university levels of education Building a Lecture Pipeline I — IV
Building a Lecture Pipeline, How This Site Turns Text Into Courses — handmade, no ads, no sign-ups. Every lecture is kept whole and shown whole: nothing is summarised, so what you see is the lecture as it was written. 35,253 words across 54 lectures in six modules, from why the pipeline had to exist, through a byte-for-byte rebuild of the shipped Scientific Python corpus, to eight fresh datasets and a capstone from a blank file to a browsable course. Every code snippet was executed during the build and is shown with its real output. Twenty-one ways through the same text, forty-six palettes on a white default with dark a click away, twelve layouts and colours you can set yourself. Written from the code that actually runs this wing, including the mistakes: every bug described here was a real one, and the fix quoted is the fix that shipped.
- 001
The Goal and the Raw Material
Lectures 1 to 6. The Goal: From Two Text Files to a Finished Course, First Contact: Opening the Raw File in Python and Anatomy of the Raw File, on through the block model and the census habit. 4,563 words, every code snippet executed with its real output shown.
live - 002
Small Parsers, Many Examples
Lectures 7 to 16. The Boundary Scan, Regular Expressions From Zero and Building the Real Heading Pattern, then fences, nine bullet alphabets, the infinite loop that almost shipped, display math, tables and callouts. 7,178 words.
live - 003
From Raw to Fix
Lectures 17 to 24. The Audit, Stripping the Scaffolding and Banners Into Headings, then the piece splitter, the stitcher that turns 64 pieces into 30 lectures, and the fix contract: readiness, conservation, idempotence. 4,794 words.
live - 004
From Fix to Perfect JSON
Lectures 25 to 34. The production parser assembled, one lecture traced block by block, harvest and renumber, strip_md, the corpus and helper builders, the search index, validation — and the proof: the pipeline rebuilt inside lecture 34 reproduces the shipped Scientific Python corpus byte for byte. 6,036 words.
live - 005
Your Own Course From Nothing
Lectures 35 to 46. Aurora Science, a six-lecture course invented for this module: author it, parse it with forty lines, break it with ten mutations, re-deliver it badly and heal it, build its corpus and helper, wire it to a viewer, catalog it, ship it, and turn every bug into a test. 7,163 words.
live - 006
Eight Datasets, Eight Parsers
Lectures 47 to 54. Eight fresh miniature sources, one challenge each: verse, glossary tables, mixed math dialects, a damaged lab notebook, missing lectures, a source with no markdown at all, a two-delivery merge, and a capstone from blank file to browsable course. 5,519 words.
live - 007
The whole course
All 54 lectures, opened without a filter. 35,253 words, 1,387 blocks, 8 tables and 0 paired examples, with a glossary of 51 entries and a search index of 3,075 words. Twenty-one ways through the same text. Six read it — one lecture entire, an index of every lecture number down the left, a reader that hands you a single section at a time, a continuous scroll, an outline of the section headings, and a syllabus of the whole course. Eight take it apart — every Spanish line beside its English, the glossary lifted out of the tables, the tables on their own, flashcards Spanish side first, the practice questions, a drill that pulls a random example in either direction, and the objectives and closing summaries. Seven keep track — homework, cultural notes, an accent-insensitive search across every block, starred lines, a page of your own notes per lecture, a progress board, and a printable handout. Forty-six colour palettes on a white default with dark one click away, eight colour roles editable by hand, twelve page layouts from a narrow reading column through ruled notebook paper and two magazine columns to a monospaced terminal, seven typefaces with a separate face for the Spanish, and sliders for text size, leading and line length. The English can be hidden to read the Spanish cold, or the Spanish hidden to work back the other way. Progress, stars and notes autosave, with JSON export and import.
live