Problem Miner

• 

  Problem Miner Test - Soap

Inspiration

Founders often start with an idea and then look for evidence that people want it. We wanted to reverse that process: start with problems people are already talking about.

Reddit contains thousands of conversations where people openly describe frustrations, workarounds, recurring difficulties, and things they wish worked better. Problem Miner explores whether those discussions can be systematically turned into useful problem-discovery signals.

What it does

Problem Miner searches Reddit using SerpApi and identifies posts and discussions that may contain real-world problems.

AI then analyzes the results to surface:

• complaints and frustrations
• recurring problems
• unmet needs
• existing workarounds
• patterns appearing across multiple discussions

The goal is not to generate startup ideas automatically, but to help founders find problems worth investigating further.

How we built it

We use SerpApi as the search layer for discovering relevant Reddit discussions.

Search results are passed through an AI analysis pipeline that extracts potential problems, summarizes the underlying pain point, and groups related findings so users can explore recurring themes rather than individual posts.

Challenges we ran into

The biggest challenge is distinguishing a meaningful problem from ordinary online complaints.

A single Reddit post does not prove that a problem is important or widespread. We therefore focused on identifying patterns, repeated frustrations, and evidence that people are already trying to solve the problem themselves.

Another challenge is keeping enough context from the original discussion for the extracted problem to remain meaningful.

Accomplishments that we're proud of

We created a simple workflow that turns Reddit search results into structured problem-discovery information.

Instead of asking AI to invent problems, Problem Miner starts with things real people are already discussing and uses AI to help organize and understand that evidence.

What we learned

We learned that problem discovery is different from idea generation.

The most useful signals are often not people explicitly saying, “I need a product for this.” They appear as repeated frustrations, awkward workarounds, unanswered questions, and situations where existing solutions are inadequate.

We also learned that search and AI work particularly well together: search provides real-world evidence, while AI helps interpret and organize it.

What's next for Problem Miner

Next, we want to improve how Problem Miner identifies recurring problems across multiple Reddit discussions and ranks them by the strength of available evidence.

We also want to make it easier for founders to move from discovering a possible problem to investigating it further through the original discussions and additional customer research.

Built With

• typescript