Methodology

This terminal exists to be reliable before it is anything else. Every record you see follows the same pipeline, and every rule below is enforced in the software, not just promised in prose.

What we track

We do not track venture funding in general. We track capital flowing into what is rebuilding the internet, tagged to four lanes: AGENTS (AI agents — the new primary users of the web), ROBOTS (robotics — new physical nodes on the network), ORBIT (space and satellite — the internet’s new physical layer), and RAILS (protocols, agent standards, identity — the new rails). A generic SaaS round, however large, does not appear here. That taxonomy is the product.

Where records come from

• SEC EDGAR filings (Form D and related disclosures), via the official EDGAR APIs.
• Company press and newsroom announcements.
• Public RSS feeds from established trade press (TechCrunch, SpaceNews, The Robot Report, Crunchbase News, and others) for corroboration and rounds filings won’t catch.

We use official APIs and public feeds only. We never scrape sites whose terms forbid it, and we never republish source content — summaries are written fresh, and every record links out to its sources.

How records are verified

Automated extraction parses each source into structured fields: company, amount, round type, investors (lead and participating), date, and lane tags. The same raise reported in multiple places is merged into one record with all sources attached. If a field can’t be supported by a source, it ships as “undisclosed” — it is never guessed. Each record carries an extraction-confidence score, and low-confidence records get extra scrutiny at review.

The human gate

Nothing publishes automatically. Every day, a human reviews the candidate set: approving records, correcting summaries, and choosing the day’s hero signal. Hero selection is significance-weighted, not size-weighted — a $20M raise in agent protocols can outrank a $200M generic AI round — and each hero pick states its reasoning and its tradeoff on the record itself, so you can audit the judgment, not just the data.

Corrections

If a record is wrong, it gets corrected in the database, not just in a follow-up post. The archive is the canonical version at all times.