Internet Atlas

Not sure what to build or how to validate your idea? Altas can find public demand for your idea/product. It can also come up with new ideas that have demand and help you generate leads.

Comment

Inspiration

Customer requests are scattered across public forums. I wanted a way to explore that demand visually, find a product worth building and identify the people to talk to first. Or for the rare few times I have extra tokens to experiment.

What it does

Internet Atlas helps founders investigate product ideas using real requests. One graph maps problems and their evidence. A second connects a chosen idea to the people who asked for it. One agent helps explore the data and draft outreach.

How we built it

The app uses React, TypeScript, Node.js and vis-network. It combines public APIs with attributed Reddit and X datasets. Local MiniLM embeddings power semantic search, while GPT 6 Luna extracts requests, reviews groups and investigates opportunities through Codex. Original authors, dates and source links stay attached. Outreach stays in drafts.

Challenges we ran into

Similar wording can describe different problems. Early results also favored isolated bug fixes over product opportunities. We added literal quote validation and independent membership review. Rate limits and archive timeouts required caching and visible source failures.

Accomplishments that we're proud of

The index contains 14,127 public records and 2,280 quote-backed requests. In one completed run, Atlas found 11 recent accounts asking to search past AI chats, built a graph of those requesters plus five historical accounts, and drafted outreach grounded in their original posts.

What we learned

More data helps only when the grouping is trustworthy. Every requester should support the same underlying problem. A public request can justify an interview; it does not establish buying intent. Showing the source makes that distinction easier to judge.

What's next for Internet Atlas

Test with founders whether the suggested cohorts lead to useful interviews. Use that feedback to improve opportunity ranking, expand community coverage and make incorrect groups easier to flag. Feedback would help me to build a better recommendation/ideation engine.

Built With

Share this project:

Updates

Submission history