Made entirely with HTML Docs

Boring source material, rebuilt for understanding.

These are real engine outputs: live HTML scenes, final narration, captions, transcripts, source evidence, MP4 fallbacks, and editable project bundles.

Make one with your agent

01 / CODEBASE → EXPLAINER

Poll a live servo at 50 Hz

A coding-interview question becomes a visual lesson about timing contracts, monotonic clocks, pacing without drift, implementation structure, and the tests that prove it works.

Codebase evidence · narrated lesson · transcript

ElevenLabs Brian · measured cue timing

Inspect the open engine

Live HTML Player · turn sound on deterministic

Storyboard

Transcript

Sources

Studio

02 / TECHNICAL QUESTION → DOCUMENT + VIDEO

Extend the estimator to a robot arm

The detailed page can hold API shape and edge cases while the video builds the multi-joint mental model: one timestamp, vector positions, elementwise rates, and strict shape checks.

Detailed page · live video · worked example

ElevenLabs Brian · measured cue timing

Inspect the open engine

Live HTML Player · turn sound on deterministic

Storyboard

Transcript

Sources

Studio

03 / DOCUMENT → EXPLAINER

Measure velocity from a large CSV log

A flat interview prompt becomes a streaming data lesson: constant memory, adjacent-row state, invalid-sample handling, deterministic output, and boundary tests.

Source grounding · narrated video · captions

ElevenLabs Brian · measured cue timing

Inspect the open engine

Live HTML Player · turn sound on deterministic

Storyboard

Transcript

Sources

Studio

Complete learning courses

Learn by building the real thing.

Each course combines source-grounded lessons, narrated explanations, practice, feedback, and a concrete finish line—not a playlist of disconnected content.

ROBOT LEARNING · 12 LESSONS

Robot Learning, Gently

Move from the observation–policy–action loop through imitation learning, reinforcement learning, action representations, world models, and VLAs—then put the theory to work in a measured ANIMA-Kiwi build.

ETH Zürich theory × practical Kiwi capstone

Open the course

LANGUAGE MODELS · 11 LESSONS

Build an LLM from First Principles

Build the stack instead of treating it as a black box: scalar autodiff, language modeling, embeddings, training diagnostics, causal attention, byte-pair tokenization, a decoder-only Transformer, and honest evaluation.

Python · intermediate · build-first

Open the course

COMPANY BUILDING · 7 LESSONS

The Inference Company Founder’s Field Course

Choose a defensible wedge, run evidence-seeking customer discovery, build and benchmark a production-shaped serving MVP, model the economics, operate it under load, and earn the right to pursue custom silicon.

Advanced · technical founders and infrastructure leaders

Open the course

WORLD MODELS · 7 LESSONS

World Models: Prediction, Planning, and Physical AI

Trace the predictive loop from latent state and imagined rollouts through Dreamer, MuZero, JEPA, planning, physical AI, and VLAs—then build an evaluation matrix that separates useful capability from hype.

Advanced · AI engineers · source-grounded

Open the course

What ships with every project

The final video is only one artifact.

Reproduce it, inspect it, edit it, render it elsewhere, or hand an exact scene selection back to Codex or Claude.

Lesson page

Responsive HTML, detailed prose, diagrams, citations

Live composition

Deterministic HTML scenes with audio-master seeking

Caption tracks

Editable HTML captions, WebVTT, and SRT

Audio manifest

Segments, final word timing, waveform, pronunciation

Quality report

Cue contact sheets, repeated seeks, visual findings

Project bundle

Portable source, storyboard, scenes, evidence, versions

Guided Studio

Edit the teaching, not a pile of anonymous layers.

Select a semantic element at an exact timestamp, inspect its evidence and spoken cue, adjust structured properties, or attach “make this clearer” for your local agent to resolve.

Scene timeline

Posters, transitions, chapters, and reorder

Narration lanes

Waveform, cues, captions, and active words

Semantic selection

Stable data-html-video-id targets

Evidence-aware

Teaching job, sources, comments, versions

Your source is already enough.

Install one skill, point an agent at the material, and ask for a document, video, the combination, or a complete course.

Install the skill Open source