← November 3 class

AI Practicum · November 3, 2026 · Module 2 of 7

Behind the scenes: how this class was built with AI

Planning by voice, designing from examples, a storyboard-first video reel, AI voice, and one connector layer for every assistant.

8:33 in the recording20 min27 chapters

Watch

Full class recording · 1:23:10 · starts at 8:33

Chapters

Each chapter starts the class video at that moment.

  1. Start planning in a chat that already has context8:33
  2. Plan by voice: a live two-way conversation9:49
  3. Voice mode on commutes and walks10:46
  4. The first draft: 28 slides11:34
  5. Grok’s “Canvas” is its artifact format12:01
  6. The first canvas deck vs. Claude Design12:33
  7. Give it examples: start from templates13:05
  8. A prior deck built with Claude Design motion13:55
  9. One slide per area, then a PDF14:50
  10. A hero reel with Higgsfield16:40
  11. Disclosure, and the models inside Higgsfield17:11
  12. ElevenLabs: voices and Eleven Reader17:37
  13. Voice cloning: original vs. AI18:27
  14. Text-to-speech and transcription18:58
  15. Don’t connect every tool to every assistant19:38
  16. One connector layer across assistants20:08
  17. Missing a connector? Have the agent build it21:13
  18. For teams: observability and access control21:25
  19. It gets faster as it learns your usage22:28
  20. Q&A: connecting an assistant to Composio23:38
  21. Zapier: where it fits24:04
  22. Relevance AI for enterprise agents25:01
  23. Storyboard before you spend credits25:45
  24. Pick a direction and render26:32
  25. The first-cut reel26:46
  26. About a minute of work: good enough?27:15
  27. Use short video to break up long content28:02

From the class

What this module covers

Before the agenda, JP showed how he produced this class: planning with an agent that already had the context, iterating the deck from a template, generating a hero reel and voice, and routing every tool through one connector layer.

Try it

Put it to work

Video generation spends credits quickly. Storyboard and approve before rendering.

Ask IT before routing company tools through any third-party connector layer.

  1. Plan your next presentation in a chat, or a voice chat, inside a project that already holds the background material.
  2. Give the assistant one example deck or template you like, and ask for one slide per topic.
  3. Export a PDF and compare it with a first draft made without the example.
  4. Optional: ask for a six-frame storyboard of a 30-second intro and approve it before rendering any video.
  5. List the tools you want every assistant to reach, and check which ones a connector layer already supports.

One slide per area

The assistant that drafted the deck. What he asked, cleaned up as a prompt.

That's too many slides for a 90-minute session. Rebuild it as one slide per area, using the template I gave you, and create a PDF.

Design from an example

Any design assistant.

Use the attached template as the design reference. Match its layout, type and colors, and don't invent a new style. Flag any slide that doesn't fit the template.

Storyboard before you render

A video tool, before you spend credits.

Before generating any video, give me a six-frame storyboard in two different styles for a 30-second intro to [topic]. Don't render anything until I pick one.

Connect once

The assistant you want on the connector layer.

Connect to Composio so you can use my other tools through it. Tell me which tools are available and which would need a custom connector.

Check your work

You're done when

Tools

Mentioned in this module

Also on the class guide

Written before class

Between classes

WhatsApp and YouTube

Join the AI Practicum WhatsApp group

QR code for the AI Practicum WhatsApp group

Hive Research Institute on YouTube