Repository documentation

Literate AI instructional courses

Start here: the continuous video play

Watch or download: Eight Hours to Make Coffee.

Sam, a skeptical C++/Make engineer, defends an eight-hour CI loop while LitAI moves useful feedback earlier without weakening required CI, independent review, native qualification or release gates. Twelve animated chapters use two approved NVIDIA Magpie voices, 86 visible captions and five source-backed project case studies with explicit limits. Running time: 12:04.896.

SRT · WebVTT · Production and limitations · Script and annotations · Authoring sources · Media receipt · Approval and hashes

The film is an illustrated walkthrough, not a recording of the five projects running. Its retained fresh evidence is a read-only adoption plan over an unchanged fixture; it does not claim apply, build, native execution or an eight-hour-to-minutes benchmark. Report timestamped feedback in the public repository's issue tracker.

Earlier cuts, retained for comparison

The animated onboarding authoring pilot adds articulated Sam/LitAI characters, an offline Motion/SVG companion and recorded NVIDIA Magpie voice auditions. The user approved the pilot voices and directed moving cut. That approval established the treatment; the separate full-film approval is bound to the published replacement's exact media hashes.

The superseded 1.2 film, Settle the Argument. Before CI., remains available with its chapter timestamps.

The cuts below are superseded creative drafts, not the recommended viewing path. The creative brief and accepted audition package record how the project-owned story evolved without prescribing a cast to children.

Three short lessons pair LitAI, a female guide, with Sam, a skeptical C++/Make engineer. Each includes two-voice narration, embedded English subtitles, and downloadable SRT/WebVTT captions. The original character artwork prompts are retained with the owned PNG assets: LitAI and Sam.

These are edited transcript replays from real sessions, not raw screen recordings. The worker lesson proves local execution and explains SSH setup; it does not claim a remote-worker run. The brownfield example actually adopted TinyXML2 and reached the retained stage after fixing a CTest parser defect. Read evidence and limitations before treating the examples as release qualification.

Rebuild or refresh

Media integrity receipts: greenfield, workers, and brownfield.

Use the Literate AI revision containing this package (the new commands are not in older wheels). On macOS with ImageMagick, ffmpeg/ffprobe, and the two installed voices, run each manifest through the standard CLI:

litai video plan docs/courses/01-greenfield-first-project.json
litai video build docs/courses/01-greenfield-first-project.json \
  --output _build/courses --font /path/to/Arial.ttf --narration say
litai video verify _build/courses/01-greenfield-first-project-UNIQUE/video-result.json \
  --manifest docs/courses/01-greenfield-first-project.json

Repeat for the other two manifests. After verification and human visual/audio review, copy the MP4, SRT, VTT and receipt into the corresponding course directory. Builds preserve previous outputs. Change the dialogue and evidence to refresh a lesson; supply recorded audio for portable narration or screen recordings for new coding demonstrations. See video authoring for supported backends, manifest fields, tools, integrity and publication boundaries.

1.3.1 animated replacement

Eight Hours to Make Coffee expands the approved Sam/LitAI treatment into a skeptical, source-backed pre-CI argument with five project case studies. The 12-chapter script, animation annotations and offline authoring/capture system are authored. All 86 approved-voice turns have been rendered into the published 12:04.896 film, which passes deterministic chapter QA and independent continuous-media verification. Fresh scoped evidence and complete listening, factual, accessibility and normal-speed playback approval are recorded against exact identities. The previous 11:52 film remains available as historical course evidence.