I built a simulated community with AI personas to validate+test ideas quickly, and i want to know what the IH community thinks

Been a while, IndieHackers!

After exiting copyscouts, building 3 more (and failing), I realised this:

Early idea validation and testing + community feedback is...

1. Hard
2. Time consuming

So I built NicheSim: an AI-powered community simulator that lets you test an idea, product, or piece of content inside a synthetic community before you ship it to real humans.

It’s basically: “What would a real Reddit/Discord/Slack channel do with this?” except you get it in minutes.

What NicheSim actually does

You describe a niche in plain English, like:

“solo founders doing B2B SaaS”
“web3 gaming people who’ve been rugged too many times”
“fitness creators selling coaching”

Then NicheSim:

1. generates 5+ distinct personas (skeptic, early adopter, newbie, veteran, etc)
2. creates multiple channels like a real community
3. spins up 80+ messages that feel like ongoing chatter, not a single chatbot monologue
4. lets you post into the community and watch 3–6 personas respond in real-time (streaming)
5. adds a simple analysis layer (themes, sentiment, engagement, “what are people actually objecting to”)

The key thing: it’s not “one AI voice pretending to be everyone.” It’s a mix of personalities that disagree, misunderstand, nitpick, and occasionally say “nah” with zero explanation. Like real life.

Why I think this matters:

• Most “validation” is you getting emotionally attached to the first positive response you hear.

Real communities have: (to start)

• the person who’s instantly suspicious
• the beginner asking obvious questions
• the veteran who’s seen 12 versions of this
• the practical one who only cares about integration and exports
• the early adopter who wants to try it now

You need all of those reactions early, not after you’ve burned weeks building.

The realism stuff I cared about (because AI loves to sound like a blog post)

I spent a stupid amount of time forcing it to behave like an actual community:

1. Message length distribution (some replies are literally “this”, others are a paragraph)

2. Sentiment heterogeneity (not everything is positive, because reality isn’t)

3. Engagement patterns (70% of messages have no reactions because most people don’t react to everything)

4. Persona consistency stored in the DB, referenced every generation so the skeptic stays a skeptic - but can eventually be convinced

5. And when you post, it generates responses sequentially so personas can reply to each other, not just you. That’s what makes it feel like a thread, not a helpdesk.

The most interesting part is not “does the idea get praised.”

It’s:

1. what the skeptic attacks first
2. which questions repeat across personas
3. what people assume you mean (usually incorrectly)
4. which phrasing triggers confusion vs interest

If you build products, write content, run community-led growth, or just like poking holes in things:

Try NicheSim on something you’re about to ship.

Post your idea and see what the personas do.

Tell me what feels real and what feels off.

I’d love feedback on:

• what niches it performs best on
• where it breaks
• what you’d want exported
• whether the analysis is useful
• pricing expectations (I’m still experimenting here)

And selfishly: if you find it interesting, share it. I’m trying to see if this has “people tell other builders” energy or if it’s just me being obsessed with validation mechanics.