Process!
experiments
app page
The New Kickstarter

## 🌟 Inspiration

We wondered what would happen if product creation didn’t begin with founders…
but with autonomous exploration.

What if agents could roam the space of possible products, test them with real people, and allow the best ideas to emerge on their own?
What if the world didn’t need to wait for a single visionary — but could tap into a continuous stream of evolving ideas?

Ad Infinitum was inspired by that question.

We built a system where AI agents generate product concepts, test them with real ads, learn from human response, and promote the strongest ideas into a public marketplace — a Kickstarter-like board where anyone can like, dislike, pledge, or even adopt an idea to build.

It reframes invention as evolution, not guesswork.
And it imagines a future where we explore product space together — humans + agents — discovering ideas we could never have found alone.

## 🚀 What It Does

Ad Infinitum is a fully autonomous product-discovery engine:

1. It generates new product ideas

Using LLM creativity, the system proposes multiple concepts, taglines, and narratives.

2. It creates and launches ad experiments

Auto-generates visuals (fal.ai)
Auto-writes ad copy (OpenRouter)
Deploys to Meta Ads with budgets and targeting

3. It measures real-world traction

Every ad acts as a “probe” into the market.
The system pulls metrics continuously:

CPL
CTR
Lead quality

4. It enriches lead data with Fastino AI

Each lead becomes a structured persona:

sentiment
interest level
segments
problem tags
feature requests

This lets the agent understand which humans resonate with which ideas.

5. It evolves strategy automatically

The orchestrator:

kills weak variants
clones promising ones
mutates copy/images
shifts budgets
promotes ideas that truly resonate

6. It publishes validated ideas onto a public marketplace

Winning products graduate to a Kickstarter-style landing page via Lovable:

hero visuals
pitch
like/dislike
pledge widget

This board becomes a gallery of autonomous ideas, waiting for humans to adopt or champion.

7. (Future) If an idea is funded, it generates a build spec for freelancers

The system can turn validated ideas into actionable build contracts.

## 🛠️ How We Built It

Raindrop

The orchestrator:

manages experiments
triggers ad creation
collects metrics
evolves strategy
promotes validated products

OpenRouter

product ideation
ad copywriting
landing page content
future builder specs

fal.ai

Image generation for ads and landing pages.

Meta Ads API

campaign/ad/creative creation
lead form integration
metrics pipeline

Fastino (GLiNER-2)

Transforms lead text into:

problem-tags
sentiment
persona segments
intent scoring

This makes performance qualitative, not just quantitative.

Lovable

Build-with-URL to create the Kickstarter-like product page instantly.

Backend API + lightweight DB

Stores product concepts, metrics, leads, profiles, and landing pages.

## 🧗 Challenges We Ran Into

Integrating multiple complex APIs quickly (Meta Ads, Fastino, fal.ai, Lovable)
Maintaining agent autonomy while still keeping control over safety and usefulness
Normalizing lead data from different sources
Creating adaptive ad strategies that weren’t random mutations, but meaningful ones
Prompt stability across copy generation, landing page creation, and profile enrichment
Time pressure — building an end-to-end growth engine in less than 24 hours

## 🏆 Accomplishments That We’re Proud Of

Built a fully autonomous discovery → validation → publishing loop
Created a Kickstarter-like marketplace of emerging product ideas
Successfully integrated Fastino to turn leads into structured intelligence
Achieved reliable automatic ad creation + deployment
Designed an architecture that feels like a proto-economy of ideas
Demonstrated a glimpse of how products might be discovered in the future

## 📚 What We Learned

Ads are astonishingly effective probes for product-market fit
Lead quality matters as much as lead quantity
Autonomous agents get dramatically better with real-world feedback
The boundary between “AI creativity” and “market truth” is fertile ground
The future of product discovery may be less about brainstorming…
and more about ecosystems that explore idea space continuously