Y Combinator put "company brain" on its request-for-startups list.
Within a month, every RAG wrapper on the market had renamed itself one.
The RFS is right about the problem
Institutional knowledge is scattered, undocumented, and walks out the door with every senior hire who leaves. Agents are operationally blind: they can query a warehouse and generate SQL, but they have no mental model of how the company works.
That's a real, large, unsolved problem. The disagreement is about what solves it.
What the current field gets right and wrong
| Where they're strong | Where they stop | |
|---|---|---|
| Context-retrieval players | Fast ingestion, broad source coverage, good UX | Retrieval can't aggregate, join, or enforce entitlement |
| Warehouse-native tools | Deep integration, existing governance | Scope ends at one platform |
| Agent frameworks | Orchestration and tool use | No opinion on what data means |
The pattern across all three: they solve roughly 40% of the problem — finding relevant material — and treat the rest as someone else's layer.
The missing 60%
Retrieval tells you what a document said about churn. Production decisions need what churn is, at what grain, for the person asking, computed over current data, with proof they were entitled to see it.
That requires:
- A typed, versioned semantic graph rather than an embedding index
- Join path proof rather than similarity ranking
- Compile-time policy rather than post-hoc filtering
- Point-in-time reproducibility rather than a citation
The honest competitive read
Context retrieval is necessary and insufficient. A company brain that only reads documents is a very good search box. One wired to governed execution is infrastructure.
By 2027 the category will split along exactly that line, and the retrieval-only products will be features inside something larger.
The full breakdown — the competitive teardown of Hyper, GBrain and Savant, the strategic separation, and the winners-and-losers call — is here:
👉 YC's Company Brain RFS: What Hyper, GBrain, and the Competition Got Right (and Wrong)
Originally published at colrows.com/blogs/yc-company-brain-rfs
Top comments (0)