the map · the contract

The Long Climb

There are no shortcuts here, and this plan refuses to pretend otherwise. Most 'learn it in 30 days' paths keep you moving while quietly teaching you to repeat instead of reason — you finish with the feeling of knowing and none of the substance. This is the opposite. You build the machine from the switch upward, derive every number yourself, and write a great deal of real code, because understanding is a side effect of building and fluency is a motor skill. Year one lays an unshakable foundation across five interlocking paths. It is the base camp of a multi-year climb — and it is real.

Five paths, one understanding

They aren't five separate courses — they're one understanding with five faces. You meet the stack building the spine; you own it overwriting a return address in hacking. Linear algebra is dry until it's the forward pass of a net you wrote. Nothing is taught in isolation because nothing is isolated.

The daily engine

The same loop, every day. This — not the module list — is what actually makes you good.

  1. 1
    One spine lesson, to the bone

    Not skimmed. You leave able to predict a case the lesson never showed you.

  2. 2
    Write real code

    Fingers on keys. You implement it, you don't watch someone implement it. This is where fluency is actually built.

  3. 3
    One hard problem

    Math or algorithms, attempted with no hints first. The daily rep that builds the problem-solving muscle.

  4. 4
    Close the loop

    Read back what you built, log what you learned, and name what you'll predict tomorrow. Spaced review keeps nothing decaying.

The rhythm

Real hours, not thirty-minute snacks. The paths run in parallel and spiral — you don't finish one to start the next.

WhenCenter of gravityAlso runningDaily
Months 1–3Spine — bits → logic → architecture → assembly → C from the metalMath (proof & discrete) · the daily problem2–3 hr/day
Months 4–6Spine — memory, the heap, the OS · Hacking begins (binary exploitation)Math (linear algebra) · Code (data structures)3 hr/day
Months 7–9Hacking — exploitation & systems · AI — ML from scratchMath (calculus, probability) · Code (algorithms)3–4 hr/day
Months 10–12AI — deep learning & autograd · Hacking — advanced · capstonesEverything, deeper3–4 hr/day
Proof of range — what you can DO

Not “familiar with.” Able to do it, from nothing, and defend every layer.

Month 3

Read a small binary's disassembly and explain each instruction. Write C that compiles in your head.

Month 6

Reason fluently about memory — stack, heap, pointers. Implement any core data structure from scratch. Prove a theorem by induction.

Month 9

Exploit a vulnerable program you fully understand. Derive and code gradient descent. Implement classical ML by hand.

Month 12

Build a neural network and an autograd engine from nothing. Own a return address on a program you wrote. Read any small binary. The foundation is complete.

Why this isn't tutorials — the eight rules
01
Descend before you ascend

Never use a layer before you understand the one beneath it. Pointers after memory. Functions after the stack. Neural nets after gradients.

02
Build it to know it

You construct the thing — an adder from gates, an allocator, an autograd engine, a shell — instead of watching someone use one. Understanding is the side effect.

03
Derive, don't memorize

Every number comes from reasoning you can redo. You are never told 'the offset is 72'; you derive 72 from the frame, every time, until it's obvious.

04
Models over recipes

A lesson's goal is a mental model that predicts what happens next — not a sequence of commands that worked once.

05
Every idea used immediately

Nothing is taught 'for later.' Learn a register, use it the same sitting. Motivation is structural, not decoration.

06
Spiral

The three deep ideas — representation, memory, abstraction — return again and again at greater depth, from the first bit to the last exploit.

07
Productive struggle

Tasks make you assemble the answer from what you were just taught. Never a step you weren't given; the reach lives in 'stretch.'

08
Fluency is a motor skill

You write a lot of real code. You cannot read your way to fluency any more than you can read your way to playing piano.

The horizon

Year one is the foundation, not the summit. Real mastery is a three-to-five year climb — that's the honest number, and you're here for it. What comes after the foundation is specialization, chosen once you know enough to choose well: kernel internals, advanced binary exploitation, research-grade AI, compilers, cryptography, or a domain that pulls you. Those are deliberately parked. Depth first. Breadth is earned, never front-loaded.

start the climb — day one of the spine