The large-scale physical design is as simple as possible (just regular wiring), the individual components are all "smart" (Christmas lights each with their own integrated IC, doesn't get more embedded than that) and addressable via a communication protocol.
Imagine how difficult the wiring would be to have 100 addressable lights that didn't use this arrangement!
The key is to drive a current and control the voltage at both ends of a string such that each LED detects modulation and bypasses (itself) or consumes power (in-line). That might be relevant where the strength of a single (e.g. steel) wire was used to support a long string. By driving relatively low (max RGB) current at higher voltages and time (e.g. PWM) sequencing the light modulation.
In a simple case, imaging there are 100 RGB lights, half of them excited on each polarity and bypassed with a 0.2V Schottky stacking to +/-10V minimum (for digital data) at <100mA. You might synchronize and send data to the smart lights each "zero crossing" and get away +/-50Vpp for the string by only having 10 (12.5%) stacked (inline) on at any one time (a controller for the LED FETs can stay awake on a capacitor). As long as you run at a few kHz you won't have noticeable flicker, capacitive and you can update with a few bits per light each cycle.
More complex designs could use FETs to improve data link efficiency of the minimum voltage and higher current (lower PWM times) could substantially reduce peak voltage to +/-5V or so.
Fortunately my place is so small a 100’ cat-5 cable reaches any point in it, so I just use wired. Wifi barely works.
Considering wifi-6, but I can’t figure out if it helps with this type of congestion or not…
The hardware world moves much slower than software one, and the if the designs work, they are rarely replaced.
The ones in the article have only 2 wires, positive and negative.
1-wire is a similarly clever protocol.