // react to sensor
  } else {
    digitalWrite(13, LOW);
  }
  delay(50);                 // simple debounce
}

MQTT publish / subscribe (Python)

import paho.mqtt.client as mqtt

BROKER, TOPIC = "broker.hivemq.com", "myhome/livingroom/temp"

# --- publisher (on the device) ---
client = mqtt.Client()
client.connect(BROKER, 1883)
client.publish(TOPIC, payload="24.5", qos=1)

# --- subscriber (on the dashboard) ---
def on_message(client, userdata, msg):
    print(msg.topic, msg.payload.decode())

sub = mqtt.Client()
sub.on_message = on_message
sub.connect(BROKER, 1883)
sub.subscribe(TOPIC, qos=1)
sub.loop_forever()

Raspberry Pi GPIO (Python)

from gpiozero import LED, MotionSensor
from time import sleep

led = LED(17)
pir = MotionSensor(4)

while True:
    if pir.motion_detected:
        led.on()
    else:
        led.off()
    sleep(0.1)

Important things to keep in mind

Puzzled? Try these interview questions

These aren't definition-recall questions — they're the kind that expose whether you've actually debugged a circuit at 1am. Tap a question to reveal the answer.