What you get
With a seat, you get the course and the tools we made it with. We guide you along the way, until using AI in an advanced way feels ordinary.
How we made the films
Each lesson is drawn live in the browser on one stage. A beat is a list of steps, and the same code plays it and draws its storyboard frame. You get the engine, the storyboard board we comment on, and the runner that hands each comment to a coding agent.
How we made the narration
Voice is built per lesson with ElevenLabs, from our direction and pronunciation files, and timed to the picture.
How we built this site
Colour, type, spacing and motion compile from one design system: colours with measured contrast, a grid that decides the layout, and motion timings worked out from what moves.
How we make marketing images
Stills and mockups start from written prompt sheets for an image model. The intro is a 3D scene of our mark, rendered frame by frame in a headless browser and cut with ffmpeg.
Our skills and rules
The written rules our agents follow for copy, interfaces, video, research and checking work, and the interface design skill we use on every screen.
All our sources
The research behind each lesson, every source marked read in full or from a summary, and dated copies of over a thousand official help pages from ChatGPT, Claude and Microsoft Copilot.
Guidance along the way
The practice chat unlocks one control per lesson, so each new feature arrives when you know what it is for.
Free, no sign-up
Prompt trainer
Articles
- Attention and virality: what the evidence says
For a short explainer, the best-supported levers are learning ones: ask viewers to predict before the answer, and name a common misconception before correcting it. 67 sources.
- Slop: the tells, the tools, and slop grenades
The current giveaways of AI-made writing, websites and video, and the tools people use to steer models away from them.
- Slop examples, line by line
Each row breaks one slop rule, with the clean line we used in the course beside it.
- Every tip in the course, checked
82 tips checked against papers and official docs: 53 hold, 28 need a caveat, one was dropped.
- Cinematography for explainer films
Staging and motion rules for the films, each with a test you can run on a single still.
Voices we listen to
- Ethan Mollick
Co-writes Wharton’s Prompting Science Reports, which test common prompting advice. We checked our tips against them.
- Andrej Karpathy
His tokenizer lecture explains why models miscounted letters, and he popularised the jagged frontier.
- Simon Willison
Writes plainly about prompt injection and what can go wrong when an AI reads your files and acts.