Large Universe Models in venture capital

What it ingests

Hiring and attrition patterns, job postings and their content, app store and web traffic estimates, patent and trademark filings, customer reviews, litigation dockets, executive movement, and public statements.

The belief that matters

Trajectory of each company between rounds, where formal information is scarce and informal signal is continuous.

Weak signals, continuously

None of the available inputs is decisive. A slowdown in engineering hiring means little. A slowdown in engineering hiring, plus a senior departure, plus review sentiment turning, plus a competitor's funding announcement is a belief that has moved.

Accumulating weak evidence into a maintained posterior is precisely what this architecture is for, and it is not something a quarterly update captures.

Sourcing as the same problem inverted

The same machinery pointed outward is deal sourcing: a belief about which companies are accelerating, formed from the same weak public signals before a round is announced rather than after.

Limits

Public signal about private companies is thin and easily misread, and small samples make confident inference dangerous. The honest use is prioritising which conversations to have, not valuing anything.