From Research to Scale

The Playbook

The operating loop behind the factory: select ideas, build with agents, launch paid tests, measure unit economics, and feed the learning back into the next decision.

1

Selection Engine

934 researched ideas · weighted ranking · founder review · launch feedback

Every idea is evaluated before a single line of code is written. The dashboard ranks opportunities by buyer, market size, users, build days, blockers, LTV/CAC, and strategic fit.

📡
Enrich
Trends + CPC
🔍
Wide Scan
9 agents × 2
🔎
Gaps
Auditor agent
🎯
Deep Dives
8–15 agents
⚖️
Synthesis
Score & verdict

Three Attack Strategies

Predator

Beat weak incumbents with AI UX

Underserved

No AI-native solution exists yet

Provisioning

Automate an expert service at 1/10 cost

Output Per Niche

  • Structured research evidence
  • Weighted score and rank
  • Known blockers before build
  • GTM assumptions before ad spend
  • Lifecycle status from brainstorm to scale
See Selection Engine →
2

AI-Native Spec

2–4 hours · Agent-executable specification

Research output feeds directly into a detailed technical specification. Every spec is designed to be executed by AI build agents — with explicit work packages, acceptance criteria, and architecture decisions.

What the Spec Includes

  • Product overview with target personas
  • Feature list with priority tiers (MVP / Post-MVP)
  • AI assistant personality and behavior spec
  • Data model and API contract
  • Work packages (parallelizable units)

Quality Criteria

  • Every feature has acceptance criteria
  • AI interactions are specified with examples
  • Stripe tier pricing defined
  • PWA behavior documented
  • Spec passes agent readability review
3

5-Day Build

~100 AI execution contexts · reusable infrastructure · human review

Specialized agents work across frontend, backend, AI pipeline, infrastructure, testing, deployment, support, ads, and investor materials. A founder-owned review loop keeps judgment in the system.

Development Engine
Agent Workforce
Research, build, QA, deploy, growth
Architecture
DoneCo + Paperclip
Task graph and progress reporting
Next.js 15
App Router + PWA
Cloud SQL
PostgreSQL 15
Prisma
ORM + Migrations
Firebase Auth
Auth + SSO
Vertex AI
Gemini models
Stripe
Billing + Webhooks
Cloud Run
Serverless Deploy
Serwist PWA
Offline + Install

Every App Includes

  • Embedded AI assistant (streaming responses)
  • Stripe subscription (3-tier pricing)
  • PWA with offline support + install prompts
  • Firebase auth (Google, email, anonymous)
  • Mobile-first responsive design

Build Process

  • Lead agent reads spec, creates work packages
  • Specialized agents claim scoped tasks
  • Continuous integration via shared task list
  • Automated testing throughout
  • Human review on Day 4–5
4

Single-Command Deploy

Automated provisioning · Zero manual infrastructure

One command provisions the entire production stack. No manual setup, no configuration drift, no forgotten steps.

GCP Project

Dedicated project with billing, IAM, and service accounts configured

Cloud SQL

PostgreSQL 15 instance with database, user, and connection configured

DNS + SSL

Custom domain mapped, SSL certificate provisioned automatically

Firebase

Auth providers, Firestore rules, and admin SDK configured

Stripe

Products, prices, and webhooks created for 3-tier pricing

Cloud Run

Deployed with min-instances=1 to eliminate cold starts

5

$3K/mo Ad Test

~$3K/month per app · 1–3 month validation · AI-generated creative with self-correction

Every app gets a controlled market test: ~$3K/month in ad spend across Meta and Google for 1–3 months. Enough volume for statistically sound unit economics decisions. AI generates creative variants, scores them with a vision model, and pauses underperformers.

Meta
Facebook + Instagram
Google
Search + Display

Optimization Cadence

Day 4
Pause weakest ad sets
Day 7
First scale/stop signal
Day 10
Reallocate budget
Day 14
Final decision

Ad Infrastructure

  • Facebook Pixel + Conversions API (CAPI)
  • GA4 event tracking for full funnel
  • AI-generated ad creative variants
  • A/B test: hooks, benefits, urgency angles

Real-Time Market Data

  • Google Trends API for demand signals
  • Keyword Planner for CPC + volume
  • Competitor ad library monitoring
  • ROAS tracked in real-time across all apps
6

ROAS Scale/Stop Decision

One metric · Day 14 decision · No ambiguity

Build speed is useful only when it is tied to distribution data. ROAS, activation, retention, and objections decide whether the factory improves, pauses, or scales.

Stop
<70%
No path to profitability
Optimize
70%+
Signal detected — improve creatives
Scale
100%+
Every ad dollar returns more than $1
Growth Portfolio
150%+
Proven winner — scale further

Why ROAS, Not CPA

  • CAC and payback matter earlier than beautiful product demos
  • ROAS shows whether paid distribution can work
  • Activation and retention explain why it works or fails
  • Support and objections feed product decisions

AI-Driven Decisions

  • AI monitors ROAS in real-time across all apps
  • AI pauses underperforming ad sets at Day 4/7/10
  • AI generates creative variants to push ROAS up
  • Framework is fully automated — humans set thresholds
7

Scale & Compound

Automated improvement loop · Portfolio compounds over time

Graduated apps enter an automated improvement cycle: user feedback drives AI app improvements, which drive higher conversion, which drives higher ROAS, which justifies more ad spend.

Focused Products in Factory
$3K/mo Ad Test · AI Monitors ROAS
ROAS ≥ 70% → Optimize Creatives
ROAS ≥ 100% → Start Scaling
Growth Portfolio → Improve → Scale More

Automated Improvement Loop

  • AI generates + tests ad creative variants
  • User feedback collected → AI improves app
  • Better app → better conversion → higher ROAS
  • Higher ROAS → more budget → more users
  • Entire loop runs with minimal human input

Portfolio Compounds

  • Each graduated app is a self-sustaining revenue engine
  • Diversification across niches reduces risk
  • Shared infrastructure reduces the marginal cost of each new product
  • Learnings from one app accelerate the next
  • Revenue grows as portfolio grows