How does Negative Space generate startup ideas?

It embeds YC company descriptions into a vector space, finds the sparsest points, and asks Claude to write a concise startup description for those gaps. The output is then compared against real companies to ensure novelty.

Is the tool open source?

Yes. The embeddings, code, and a Claude skill are available on GitHub at https://github.com/nh66labs/negative-space.

Do I need access to Claude to use Negative Space?

The current implementation relies on Claude to generate the startup descriptions, so you will need an API key or a local deployment that can run the Claude model.

PromotedAdvertise

Ask AI

Negative Space

Startup ideas from the gaps between YC companies

ai startup ideation embeddings

NnoufalibrahimNov 3, 2026AI Shown on Hacker News by noufalibrahim Your product? Claim it

Visit

About

Negative Space is a generative tool that scans YC company descriptions, embeds them into a vector space, and identifies the sparsest points. It then prompts Claude to write realistic startup concepts for those gaps, producing ideas that fit the market yet remain untapped. The workflow is fully automated and can be run locally or via a simple web interface.

The case

Who it is for

Founders and product managers who want a systematic way to discover untapped startup ideas based on the YC ecosystem.

What is broken today

Finding genuinely new startup ideas often involves sifting through vast amounts of existing company data and relying on intuition or manual brainstorming, which is time‑consuming and may miss subtle gaps in the market.

What it does about it

Negative Space embeds every YC company description into a high‑dimensional space, blends related entries, then identifies points that are farthest from all existing data. It prompts Claude to write a concise description of a potential startup that could occupy that point, and finally measures the generated idea against real companies to ensure uniqueness.

Why not the obvious alternative

Unlike generic idea generators or market research tools, Negative Space uses a data‑driven embedding approach to pinpoint genuinely underserved niches within a well‑defined ecosystem. It leverages Claude’s language model to produce realistic, ready‑to‑pitch concepts rather than vague suggestions.

Details

CategoryAI

LaunchedNov 3, 2026

Built by the makerYes

Questions

How does Negative Space generate startup ideas?: It embeds YC company descriptions into a vector space, finds the sparsest points, and asks Claude to write a concise startup description for those gaps. The output is then compared against real companies to ensure novelty.
Is the tool open source?: Yes. The embeddings, code, and a Claude skill are available on GitHub at https://github.com/nh66labs/negative-space.
Do I need access to Claude to use Negative Space?: The current implementation relies on Claude to generate the startup descriptions, so you will need an API key or a local deployment that can run the Claude model.

Discussion

No comments yet. Be the first to give the maker feedback.