One wire, five LEDs

Each LED has its own chip

An ordinary RGB LED is three LEDs in one package, and your board drives each colour from its own pin. The five on the TK33 are WS2812B LEDs: the same three colours plus a small driver chip, all in a 5 mm square. The chip holds three brightness values, one per colour, and keeps the LEDs at them by itself.

So your board never drives an LED. It sends numbers down one wire, called DATA, and the chips do the rest.

The first 24 bits are yours, LED1

A colour is three bytes, 24 bits: green, red, blue, 0 to 255 each. A frame for this bar is five of those, 120 bits, sent as one stream.

The top LED, LED1, keeps the first 24 bits that reach it. Everything after them it sends out of its own output pin, DOUT, rebuilt as a clean signal, into the next LED's input, DIN. LED2 does the same with what is left, and so on down. When the line then stays low for at least 80 µs, every LED shows what it kept.

That is why no LED has an address. Which colour an LED gets depends only on where it sits in the chain: LED1 is pixel 0 in a sketch, and the bottom LED is pixel 4.

Where the stream stops

LED5 keeps the last 24 bits and passes on whatever follows. On this board that goes nowhere: LED5's DOUT is soldered to a pad that no trace reaches. A sketch that sends ten colours lights these five and loses the other five. Where the chain ends is about what that means for a second bar.

What it costs

The timing is the price. A 0 and a 1 are told apart by a few hundred nanoseconds of pulse width, far too fine for digitalWrite(). A library does it: on an ESP32 it hands the bits to a peripheral built for exact pulses, so nothing else in the sketch can upset them. Bits as pulse widths shows the rule the LED reads by.