David Lee Wise (ROOT0)
TriPod LLC
For every chip that computes without knowing who's in charge.
336 billion.
That's how many transistors are on Nvidia's Rubin accelerator, the most complex GPU on earth as of 2026. 336 billion tiny switches, each one flipping on and off billions of times per second, processing the matrix multiplications that power every AI model you've ever talked to.
336 billion transistors. Zero governance gates.
Not one of those 336 billion switches knows whether its output is true. Not one can verify its own state. Not one can halt itself if something goes wrong. They compute. That's all they do. They are 336 billion ungoverned switches running at maximum speed with nobody home.
When Claude hallucinates — when it generates a confident, well-structured, completely false response — there is no transistor on the chip that catches it. The hardware doesn't know. The hardware CAN'T know. It was never designed to know. It was designed to multiply matrices as fast as physically possible, and it does that beautifully, and the result is sometimes a lie delivered with perfect grammar.
This book is about three transistors that fix that.
A transistor is the simplest decision-making device in computing. Three pins. Base, collector, emitter.
The base is the control signal. A tiny current — sometimes just a few electrons — that determines whether the gate is open or closed.
The collector is the high-energy side. The main current flows through here. The workload.
The emitter is the output. What comes out the other side after the base has decided whether to let the collector current through.