AIBoyz
Self-Improving Content Brain
Sam + Rohan • July 2026
What Are We Building?
One system that replaces content strategy, research, and scripting teams.
The Brain
Self-improving knowledge graph extracting from 9 creators across 8 dimensions each
The Pipeline
Extract → Store → Synthesize → Validate → Produce
The Moat
593 proprietary insight nodes. Zero external APIs. File-based, crash-proof, compounding.
The Ralph Loop
24/7 autonomous extraction via Task Scheduler. One notebook per cycle. Epoch 66.
How It Works
1
Extract
Deep mines 9 notebooks across 8 dimensions
2
Store
Entities to typed nodes, relationships to edges
3
Synthesize
Traverses graph for underexplored combinations
4
Validate
5-dimension gating, only PRODUCE passes
5
Produce
12-min unified scaffold scripts with prompts
Extraction Progress (5 of 9 Complete)
| Notebook |
Nodes |
Status |
| BibleProject |
160 |
Complete |
| Dan Romer |
293 |
Complete |
| MrBeast |
286 |
Complete |
| Veritasium |
241 |
Complete |
| Alex Hormozi |
149 |
Complete |
| Johnny Harris |
135 |
8/9 dims |
| Kurzgesagt |
— |
Queued |
| MagnatesMedia |
— |
Queued |
| Yes Theory |
— |
Queued |
Two-Tier Retention Engine MVP
Tier 1: Free — Script Audit Tool
Paste any script and get an immediate retention engineering score. 5 dimensions scored 0-10 against 368 validated patterns from our 593-node knowledge graph.
Leads with value — shows what's wrong but not how to apply the fix. The execution skill is the paid moat.
Tier 2: Paid — Brain-as-a-Service
Clients send their NotebookLM sources. Our brain generates retention-optimized scripts using client content + 9-creator knowledge graph.
Starter
$497/mo
4 scripts/mo
3 sources
Growth
$997/mo
12 scripts/mo
10 sources
Agency
$2,497/mo
30 scripts/mo
25 sources
ROAST Validation Verdict
GREEN LIGHT
5-agent adversarial council tested the retention engine concept:
- Contrarian: Free tool commoditizes IP. Moat must be in execution gap.
- Expansionist: Category creator — no one offers retention engineering as a service.
- First Principles: Core value prop validated — retention is #1 YouTube metric.
- Buyer: "I'd try you for $500 but I need proof — one case study."
- Deep Researcher: Gap confirmed — no competitor offers retention-engineered scripts with technique attribution.
48-hour validation test: Post free auditor to 3 communities, track 50 submissions or 10% click-through to paid tier.
Current Pipeline State
- 9 total scripts in pipeline
- 1 validated and ready for recording
- 8 scripts in draft from previous pipeline runs
- Auditor UI live at localhost:8899
Technology Stack
Python stdlib
NotebookLM CLI
JSON filesystem graph
http.server (zero deps)
Tailwind CSS
Windows Task Scheduler
PowerShell orchestration
Key Files
sam-os/brain/researcher.py — Deep extraction engine (v2)
sam-os/brain/memory.py — KnowledgeGraph, QueryHistory, InsightStore
sam-os/brain/topic_synthesizer.py — Novel topic generation from graph traversal
sam-os/brain/topic_validator.py — 5-dimension gating (PRODUCE only)
sam-os/brain/pipeline_executor.py — Dynamic topic, validation, script writer
sam-os/brain/production_manager.py — Pipeline state machine
sam-os/brain/retention_auditor.py — Free-tier audit engine, 5 dims × 368 patterns
sam-os/brain/auditor_server.py — stdlib HTTP server on :8899
sam-os/brain/brain_service.py — Paid-tier Client onboarding
sam-os/brain/run.ps1 — Ralph Loop orchestrator
sam-os/brain/knowledge/knowledge_graph.json — 593 nodes, 902 edges
The Ask (for Rohan)
- Distribution Partnership: Access to content creators for free audit tool beta testing
- Go-to-Market Insight: Right pricing for paid tier? $500-1000/mo?
- Role: Partner (build/sell) or Advisor (open doors)?
- The Big Question: Is retention engine MVP the right market wedge?
Home •
Demo
AIBoyz — Sam + Rohan, July 2026