Retention Auditor
Analyzes any script against 368 validated retention patterns. Scores on 5 dimensions. Shows exactly what's missing and how to fix it.
Sample Script Input
"There's a secret the top 1% of creators figured out. And 99% never will. What if I told you that every viral video follows the exact same structure? Here's what I discovered after analyzing 500+ videos from MrBeast, Veritasium, and BibleProject..."
6.9
/ 10
B+
166 words|Est. 1:06
Hook Strength9.7/10
9 signal patterns detected
Curiosity Loop5.5/10
Could stack more questions before answers
Pacing & Density2.0/10
Weak — needs pattern interrupts every 90s
Persuasion & CTA9.2/10
Strong objection handling + CTA
Retention Engineering7.7/10
Good distribution of engagement hooks
How to Improve
Your curiosity loop is decent (5.5/10). To hit 8+: combine two patterns — stack questions AND use a delayed reveal.
Insert a pattern interrupt every 90 seconds. Use transitional phrases like "But here's where it gets interesting."
Your first line hooks (9.7/10). Maintain that energy through the foundation section.
Try It Live
Start the local server:
python3 sam-os/brain/auditor_server.py
Open http://localhost:8899 — paste any script, get an instant retention score.
Behind the Demo
Retention Auditor
593-node knowledge graph, 368 techniques, 5 dimensions
Server
Python http.server (stdlib, zero deps)
Pattern Detection
Regex signal matching against validated retention patterns
Data Source
9 NotebookLM creator notebooks, 151 insights
AIBoyz — Sam + Rohan, July 2026