Crowdmind is a full simulation environment, not a chatbot with a marketing label. Each AI persona has demographics, psychographics, a personal backstory, communication style, and a baseline disposition (skeptical, neutral, enthusiastic). Personas are built to have real objections and reasons to say no, not just agree with whatever you show them. Unlimited personas, any scale. Generate 5 or 500. Organize them into panels, batch-generate more anytime without starting over.
No reviews yetBe the first to leave a review for CrowdMind AI
Maker
📌
Hey Product Hunt 👋
By building my Startup i realized I needed to test pricing and pitch copy on people before burning real relationships and real time on conversations that might just fall flat.
So I built the thing I actually needed, a way to simulate dozens of AI personas with real backstories, real skepticism, real objections, and get an honest opinion before I ever talk to an actual human.
A few things I care about a lot with this project:
It's fully open-source. No black box, no "trust us" — you can see exactly how personas are generated, how the anti-sycophancy prompting works, and how the confidence/diversity scoring is calculated.
It's local-first. Your ideas, your unreleased pricing, your pitch decks stay on your machine. Only the LLM call itself goes out, and you choose the provider (Claude, GPT, Gemini, OpenRouter, or fully local).
It doesn't pretend to replace real research. This is a fast, honest first filter, not a substitute for talking to actual customers. If anything, it's meant to make you more ready for those real conversations, not avoid them.
Report
Would love to see a way to upload your own past customer interviews or survey responses so the personas can be calibrated to feel even closer to your actual audience. That kind of grounding data layer would make the simulations way more trustworthy for real product decisions.
Congratulations on the launch! Since it's open source I can see the generation logic directly, which is the right call. The question the code can't fully answer for me: in practice, do large panels actually hold disagreement, or do you find they flatten toward a consensus voice as the panel grows? Curious whether you've measured that.
Would love to see a way to upload your own past customer interviews or survey responses so the personas can be calibrated to feel even closer to your actual audience. That kind of grounding data layer would make the simulations way more trustworthy for real product decisions.
@mitmarazrqe actually it can be done
Congratulations on the launch! Since it's open source I can see the generation logic directly, which is the right call. The question the code can't fully answer for me: in practice, do large panels actually hold disagreement, or do you find they flatten toward a consensus voice as the panel grows? Curious whether you've measured that.