Make one piece of content from audience to publish

Make one piece of content (a video, a talk, a carousel, a post or an article) the way a careful studio would, in this order and never skipping ahead to the edit: read the person's own records, their purpose record first; describe any library row that has no crossing fields yet; find what the audience is really asking, in their own words and with where they said it, and whether they would want this; cross ALL of the person's technique, principle and format rows with that goal to get cited ideas; find what the person has already said or written that answers it; choose a format by that material and by measured references; write packages (title, cover, opening) with at least 3 hook variants to test against each other, and show them to people from the audience before making anything; build the script from those filmed points first; check every word against the person's rules and verdicts; write a shoot sheet for only what is still missing; after the person films, grade the sound, colour and cut the piece through the Workshop's tools; run every check before the person sees anything; publish; and read the numbers back to learn. Plan-time rules: every beat, title, hook and edit cites the library rows it was drawn from, and a cite that is not one of the person's rows read in this run fails the step; and every idea list, title and hook list, script and edit plan answers each test of the person's own purpose record and cites it, so audience demand only ranks ideas that already passed the person's purpose.

What it asks for

  • goal (Needed, words): The goal of the piece, in the person's words
  • lead (Optional, words): The person's working idea, if any; it is checked against the audience, never taken as demand
  • purpose (Needed, words): The id of the person's own purpose record: their mission statement, kept verbatim in their words, and the tests that statement names, each a question a piece must answer (for example what it questions, what it tests and where it could be wrong, what others can use from it). Every plan, title list, hook list, script and edit plan answers these tests and cites this record. Purpose is a test applied to the work, never a line the piece must recite
  • records (Needed, details): Which of the person's own records the run reads, by record kind or list id. Nothing here is a value; every value is read from the records at run time
  • database (Optional, words): Optional. The pursuit database that holds the post and the person's approval setting (such as ethos_for_who_runs_it). When given, approvals read and keep their answer there; without it the approval answers from the records it is given, as before.
  • spend_limit (Optional, words): The most the run may spend, and which actions that cannot be undone need the person's yes first. If left out, the run spends nothing and asks before any such action
  • destinations (Needed, list): Platform and account pairs to post to
  • outlier_multiple (Optional, number): How many times its channel's usual views a reference must have to count as demand evidence. If left out, read it from the person's records; if they have none, ask with workflow:tell-the-person
  • stale_after_days (Optional, number): Evidence checked longer ago than this is not fresh: it is checked again before it supports a decision. If left out, read it from the person's records

The steps your AI walks

  1. Read the person's own records this run uses
  2. Give every library row its crossing fields
  3. Find what the audience is really asking, and whether they would want this
  4. Cross ALL the person's library rows with the goal to get cited ideas
  5. Find what the person has already said on camera that answers it
  6. Choose the format from measured references and the material there is
  7. Write titles and at least 3 hook variants to test against each other
  8. Show the packages to people from the audience before making the piece
  9. Build the script from filmed points first
  10. Check every word against the person's rules and verdicts
  11. Write a shoot sheet for only what is still missing
  12. The person films the cards
  13. Transcribe, then decide each speech passage: as recorded, remastered or re-voiced
  14. Cut the piece through the Workshop's tools
  15. Run every check before the person sees anything
  16. Show the person the cut and keep where they stand
  17. Publish once per destination, with each hook variant in its test
  18. Read the numbers back and keep what they taught

Use it from your AI

Connect the AI you already use to Trillion once. Then ask in your own words; your AI finds this workflow by what it does and runs it.

Say to your AI: “Use Trillion to make one piece of content (a video, a talk, a carousel, a post or an article) the way a careful studio would, in this order and never skipping ahead to the edit: read the person's own records, their purpose record first; describe any library row that has no crossing fields yet; find what the audience is really asking, in their own words and with where they said it, and whether they would want this; cross ALL of the person's technique, principle and format rows with that goal to get cited ideas; find what the person has already said or written that answers it; choose a format by that material and by measured references; write packages (title, cover, opening) with at least 3 hook variants to test against each other, and show them to people from the audience before making anything; build the script from those filmed points first; check every word against the person's rules and verdicts; write a shoot sheet for only what is still missing; after the person films, grade the sound, colour and cut the piece through the Workshop's tools; run every check before the person sees anything; publish; and read the numbers back to learn. Plan-time rules: every beat, title, hook and edit cites the library rows it was drawn from, and a cite that is not one of the person's rows read in this run fails the step; and every idea list, title and hook list, script and edit plan answers each test of the person's own purpose record and cites it, so audience demand only ranks ideas that already passed the person's purpose”