Where the network already talks
Every analysis produces a US Context Pack before it produces an opinion. Discovery then concentrates where the surrounding graph is populated enough for adjacency to mean something — and holds each claim against the date it was true.
A company is never evaluated in a vacuum. It is evaluated inside a jurisdiction.
Every analysis produces a US Context Pack first. Luna reasons over the startup inside that environment — not against a generic global template that treats a CMS reimbursement change and an EU directive as interchangeable weather.
Corporate evidence resolves against incorporation state and Delaware status, registered entity, funding vehicles, Form D filings and subsidiaries. Capital resolves against Form D, accelerator participation, syndicate activity, and historical rounds with implied valuation where derivable.
A health startup is never analysed against generic global healthcare trends. It is analysed against the reimbursement path that decides whether anyone can pay for it.
Twelve clusters, ranked by graph density.
First-generation discovery concentrates on the twelve densest US clusters — not to exclude companies elsewhere, but because network context resolves far more reliably when the surrounding graph is populated. A founder in a dense cluster arrives with legible adjacency.
SF Bay Area
Compiler, runtime and model-layer talent density is unmatched nationally.
- Tracked
- 4,820
- Adjacencies
- 8
- Graph weight
- 1.00
IP intelligence is the differentiator.
USPTO, WIPO, EPO, Google Patents, trademarks, open-source repositories, licensing relationships and the papers behind them. The engine does not ask whether a company says it has IP. It asks whether IP exists, what it protects, and who actually owns it.
- Does actual IP exist?
- What does it protect?
- Who is named as inventor?
- Was it assigned to the company?
- What is its remaining life?
- How broad are the claims?
- How easily can competitors design around it?
- Does it correlate with the shipped product?
- Is assignment recorded, or merely asserted?
- Did founders create related IP before incorporation?
- Is there university ownership risk?
- Could a former employer have a claim?
The three flagged questions are where diligence actually breaks. “Core technology was developed while the founder was employed at Company X” is a sentence no pitch deck contains and no generic analysis surfaces.
“They have thirty employees” is trivia. The path to thirty is the signal.
A 2.31× expansion in eight months, and the shape of it tells you when they raised, when they closed something, and what they decided to build. Applied across every observable series, this stops being a snapshot and becomes an early-warning system.
Eight monthly observations per series, all from public sources. Endpoint marked.