# Sep 25, 14:13

![David's desk setup, where the session ran from Chip's drive mirrored to his monitor](photos/photo-a78481fc-58ba-477b-92c9-bd6e76194e27.jpg)

![David at work in his room, with the mic set up for the recording](photos/photo-df5a81fe-399c-450e-92f1-fa0d299b7925.jpg)

## Overview

Chip visits David Avalon, an award-winning director and producer, to see how David's tool Story Sherpa (now renamed Box Box Story) turns rough ideas into a commercial script. Chip brought three treatments that ChatGPT wrote for a 60-second, AI-assisted cinematic online commercial for his friend Anthony's product, New Normal Security, and its scanner, NewScan. David fed the treatments and the website into the full Story Sherpa workflow. It extracted insights, scored each treatment against its rubrics, merged the best parts into a hybrid that scored much higher, and wrote a beat-by-beat script about a vibe coder named Jordan. The two refined the "doubt" beat and the end card together. The session ended with a finished script saved as Markdown in a new "Anthony new" folder.

## Key points

### The commercial and its story

- The product is a security tool for vibe coders. Story Sherpa's core truth: "The ad isn't selling a scanner. It's selling a permission to ship."
- StoryBrand casts the developer as hero, NewScan as guide, and the coding agent as the hero's weapon.
- The villain is "the shrug": the coding agent answering "are we secure?" with "I recommend a comprehensive security review."
- Problems: external, "he can't tell whether the app is safe"; internal, "am I about to be the guy who leaked his users' data"; philosophical, if AI changed how software gets built, security shouldn't still need a different workflow.
- Plan: 1) hand it to your agent, 2) NewScan finds what's real, 3) fix it and ship. The transformation runs "from hoping it's secure to shipping with evidence."
- The script follows Jordan, late twenties, at 2:40 a.m. with a "Ship it" sticky note, hovering over "Deploy to production." The light shifts from cool night to warm dawn as subtext.
- Story Sherpa flagged that the website speaks mostly to pen testers and security teams, while the ad targets vibe coders.

### Scores and the hybrid treatment

- StoryBrand rubric (out of 35): treatment one 25, treatment two 22, hybrid 32.
- Four Words, Five Lines rubric (out of 45): treatments 30, 35 and 38; hybrid 41.
- Overall, the treatments scored about 70–71; the hybrid reached 91.
- The rubric shows red X, yellow warning and green marks per beat, so you can compare versions by number rather than gut feel.

### How Story Sherpa works

- It has five parts: insight extraction, StoryBrand, Four Words Five Lines, the story rubric, and the emotional index. A fifth, top-level file ties them together, "like Claude MD."
- To install it, add the Markdown skill files to your own repo. It works with Claude, Copilot, Gemini, Grok or any LLM.
- Four Words (from Ted Sorensen, JFK's speechwriter): clarity, brevity, charity, levity. Five Lines include headline, front line and bottom line.
- The bottom line is not a sub-headline. It does two separate jobs: an emotional close (the hero's win, usually shown) and a credibility anchor.
- The emotional index weighs rhetorical appeals: pathos 50%, then logos and ethos, with kairos (timing) at only 10% but flagged first as the "ignition."
- David calls it "the algebra of storytelling": he can teach the algebra, but not the art.
- It also teaches. You can ask what any term means ("the doubt," "Kairos") mid-session.
- It can analyse an existing ad (e.g. a Coca-Cola spot) using a local tool such as WhisperX to describe key frames and transcribe voiceover.

### Refining the script live

- The "doubt" beat (seconds 14–22) is "the hinge of the whole spot." Story Sherpa admitted a sub-second profile flash would not read.
- It offered three fixes: the wrong face, a customer email, and a key that opens every door. They chose "the wrong face": "Logged in as Jordan" above a stranger, Maria Lopez, with her address and card. The flashes drop from three to two.
- Chip disliked the end card "Your agent built it. Now let it find what's real." ChatGPT's line won instead: **"Your customers trust you with more than their password. NewScan. Free."**
- Chip and David agree ChatGPT writes headlines and taglines better than Claude, while Claude is stronger at code and analysis.
- Both stress that the script matters most. Fixing a gap on set means another day rate of about $1,600 or more.

### Background

- David and Chip worked at Human Code, an early-1990s interactive media company in Austin that made CD-ROMs and early websites and built its own no-code tool, CD Factory.
- Story Sherpa came together about a year ago as LLMs spread, but draws on 30–35 years of design, information architecture, filmmaking and commercial work. *Save the Cat* was the turning point.
- David says a Human Code insight session like this would have cost about $7,900; Story Sherpa did it in about five minutes.

## Action items

- Integrate ChatGPT into the workflow locally, since the cloud chat can't call the "question codex" skill (possibly this weekend).
- Add a Markdown version of the production script to Story Sherpa's output.
- Add a podcast or mulligan video showing how to use Story Sherpa to the /next list.
- Consider a mulligan with screenshots and a couple of avatars built from this session.
- Record the next stage as a separate session.
