The Energy-First Turn: Computing as a Function of Energy
Electricity, not transistors, is now what schedules the next model. This course teaches you to price computing in joules and to tell a well-founded energy claim from a flattering one: where the joules actually go, why data movement rather than arithmetic is the bill, which substrates escape it, and what evidence would count as proof. You leave able to audit an efficiency claim rather than repeat it, and to tell a claim a cleverer algorithm can destroy from one it cannot.
▶ Start the course ← All coursesWhere this sits, and what moves it.
Binding constraint · Data movement. Physics permits arithmetic millions of times cheaper than we build; almost none of the bill is the arithmetic, and that gap is where every serious escape route is aimed.
Computing got cheaper by making transistors smaller, and energy was an operating detail. That electricity rather than lithography would become what schedules the next model was a forecast, not a constraint you could feel.
It is the constraint now, and this course prices it in joules. The discipline the course insists on matters more than any single number: a modelled figure, a stand-in measurement, and a metered one are three different claims, and most published energy comparisons quietly mix them.
A new substrate has to clear two bars at once -- beat the incumbent on work we already do, and do something the incumbent cannot do at all. Either alone has been cleared before by machines that did not survive. The course teaches you to do the destroying yourself, because a claim a cleverer algorithm can dissolve is the common failure mode and it is detectable in advance.
Every hard thing was impossible until the constraint that made it impossible was named. How we read a frontier →
The bill and the floor
Physics permits a computer millions of times cheaper than the one we build. Find the floor, then find where the money actually goes, and discover that it is not the arithmetic.
- L3How cheap is computing allowed to be?Before computing it: how many times the physical minimum do you think one multiply-accumulate costs?Compute the Landauer limit yourself and measure how far above it a real operation sits. The gap is the headroom the whole field is arguing about.→
- L3The bill is data movement, not arithmeticIn a workload with no off-chip fetches at all, what share of the energy is the arithmetic?Build the energy ledger of a small workload and find out which side it is actually on. This is the fact that decides what a cheaper computer must look like.→
Settle, don't shuttle
Meet the move that every serious escape route shares: treating computation as relaxation rather than as a sequence of fetches and multiplies.
Measure it completely
Learn the discipline that decides whether this paradigm survives: making a modelled number, a stand-in measurement and a metered one permanently distinguishable.
The bar a new machine must clear
Learn to tell a claim that a cleverer algorithm can destroy from one that it cannot, by doing the destroying yourself.