Show HN: Interactive 3D map of 3k YC companies by similarity (Three.js and D3)

Try it: We mapped 2,970 Y Combinator companies in 3D by product similarity - 57% of YC's entire portfolio since 2005. Jump right in and explore: pan/zoom around, click for detailed profiles, explore clusters. No signup required, works best on desktop.

Why

Sometimes it's hard to grasp the bigger picture and see the forest for the trees. This map gives an overall representation of the whole space - companies, clusters, and outliers (companies that don't have obvious similar counterparts). For example, there are three companies supporting innovation labs. Also it's fun to see how LegalTech is very compact but very far away cluster.

What stood out

A surprisingly large Government & Public Sector Solutions cluster and Accounting for SMBs.

LegalTech forms a compact but isolated cluster. And three lone outliers — all research lab equipment companies.

How it works

Input: company URL → Our own crawler extracts product signals from public pages. Representation: ML embeddings (multidimensional similarity). Projection: UMAP algorithm reduces to 3D coordinates. Clustering: hybrid algorithm groups similar companies; clusters auto-labeled by AI. Anthropic has similar approach in their Clio paper, if someone is interested. UI: three.js scene + d3 interactions (search/filter/hover). Visualization inspired by Music Galaxy (https://galaxy.spotifytrack.net/)

Scope & coverage

Dataset: 2,970 companies from YC's 5,200+ portfolio (Summer 2005 to August 2025). Coverage includes active companies plus many acquired/defunct ones we could access. We respected robots.txt; some sites missing or had limited crawlable content. UMAP is non-deterministic; we use consistent seeds but local neighborhoods can shift. Similarity ≠ endorsement; not affiliated with YC!

Looking for feedback

This sits on top of our B2B company mapping engine and we are launching more products. Would letting you drop in any company URL to see where it lands be useful? API?

Roadmap

Projecting any company in this space for comparison; Regular refresh as companies pivot and grow; API to get consteallation for your own project.

Learnings Visualization started as a small project with 5 moving points in 3D space and escalated to a full solution based on three.js and d3. This wouldn't be possible without AI assistance powering data collection, embeddings, and clustering.

Happy to dig into the technical details if folks are curious.