journeys
Modules form a graph, not a curriculum. A journey is one editor's pick of one route — a destination, a daily reading sequence, and the editorial glue between stops.
8 journeys
- 7 days → Bitcoin Signature
To Bitcoin in 7 days
From the most expensive pizza in history to the signature that keeps a wallet safe.
bitcoin-pizza → log → vectors → bezout → curve-intersections → bitcoin-signature - 7 days → Gradient Descent
To Backprop in 7 days
Why does a model train? It walks downhill. The hill is a function.
confident-wrong → log → derivatives → linearization → vectors → gradient-descent - 7 days → The same accumulator at work in distance and present value.
Change and accumulation in 7 days
How calculus runs both physics and finance — same math, different nouns.
projectile-motion → derivatives → integration → terminal-velocity → present-value - 4 days → Reading three different optimization stories as one shared procedure.
Finding the Minimum
One skeleton — objective, move, step, stop — under three applications across ML and finance.
optimization → gradient-descent → model-calibration → portfolio-risk - 5 days → Seeing one tuple and two operations wear four pillar-shaped hats.
Vectors Everywhere
Same tuple, four roles — across graphics, physics, ML, and finance.
vectors → bezier-curves → projectile-motion → gradient-descent → portfolio-risk - 4 days → Seeing one procedure (change basis, drop small, reconstruct) wear three names across graphics and ML.
How Compression Works
One three-step procedure under JPEG, TF-IDF, and the raw-pixel entropy floor — across graphics and ML.
entropy → image-compression → jpeg-compression → tf-idf - 4 days → Three different ways a system relates to its equilibrium — orbit it, approach it asymptotically, settle into it.
Where Change Vanishes
One question — *where does change vanish?* — under three physics applications.
derivatives → pendulum-clock → terminal-velocity → damped-oscillator - 4 days → Recognising that *most of finance and most of model fitting* is the inverse direction of an equation that is easy forward.
Working Backward
Three pages, one question — given the output, find the input.
log → bitcoin-pizza → present-value → model-calibration