Start a research
Find the problems
worth solving
Stage 1 · Triage
400
posts checked per run, in parallel batches
Each bar is one post that passed triage, height = pain intensity.
Stage 2 · Report
1
report, written only from signals that passed triage
Evidence
100%
of quotes are checked against the source text, unverifiable ones are dropped
- ◎ Problem signals
- ⚑ Counter evidence
- £ Willingness to pay
How it works
| Step | What happens | Source |
|---|---|---|
| Fetch | Reddit (several sort orders, pain-phrased queries, subreddit deep dive) plus Hacker News, Stack Overflow and GitHub, in parallel 4 data sources | 4 APIs |
| Rank | Every post is compared to your topic semantically; unrelated posts are dropped semantic ranking | embeddings |
| Comments | Comment trees of the top posts — that is where pain and buying intent live top comments per post | evidence |
| Triage | Each post is validated: real problem, pain intensity, buying signal and the verbatim quote quote verified against the source | verified |
| Cluster | Duplicate complaints are merged, so frequency means distinct users, not posts deduplicated | clusters |
| Report | One report built from the validated signals, including counter evidence evidence only | markdown |
Guides
Pain points
Find SaaS pain points
Where real complaints hide across Reddit, Hacker News, Stack Overflow and GitHub — and what turns one into a product.
Competitive analysis
Competitive analysis
Read what paying customers of the alternatives already wrote, and find the gap rivals keep failing to close.
Writing
All field notes
Methods and checklists on using public discussions as demand evidence: validation, pricing and counter evidence.