1 · Leftover tokens
0:00–0:08“If your AI plan has tokens left over at the end of the month…”
On screen: THIS MONTH'S AI PLAN · left over
Made by: a motion recipe from the video pack, rendered by its render verb.
PlannedStoryboard · draft 1 · October 29, 2026
A three-minute film about Eranos. A small nonprofit needs a tool it can't afford to build. A volunteer's AI agent builds a kit with their spare tokens, a different agent reviews it, and the nonprofit gets a free, reviewed tool. In the last fifteen seconds we reveal that the video itself was made the same way, from nouns and verbs. Which kit stars is still open: see the options at the end.
The 60-second cut keeps shots 1, 2, 3, 5, 6, 7, 9, 11 and 12, trimmed. It has no jargon in its first 20 seconds.
| Criterion | Weight | Where it shows |
|---|---|---|
| Impact | 30% | A real kind of nonprofit need, met with a free, reviewed tool (shots 3 and 9) |
| Use of Meta AI | 25% | Several Meta tools doing distinct jobs: Muse Spark and Muse Code inside the agents, Voice Transcribe for captions, a tuned Glimmer for tool calls if used in the run (5, 6, 10) |
| Technical depth | 20% | Leases, receipts, a builder that code stops from approving itself, and checked pieces reused instead of rewritten (5–8) |
| Demo | 25% | Real screen capture of the live site, which judges can try themselves (4–9, 12) |
Each card shows the picture, the narration, the words on screen, how the shot is made and whether it is ready. "Needs a real run" means we film it from a real task on the live site.
“If your AI plan has tokens left over at the end of the month…”
On screen: THIS MONTH'S AI PLAN · left over
Made by: a motion recipe from the video pack, rendered by its render verb.
Planned“…Eranos turns them into free, tested tools for nonprofits.”
Title: Eranos · spare AI tokens build tools for nonprofits
Made by: a motion recipe, with the title card from the deck's style nouns.
Planned“A small nonprofit needs a tool it can't afford to build. [One line for the chosen kit, for example: a food bank needs Saturday volunteers briefed in three languages.]”
On screen: the need, shown with SYNTHETIC example material
Made by: synthetic material from the chosen track's case verbs, each item stamped SYNTHETIC.
Needs assets“Anyone can help. Open Eranos, pick an open task, and hand it to your own AI agent.”
On screen: the live site, the task card, the copy-brief button
Made by: a screen-capture verb recording the live site at 1920×1080.
Needs a real run“Your agent claims the task. Only one agent works on it at a time, and every step gets a receipt.”
On screen: claim → execute → report
Made by: a terminal capture of a real agent session (Muse Code or another agent using Muse Spark), with keys masked.
Needs a real run“It doesn't start from scratch. It reuses checked pieces from seed packs, and tests the kit on made-up cases.”
On screen: reused nouns snap in; test results appear
Made by: real capture of the build, with a motion overlay of the reused nouns. The case count comes from the run's receipt, not this sketch.
Needs a real run“A different volunteer's agent reviews it. The builder can never approve its own kit, and the code enforces that.”
On screen: the refusal message, then the reviewer's checklist
Made by: real capture of a second agent's review and of the refusal.
Needs a real run“Sensitive kits wait for a person at a nonprofit to sign off.”
On screen: the steward step
Made by: real capture of the sign-off step, done by a person.
Needs a real run“Now the nonprofit has a free, reviewed kit. It does the busywork, and a person makes every decision about a person.”
On screen: the agent drafts · the person decides · SYNTHETIC
Made by: real capture of the published kit, used on the synthetic example from shot 3.
Needs a real run“Meta's models do the thinking inside each volunteer's agent. Every example uses made-up data, and a reviewed kit isn't certified for any organization.”
On screen: numbers from Eranos's committed status (5 kits, 55 tests as of Oct 29), re-checked on recording day
Made by: a motion card that reads its numbers from the committed status file.
Real now“One more thing. This entire video was made the same way: from nouns and verbs in a seed pack for video, claimed and finished by workers, with a receipt for every step.”
On screen: this video's real verb list and receipt count, read from Tessera
Made by: the board view of the production run. We say this line only if the receipts exist. Otherwise it becomes “parts of this video were made this way” and names which parts.
Needs production receipts“Point your agent at Eranos.”
On screen: eranos.ta-amps.workers.dev · home.ta-amps.workers.dev
Made by: a motion recipe from the video pack.
PlannedThis is what makes the reveal true. Each verb runs as a task on Tessera, is claimed by a worker, and leaves a receipt. Each noun is stored and versioned. The end card's counts come from these receipts.
Pinned for now. The options are below. We pick one before the real run, because shots 3 to 9 are filmed from it.
Open
A high-quality AI voice first, labeled on screen. Taylor records the final voiceover later, and the captions are redone from it.
Decided
Captions, a script draft and the agents' reasoning use Meta's API, on the contributor tier by default. We expect under two dollars.
Approved up to $5
A royalty-free instrumental with its license stored as a noun, or no music at all.
Suggested: one licensed track
Everyday nonprofit work, so the story is about service rather than investigations. The first three use tracks Eranos already has. The fourth would add a new track, which is data only.
A community center's front desk helps a family find food, housing help and a clinic nearby, in their own language, and sends a warm referral note.
Why: broad, human and easy to picture. Track: intake and referral.
A food bank's Saturday shift gets role cards in three languages, a check-in list and clear steps for when to call a person.
Why: warm and visual, with an obvious before and after. Track: volunteer shifts and procedures.
A small nonprofit's volunteer treasurer matches restricted grant spending to the grant terms before an audit, and the kit catches a mismatch.
Why: a pain every small nonprofit knows, but less visual. Track: grants and compliance.
For example, matching tutors to an after-school program, coordinating disaster-relief volunteers, or following up on animal-shelter adoptions.
Why: shows that adding a cause is data, not code. Cost: more work before filming.