---
id: "202610132129"
title: "Rhizome AI: one founder, a product made from public documents"
type: literature
status: partial
source: "Matheus Hoffmann, messages to friend3, 2026-10-13: his own words, thinking out loud; nothing here is friend3's"
tags: [agents, documents, startups, brazil]
created: 2026-10-13
---
<!-- cite: main="Matheus Hoffmann"; last="Matheus Hoffmann" -->

# Rhizome AI: one founder, a product made from public documents

My example to friend3 of where this goes. Companies have worked for years to read PDF files
well, to build large applications on top of them. Rhizome AI is one of these, and Y Combinator
accepted it with one person.

What the sources say, read on 13 October 2026:

- Rhizome AI's site: the product answers a question about the positions of the FDA and the EMA,
  "citing up to 1,000 docs from health authority databases". It carries the badge "Backed by Y
  Combinator" (<https://rhizomeai.com/>).
- Y Combinator's page: batch Winter 2026, team size 1, one founder listed, and the one-line
  description "Agent Platform for Life Sciences"
  (<https://www.ycombinator.com/companies/rhizome-ai>).

**Status: partial.** The two points above are verified from those pages. Two things are not:

- "They have optimised PDF processing for years" is a general claim. No source was fetched.
- "There is no product like this for Brazil" was not checked. A search for a product that does
  the same over ANVISA's documents would settle it.

**[ANALYSIS]** The product reads public documents of regulators, not private papers. So it shows
that one person with agents can build on a large body of documents. It does not show, alone,
that private science is reverse engineered. That claim is in [[202610132127-private-code-and-private-science-lose-their-value]].

Easy version in Portuguese: [easy/202610132129-rhizome-ai-one-founder-on-documents.easy.pt-BR.md](easy/202610132129-rhizome-ai-one-founder-on-documents.easy.pt-BR.md).