notebooks8 so far, more in the ink
The notebooks
I keep notebooks for the things I want to understand. Drawings first, words in the margins, and always the urge to take the thing apart to see how it moves.
Each one takes a single question that has caught my curiosity and answers it with machines you can poke: switches to flip, clocks to step, caches to thrash. Read them top to bottom, or skip straight to the instruments.
What a CPU is, and how it works
On the clock that never stops Everything from NANDCounting with wiresFetch, decode, executeThe assembly lineFar from homeThe cacheBetting on the future open the notebook →What a GPU is, and how it works
On the patience of crowds A thousand slow brushesThirty-two in lockstepNever idleOne trip, not thirty-twoBring it in onceTensor coresGrid, block, warp, thread open the notebook →What a TPU is, and how it works
On a heart that beats in numbers A chip for one equationThe systolic arrayGood-enough numbersFeeding the beastCompiled, not scheduledA pod is one computer open the notebook →CPU vs GPU vs TPU
On three bets and one roof Three bets on siliconSame job, three runnersThe rooflineOne step, three chipsThe cost of movingWhich one, when open the notebook →The architecture of an ahead-of-time compiler
On turning words into motion Before you press runWordsGrammarMeaningOne name, one valueThe optimiserSixteen registersBytes, symbols, a binary open the notebook →How a large language model works
On guessing the next word Words into numbersOne job: the next wordA model you can trainMeaning as directionPaying attentionThe context window open the notebook →Anatomy of an AI agent
On a loop that uses tools A mind without handsThe loopTools are a contractA desk that fills upShapes of an agentKnowing when to stop open the notebook →How agents remember
On what to keep in mind A brilliant goldfishCutting the book into cardsMeaning as distanceFinding the right cardThe open-book examWhat’s worth remembering open the notebook →// more notebooks are on the bench