Clawdemy Lessons

Overview of the cross-validation lesson: learning outcomes, prerequisites, and time for train/validation/test splits, k-fold CV, and data-leakage traps.

Overview of the classification metrics lesson: what you'll learn about the confusion matrix, precision, recall, ROC and AUC, plus prerequisites and read time.

Overview of the bias-variance lesson: what you will learn, where it fits, the prerequisite to have first, and the time and difficulty before you start.

What the scaling-laws lesson covers: the power-law form, the Kaplan to Chinchilla shift, computing an optimal budget split, prerequisites, and read time.

Learning outcomes, prerequisites, and scope for the reasoning and RLVR lesson: the RLVR loop, GRPO, and how it caps the build-an-LLM pipeline.

Orientation for the post-training lesson: what SFT and preference tuning (RLHF, DPO) do, the prerequisites, and the technical-primer scope you can expect.

What the parallelism lesson covers: the three classic schemes, FSDP/ZeRO sharding, placement rules, and how 3D parallelism scales frontier training.

What this lesson covers: GPU kernels, why fusion is the single biggest performance lever, the Triton and XLA paths, and the FlashAttention case study.

Overview of what makes LLM inference expensive: the prefill/decode split, the KV cache, and the batching and quantization techniques, with prerequisites.

How GPUs and TPUs execute model math: SIMT cores and memory tiers, the physics behind arithmetic intensity, and the TPU's systolic-array bet on matmul.

An overview of how modern language models are evaluated: the four benchmark formats, why evaluation is hard, contamination defenses, and the layered stack.

What you will learn about where LLM training tokens come from: six source categories, open datasets, the raw-to-final funnel, and sampling weights.

What this lesson covers and the skills it builds: filtering, deduplication, mixing, and synthetic data, with prerequisites and the technical-not-legal scope.

What this lesson on vectors covers: the three views, the two defining operations, prerequisites (none), and read and practice times.

Overview of the spans and basis lesson: what you will learn, prerequisites, the light algebra involved, and read and practice time estimates.

Overview of rectangular matrices: read a shape to find input and output dimensions, tell embeddings from projections, and place column space and null space.

Overview of the matrix multiplication lesson: composition of transformations, computing products column by column, prerequisites, time, and outcomes.

What this lesson on linear transformations covers: a matrix as the record of where basis vectors land, plus prerequisites, timing, and outcomes.

Orientation to inverses, column space, and null space: what you will learn, prerequisites, and the three equivalent tests for whether a matrix is invertible.

Overview of the eigenvectors lesson: what you will learn, how diagonalization delivers the change-of-basis payoff, prerequisites, and time and difficulty.

A learning guide to the dot product: the two formulas, projection, duality, and its role in attention and cosine similarity, plus prerequisites and timing.

A guide to the determinant: what it scales, how to read it off the unit square, the ad-bc formula, why zero means collapse, and the product rule.

A brief on the 2D cross product: compute the signed area two vectors span, read its sign as orientation, and see it as the determinant you know.

A roadmap to the Cramer's rule lesson: prerequisites, the area-ratio derivation you will build, and what solving a 2x2 system as a determinant ratio covers.

Overview of the change of basis lesson: what you'll learn, the prerequisites (inverses, transformations), and how it sets up eigenvectors.