A social-listening system for finding high-intensity pain points

Most "market research" is just reading your own assumptions back to yourself. The fix is a system that mines real conversations for pain you can price.

The loop

Working AI-natively means treating the model as a research analyst, not a search box. Here's the prompt I open every session with:

You are a senior product researcher. I'll paste raw forum threads, reviews, and support tickets. Extract every distinct pain point. For each: quote the exact language, rate intensity 1-5, and note whether the person is already paying to solve it.

Feed it a batch, then cluster the output. The threads that score 4-5 on intensity and show existing spend are where you build.

Automate the collection

I pull sources with a tiny script so I'm never copy-pasting by hand:

npm run listen -- --subreddit=ProductManagement --limit=200

Everything lands in one file. The model does the reading. You do the deciding.