ANC Media

AI Factory

An autonomous multi-agent platform that researches markets, creates digital products, and publishes them — end to end, without human intervention. Built natively on Google Cloud.

In production since May 2026 · Private beta

From market signal to published product — autonomously.

Creating a digital product traditionally takes a team days: research, design, production, quality control, publishing. AI Factory runs that entire chain autonomously. A fleet of 15 specialized AI agents finds the opportunity, validates it against live market data, produces the asset, verifies its quality and publishes it. A human makes exactly one decision per product: approve or reject.

15
Specialized agents
5
Pipeline stages
1
Human decision per product
How it works

Five stages, fully orchestrated.

Each opportunity moves through the same pipeline. Specialized agents hand off at every stage; humans approve at the gates.

01

Market Scout

AI agents scan marketplaces, social platforms and search trends for profitable product opportunities.

02

Strategy & Validation

A strategist agent scores each opportunity; only ideas above the quality threshold proceed.

03

Multi-Agent Production

Specialized agents generate the product: copy, design, audio, video, layout.

04

Quality Gate

Automated checks verify completeness, originality and platform compliance before anything ships.

05

Automated Publishing

Finished products are published to marketplaces and content platforms, with performance data fed back into the system.

Architecture

Under the hood

Every product flows through the same multi-agent pipeline. Each node is a specialized AI agent powered by Vertex AI; the only human touchpoint is the approval gate.

AI Agent (Vertex AI)
Human Decision Gate
GCP Infrastructure
Infrastructure

Built on Google Cloud

Cloud Run

ORCHESTRATION

FastAPI orchestrator, us-central1. Scale-to-zero: costs nothing between pipeline runs, spins up in seconds when a job starts.

Vertex AI (Gemini)

INTELLIGENCE

Three model tiers — Lite for routing, Flash for production, Pro for strategic decisions. Automatic fallback chain if a model hits quota.

Cloud Firestore

STATE

Two-layer data store: raw pipeline logs in one layer, distilled knowledge (what worked, what failed) in another. Agents read the distilled layer before every decision.

Firebase + GCS

DELIVERY

Firebase hosts the console and handles auth. Cloud Storage holds produced assets. GitHub Actions deploys to Cloud Run on every push.

Every agent call is a Vertex AI inference. Every state change is a Firestore write. Every produced asset lands in Cloud Storage. The orchestrator runs on Cloud Run. No external AI providers — the entire stack is Google Cloud.

Output

What the platform produces

The platform currently produces and publishes digital products across leading marketplaces and content platforms: printable planners and templates, Notion templates, and short-form video content.

Printable planners & templates
Notion templates
Short-form video content