Tracing How an AI Plans Ahead
A hand-drawn research explainer turns an invisible AI process into a visible, changeable circuit through one worked example.
Field notes / why it works
A black shape resolves into a literal box by 0:01.5, making the opening abstraction visible before the narrator finishes naming it. At 1:11–1:16, word strips and empty response boxes turn a research claim into a prediction the viewer can follow; that interval contains the strongest replay peak. The changed circuit and alternate outcome at 1:43–1:56 provide a second, causal payoff.
Plate 01 / The format in one breath
A calm narrator starts with an invisible system framed as a tangible “black box,” then uses hand-drawn objects to show how a new way of observing it works. A single concrete example occupies the center, first showing an outcome and then changing one part of the mechanism to show cause and effect. The last 30 seconds explain why the finding matters and point to the underlying research.
Why it works
- The visual metaphor arrives almost immediately: a black shape enters at 0:00.75, becomes a small central box by 0:01.5, and supports the narrator's opening claim by 0:03.
- Abstract research becomes a stepwise, visible test. The sentence strips and empty boxes at 1:11–1:16 draw attention to a possible next word; the replay peak is 1:13.9–1:15.7.
- The example has a second reveal. At 1:41–1:56, a circuit is changed and the illustrated output changes, making the method feel consequential rather than merely descriptive. Another replay peak falls at 1:52.6–1:54.4.
Plate 02 / Format card
| Reference | |
|---|---|
| Platform · aspect · length | YouTube · 16:9 · 640×360 source · 24 fps · 2:55 |
| Pace | 17 measured shots, 16 cuts, 5.5 cuts/min; 18 graphic steps within shots, 11.7 total picture changes/min; mean shot 10.3 s, median 6.17 s. The 79.17 s example is one measured shot with many internal changes. |
| Script | 491 spoken words · 171 wpm overall, about 191 wpm during speech · 26 sentences averaging 18.9 words · no spoken questions |
| Voice | One low-register, expressive, clear adult narrator; explanatory and assured |
| On screen | Essentially 100% flat, hand-drawn motion graphics; no visible presenter, live-action footage, or software UI |
| Captions | No running subtitles; only purposeful labels, sentence strips, outlined word slots, and an end-card URL |
| Sound | Low tonal bed inferred from measurements; approximately 14 dB beneath voice; no clearly established cut-synced effects; −15.3 LUFS integrated, −1.9 dBFS true peak |
| Structure | Invisible problem → method → one worked example → intervention → implications → research CTA |
Pro / 13 plates + recipe
The first plates are open. The full dissection is Pro.
Get the timed shots, type and sound specs, and the runnable recipe.
Open the full blueprint- Plate 03 / The hook, frame by frame (first 3 seconds)Pro
- Plate 04 / Structure (beat sheet)Pro
- Plate 05 / Shot grammarPro
- Plate 06 / Visual style guidePro
- Plate 07 / Voice and deliveryPro
- Plate 08 / Sound designPro
- Plate 09 / PackagingPro
- Plate 10 / Script templatePro
- Plate 11 / Production planPro
- Plate 12 / PromptsPro
- Plate 13 / Edit recipePro
- Plate 14 / QA checklistPro
- Plate 15 / Make it yoursPro
- Recipe / runnable promptPro