Launching today

Synthetic Consumer Research
Test product ideas on AI customers before you build
1 follower
Test product ideas on AI customers before you build
1 follower
AI rates every product idea 3 or 4 out of 5. This open-source agent skill uses Semantic Similarity Rating (arXiv 2510.08338): AI personas answer in words, scored by meaning into realistic ratings. Get purchase intent, segments and objections.




Launch Team

CloudflareDeploy serverless functions globally in seconds
Promoted
Hi Product Hunt! I built this after reading a 2025 paper from PyMC Labs and Colgate-Palmolive (arXiv 2510.08338).
If you ask an AI "would people buy this? rate it 1 to 5", it says 3 or 4 almost every time. The paper's fix: let each AI customer answer in their own words, then score the answer by meaning against reference sentences for each rating. On 57 real product surveys with 9,300 people, that matched real results 90% as well as people match themselves when they retake a survey.
This open-source skill runs the whole thing in Claude Code, Codex, Cursor or Antigravity: it writes personas for your market, asks each one separately, and gives you purchase intent, results by age and income, and the main objections in their own words. There's also a free demo on Hugging Face.
It's a fast first filter, not a replacement for real customers. I'd love feedback, especially from anyone with real survey data to test it against.