FROM FIRST PRINCIPLES TO SYSTEMS

How language models actually work

From “what is a parameter?” to streaming a 2.8-trillion-parameter model off disk on an 8 GB laptop. Every claim cited, every lab grounded in real numbers, nothing behind a login.

8
finished tracks
79
lessons
2
EN + PT-BR
  1. INTUITION
  2. MECHANISM
  3. EVIDENCE
  4. SYSTEMS

SECTION 01 · TOKEN TO SYSTEM

LEARNING PROTOCOL

Every lesson is a complete investigation

You do not just consume content. You follow the mechanism, test the idea, and prove you can explain it.

  1. 01Concept

    The actual mechanism, with no hand-waving.

  2. 02Analogy

    The same idea in plain language.

  3. 03Lab

    Real numbers, computed live.

  4. 04Teach-back

    You rebuild the idea in your own words.

  5. 05Quiz

    The right answer comes with the reason.

A lesson finishes only when your explanation has substance and the quiz is right. Your writing stays in this browser.

CURRICULUM SURVEY

The complete path, layer by layer

Eight published tracks form one continuous descent, from foundations to efficient inference.

  1. 01Foundations · 8 lessonsOrientation & FoundationsBuild an accurate mental model of language models and the mathematics needed for everything that follows.Enter track
  2. 02Foundations · 7 lessonsFrom Text to TensorsSee how text becomes tokens, vectors, and numerical structures a neural network can process.Enter track
  3. 03Foundations · 9 lessonsNeural Network FundamentalsLearn how neural networks represent functions, measure error, propagate gradients, and improve through optimization.Enter track
  4. 04Core · 5 lessonsSequence ModelsFollow the path from recurrent state to encoder-decoder models and attention, the immediate ancestors of the Transformer.Enter track
  5. 05Core · 14 lessonsThe TransformerAssemble the architecture behind modern LLMs, beginning with self-attention and its query, key, and value operations.Enter track
  6. 06Advanced · 13 lessonsPretraining at ScaleDesign the objectives, data, optimization, parallelism, and cost model behind a serious language-model pretraining run.Enter track
  7. 07Advanced · 11 lessonsPost-training & AlignmentTurn pretrained predictors into useful assistants through instruction tuning, preference learning, reinforcement learning, and distillation.Enter track
  8. 08Advanced · 12 lessonsInference & EfficiencyUnderstand how trained models generate text and how decoding choices trade off diversity, coherence, latency, and cost.Enter track

More tracks are being written. Everything published is complete.

DEPTH WITHOUT SURVEILLANCE

No account. No paywall. No tracking.

Your progress belongs to your device. The code and curriculum remain open.

Start the descent