I'm building a market simulator that tests your GTM with AI agents. Here's what I learned.

I spent 6 months building a SaaS product nobody wanted. The validation methods available — surveys, focus groups, 20 customer interviews — didn't catch the fundamental assumption that was wrong.

So I started building Sim-In-Silico: a platform that simulates your target market with AI agents before you invest in building.

Each agent has personality, memory, social connections. They reason through whether to adopt, get influenced by peers, form genuine preferences. The simulation outputs things like adoption curves, churn reasons, and which segments respond to which positioning.

What surprised me:

One test found that a free tier was so generous nobody ever upgraded. That's insight you'd normally burn months and real money to discover.

Another found that word-of-mouth drove 62% of late adoptions — meaning the product's shareability mattered more than its feature set for long-term growth.

The technical bet:

Every funded player in this space (Simile raised $100M, Aaru hit $1B valuation, Listen Labs raised $96.6M) went enterprise at $100K+/year. They simulate individual opinions — "would this person buy X?"

We simulate markets. Agents influence each other through social networks. Early adopters convince skeptics. Word-of-mouth cascades through real network topology. This produces emergent behaviors that individual-opinion tools structurally cannot capture.

And we're going PLG — free to start, $49/mo for full power. The self-serve play for the 99% of teams that can't afford enterprise synthetic research.

Where I'm at:

Stack is Next.js + Convex + Python simulation engine, mostly vibe-coded with Claude Code. Still early — 8 followers on X, just started posting on Reddit this week. Happy to share what's working and what isn't for distribution.

Site: https://siminsilico.com

What product hypothesis would you test first if you had a market simulator?