Methodology

The evidence ladder: company-reported → publicly shown → independent support

Every piece of evidence about a company is categorized into exactly one of three buckets, and these buckets are never blended on a single feed.

• Company-reported — what the company says. Press releases, marketing copy, interviews where the company is the source. Recorded for completeness, never weighted as independent support.
• Publicly shown — what has been publicly demonstrated. Demos, conference presentations, third-party visits. Independent of company narrative but not yet independently tested.
• Independent support on record — third-party tests, peer-reviewed papers, government grants, certifications. We place evidence here only when it is third-party AND independent AND published. This rule is enforced automatically — nothing reaches this level without all three.

Score dimensions

These roll up into a single Coverage Score (0–100) — a summary of how much reviewed, source-linked material we have assembled on a company. It is not a valuation or a rating of whether the technology works. Six model-estimated areas feed it, each on a 0–100 scale:

• Technical Progress — how much engineering/technical activity is on record. Not an assessment that the engineering is sound.
• Independent Evidence — how much arms-length, third-party corroboration exists. Not a verdict that claims are validated.
• Commercial Progress — movement toward market. Not proof of revenue or a shipping product.
• Funding & Partnerships — funding and partnership activity on record. Not audited financials.
• Activity Trend — how active a company has been recently. Not a prediction of outcome.
• Claims vs. Evidence — how far a company’s own claims run ahead of independent evidence (subtracted). A coverage gap, not an accusation.

Three further areas are derived from those above:

• Technical Maturity — technical progress plus demonstrated artifacts. Not a certification that hardware works.
• Commercialization Readiness — commercial progress plus funding & partnerships. Not proof of a product.
• Institutional Credibility — independent evidence plus recognized-institution involvement. Not an endorsement.

Source Confidence

A separate 0–100 score answering: how much independent, recent, and consistent sourcing stands behind this profile? Computed from source count, source quality, independent confirmation, recency, and consistency across sources. Active critical risks deduct from the score. It measures how well-sourced our picture is — not proof that the technology works.

What scores do — and do not — mean

Scores are relative rankings on a defined methodology. They reflect the structured evidence the platform has indexed at a point in time.

Scores are not investment advice. They are not predictions of returns. They are not endorsements. A high score does not mean a company is a good investment; a low score does not mean a company is not. They are inputs for your own analysis.

Source rules

• Every patent record links to a source URL. We do not invent or paraphrase patents.
• Every verification event includes a source URL and a confidence score (0–100).
• Every research source includes a URL, a confidence score, and a verification status.
• Press-release claims about patents are recorded as company_statement sources, not as patents.

Approval workflow

AI agents draft and propose; admins approve and publish. Nothing posted by an agent is public-facing until reviewed. Reports remain in published=false state until a human flips the publish flag.