In 1995, connecting a printer, modem, or joystick to a PC meant a tangle of parallel ports, serial ports, PS/2 plugs, and game ports, plus driver floppies and hand-assigned IRQ channels that crashed the machine when two devices collided. Seven companies, Intel, Microsoft, IBM, Compaq, DEC, NEC, and Nortel, set out to replace all of it with one plug and play, hot-swappable port that carried its own power. This episode traces the Universal Serial Bus from that fix to the cable that now runs nearly everything on your desk.

The story turns on a design choice: a tiered star topology with a single host in charge and up to 127 peripherals that speak only when spoken to. Putting all the intelligence in the computer made mice and keyboards almost free to build, and that economics let USB outlast the technically superior FireWire. The same trusting, low-cost design also left an opening for BadUSB sticks that pose as keyboards and USB killers that fry a motherboard with its own electricity.

• USB 2.0 reached 480 megabits per second in 2000 but ran half-duplex, like a walkie-talkie. USB 3.0 arrived in 2008 with five extra wires, full duplex, and 5 gigabit SuperSpeed.
• The 8b/10b encoding in USB 3.0 kept sender and receiver clocks in sync at a cost of 20 percent overhead. A later 128b/132b scheme cut that to about 3 percent.
• Names like USB 3.2 Gen 2×2 were engineering descriptions, not marketing, and the standards body eventually switched to plain speed labels such as USB 20Gbps.
• USB-C dropped the old Type-A and Type-B roles in favor of configuration channel pins that negotiate who is host and who supplies power, scaling from 2.5 watts in the original spec to 240 watts at 48 volts.
• A USB killer charges internal capacitors from the port’s 5 volts, steps up past 200 volts, and fires it back down the data lines. Unlike Ethernet, USB was built with little electrical isolation to keep chips cheap.