Models finally read whole papers reliably. Extraction stopped being the hard part. What’s left (synthesis depth and trust) is a corpus-and-method problem. That’s ours.
04 · The Road
v1 earns the trust. v2 compounds it.
We ship the evidence base first (search, screening, the table) and the review writer on top of it, drafting only from evidence the scholar has already approved. v2 turns those reviews into a knowledge engine. The sequence isn’t caution. It’s the strategy.
-
v1 building now
The evidence base + the first draft
Curated search over the top-tier corpus, systematic screening with PRISMA, the quote-grounded comparison table, and a first draft of the review written only from that approved evidence: organized by theory and tension, every claim traced to a DOI, always labeled a draft. The bar is explicit: deeper than Elicit, grounded without exception.
curated search · screening + PRISMA · extraction table · drafted review · reproducible method log
-
v2 next
The knowledge engine
Every review a scholar runs teaches the system: verified paper cards accumulate into a standing knowledge base over the corpus: the field, as structured data, filterable by theory, method, construct, and finding.
accumulating paper cards · theory × method × finding filters · corpus-wide questions
-
v2+ then
The gap engine
Read the field’s own future-work statements, absence patterns, and unexplained contradictions across that knowledge base to surface the questions nobody has answered yet. Research ideas: grounded, not hallucinated.
clustered future-work statements · absence maps · contradiction mining · grounded gaps
Why now
Reviews are due for infrastructure. Every business school runs on them, from PhD seminars to tenure files, and the tooling is a reference manager and a spreadsheet.
The method generalizes. Business first: 148 journals across the lists the field trusts, a clean, finite, high-status boundary. Then the same instrument, corpus by corpus, field by field.
The ask
Let’s talk.
We’re assembling early partners for v1. If you back research infrastructure (or know the scholars who would stress-test this), write to us.