This one started as a Zoom, not a dunk session. Same rule I use on athletes: if you only type it, you leave half the context on the table.
The reason we tell people to talk to the AI instead of typing is the same reason you should hit record. A video of you doing the work is more context than a bullet list. YouTube transcribes it. That transcript becomes an article. The article teaches the next agent. That is the same loop as a meta-article after a task — except you are in the footage.
Learn, Do, Teach is not a poster. On the Paul Ryazanov Knowledge Panel work I learned the inventory, the agent did the inventory, then I recorded the walkthrough so the next person does not start from zero. Visual learners watch. Readers get the article. The video also catches what the agent missed in QA.
Content → Checklist → Software. Skip to software and it fails. The video plus the meta-article is the content. Bold the sheet columns, paste the live URL back — that is the checklist. Then a Claude skill can repeat it. I am guilty of not recording enough. This post is me doing the third step on that Zoom instead of leaving it in Descript.
If you dunk, the equivalent is: do not leave 200 long-form videos and 80 podcasts in a private folder. Put them on a site Google can quote. I said that out loud at Dunk Camp. This is the contractor version of the same sentence.
Related: repurposing YouTube into articles, Descript + ChatGPT, one episode into 15 pieces.
About The Author
