Case Study · AI Product Design · Live at sceneone.net
SceneOne
“Democratizing Hollywood, one script at a time.”
An AI-powered screenplay feedback tool that gives every writer the same studio-grade coverage Hollywood charges $100 for — free, in minutes, graded against the same framework the industry uses. Live at sceneone.net.
Professional script coverage costs $30–$100 per read, graded against Save the Cat — the 20-year industry-standard framework development executives actually use. The writers who most need that feedback are exactly the ones who can’t afford it. They submit blind, get rejected without explanation, and never learn why.
The Problem
$100 to find out if your script is ready.
Hollywood readers evaluate structure, dialogue, pacing, conflict, and visual storytelling, then send notes. Writers with money get those notes. Writers without money just wonder why they keep getting rejected.
The Solution
The same feedback. Free. In minutes.
SceneOne grades every script against Save the Cat — page-specific, jargon-free, with a Fix It loop that closes right inside the app.
The Insight
Who this is actually built for.
“The truth about some people is that they are filmmakers who haven’t made their films yet.”
— Kevin Smith
I never made it through film school — the money didn’t work out — but I never stopped writing. That line is why SceneOne exists: most of the people it serves aren’t hobbyists, they’re filmmakers stuck outside a door that costs $100 just to knock on. We grade craft. Not taste. Never taste.
The Decisions
Every choice made for the writer, not around them.
Decision
Save the Cat as the Grading Framework
AI feedback without a documented methodology is just opinion. Save the Cat gives every score a traceable foundation — the same blueprint Hollywood uses — stated clearly in onboarding so writers know exactly what they’re being judged against.
High Priority
Decision
Creative Choice — Score Goes UP, Not Down
Breaking a rule on purpose is craft. Breaking it without knowing is something to fix. Only the writer knows which one it is. Tap Creative Choice and the note is logged, locked, and removed from scoring — so the score rises. No coverage tool has done this before.
Highest Priority
Decision
Plain English, No Exceptions
If the notes are hard to understand, the tool fails its mission. Every note uses the writer’s own lines as examples, with page numbers, in plain language. The test: could someone who never took a screenwriting class read this note and know exactly what to do?
High Priority
The Iterations
A few real fights, not one clean pass.
Fight 01
Generic Feedback → Save the Cat as Real Guardrails
Early feedback read like it could’ve been notes on any script. Rebuilding the grading prompt around Save the Cat structure as actual guidelines — not vague inspiration — sharpened the analysis immediately. Unexpected side effect: once the structure made gaps visible, I started noticing weak scenes in my own test scripts myself, not because the AI told me to but because the layout made the problems obvious. The tool didn’t write better scenes for me. It helped me see my own script better.
Fight 02
Hardcoded References → Real Page Citations
Some sections were pulling hardcoded data instead of referencing the actual script. Forced the prompt to cite real page numbers tied to the specific scenes it critiqued — no more evidence that wasn’t actually evidence.
Fight 03
A Sync Bug Between Writer and Exec Views
Writers could toggle “list on discovery page” on their side, but it didn’t reflect on the exec dashboard. Took several passes to trace down and fix.
Each fight taught me something I now check by default: don’t trust a model to infer structure, don’t imply real-time sync until you’ve verified both sides actually talk to each other, and generic output is usually a prompt problem — not a model limitation.
The Final UI
Nine screens. One cohesive system.
Upload → Coverage Report → optional Discovery Dashboard listing. A few things that make it feel built, not prototyped:
🗺
Pacing Heatmap
Every page scored and color-coded, orange to cyan. Hover any cell for what’s happening on that page. Dead zones are impossible to ignore.
🔧
Fix It Inline
Every flagged note opens a small scene editor. Revise, hit Re-Analyze, and the score updates in seconds — no re-uploading the whole script.
🎬
Logline + Story DNA
A pitch-ready logline and comp titles (e.g. “40% Whiplash · 35% Black Swan”) generated from your script. Editable suggestions, never verdicts.
👤
Co-Writer Credits
Add up to 3 co-writers on upload. Every collaborator is notified on upload, listing, and reader requests.
A two-sided marketplace runs underneath it: writers opt in to list their logline and Story DNA, verified industry readers (IMDbPro + company email + LinkedIn) browse and request access, and get a 14-day read window. No response within 72 hours auto-declines — scripts stay private by default. Three randomized cinematic loading sequences — Star Wars crawl, film countdown, clapperboard — turn the one moment a writer is captive into the product’s personality.
Production-hardened, not just demo-ready: server-side JWT enforcement on the admin panel, XSS-safe event delegation across the app, RLS-enforced 14-day access windows, and a live Supabase Realtime exec dashboard with zero polling.