~/curriculum/record
Every load-bearing term the course teaches, and what this browser has done with each one. Three columns, and each is an event rather than a mark: met is the teaching module marked complete, recalled is a recall deck in it finished, and applied is a check in it answered. Nothing is scored.
A concept you have applied comes round again on a fixed schedule — after one day, then four, then twelve, then thirty — and reviewing it moves it to the next interval. Reading does not count; only answering does.
look first: which module carries the most arcs, and how far back does the longest one reach?
Part 1 — Foundations of Agents
| concept | met | recalled | applied | taught in | review |
|---|---|---|---|---|---|
| confabulation (hallucination) | · | · | · | 1.01 | — |
| context window | · | · | · | 1.01 | — |
| grounding gap | · | · | · | 1.01 | — |
| human-in-the-loop (HITL) | · | · | · | 1.01 | — |
| large language model (LLM) | · | · | · | 1.01 | — |
| systemic bias | · | · | · | 1.01 | — |
| token | · | · | · | 1.01 | — |
| capability floor | · | · | · | 1.02 | — |
| open weights | · | · | · | 1.02 | — |
| quantization | · | · | · | 1.02 | — |
| system prompt | · | · | · | 1.02 | — |
| agent | · | · | · | 1.03 | — |
| agent loop | · | · | · | 1.03 | — |
| block anatomy | · | · | · | 1.03 | — |
| doctrine | · | · | · | 1.03 | — |
| adversarial persona | · | · | · | 1.04 | — |
| critic loop | · | · | · | 1.04 | — |
| quality bar | · | · | · | 1.04 | — |
| council | · | · | · | 1.05 | — |
| register law | · | · | · | 1.05 | — |
| topology | · | · | · | 1.05 | — |
Part 2 — Agent Design and Evaluation
Part 3 — Mastering the Principles of LLMs
| concept | met | recalled | applied | taught in | review |
|---|---|---|---|---|---|
| activations | · | · | · | 3.01 | — |
| autoregressive | · | · | · | 3.01 | — |
| logits | · | · | · | 3.01 | — |
| residual stream | · | · | · | 3.01 | — |
| sampling | · | · | · | 3.01 | — |
| softmax | · | · | · | 3.01 | — |
| temperature | · | · | · | 3.01 | — |
| thought state (h) | · | · | · | 3.01 | — |
| tokenizer | · | · | · | 3.01 | — |
| unembedding table | · | · | · | 3.01 | — |
| cosine similarity | · | · | · | 3.02 | — |
| weights | · | · | · | 3.02 | — |
| attention head | · | · | · | 3.03 | — |
| causal mask | · | · | · | 3.03 | — |
| fact collision | · | · | · | 3.03 | — |
| MLP (multi-layer perceptron) | · | · | · | 3.03 | — |
| query, key, value | · | · | · | 3.03 | — |
| superposition | · | · | · | 3.03 | — |
| three sources (h_initial, h_ctx, h_prior) | · | · | · | 3.03 | — |
| in-context learning | · | · | · | 3.04 | — |
| prior (the) | · | · | · | 3.04 | — |
| chain of thought | · | · | · | 3.05 | — |
| scratchpad (thinking trace) | · | · | · | 3.05 | — |
| fine-tuning | · | · | · | 3.06 | — |
| retrieval-augmented generation (RAG) | · | · | · | 3.06 | — |
· not reached · ■ done · ▲ due for review
Clearing your browser's storage for this site clears the record. There is no copy anywhere else — the glossary defines every term above, and the syllabus says which chapter covers what.