A due-diligence knowledge base where every memo compounds.

We turned S-1 prospectuses, 10-Ks, merger proxies, and deal press releases into a living, interlinked due-diligence base of 139 cross-linked pages using a four-stage process, then hardened it with an integrity layer that fact-checks itself. Here's how it works at a high level.

01 - Extract

We read the filings and identified the knowledge structures inside them: companies (Toast, Procore, Veeva, ServiceTitan, Samsara), deals (nine IPOs from 2013 to 2024, the 2024-25 take-private wave), vertical markets (restaurant tech, construction tech, edtech), and the people behind them. Each source was read once, deeply, and decomposed into atomic, structured fragments - exact figures with their fiscal periods, never rounded away.

02 - Structure

We designed a template for each page type - company, deal, market, person, thesis, source, synthesis - so every fragment follows a consistent schema. A company page always carries its metrics table with fiscal periods and sources; a deal page always reads like a memo: overview, rationale, multiples, diligence notes, aftermath. That consistency is what makes the system compoundable - and it caught real traps, like fiscal years ending January 31 and retention metrics whose definitions differ by issuer and are not comparable without a caveat.

03 - Connect

Every fragment was cross-linked into a knowledge graph - 1,903 internal links, all resolving. The PowerSchool take-private links to the Bain Capital page, the S-1 it IPO'd on, the education-technology market it left a public-comp gap in, and the take-private-wave synthesis that compares its 37% premium to Instructure's 16% and Olo's 65%. Where sources disagree - an announced deal price versus the closing consideration in the acquirer's 10-K - both are kept and the conflict is flagged, not silently resolved.

04 - Compound

This is the pitch made literal: every memo feeds the living knowledge base. A thesis page like "Retention Quality" cites down into five company pages; a question like "why did PE firms take Olo, Instructure, and PowerSchool private" draws across three deal memos and files its answer back into the base, making the next query richer. Knowledge from past deals compounds into the next one instead of dying in a folder.

The integrity layer

A diligence base is only useful if an investor can trust every number. Every load-bearing figure - revenue, retention, deal value, premium - carries its fiscal period and traces to a source page naming the exact SEC accession or document. Where a figure only exists in press coverage, it is flagged, not asserted; 53 pages carry explicit uncertainty flags rather than confident guesses.

Before sign-off, an adversarial fact-checking pass attacked the base's load-bearing claims against primary sources. All fourteen traced claims matched their filings exactly - and the pass caught a boutique-bank valuation multiple that did not exist in the report it was attributed to, corrected it from the actual document, and fixed a quarterly take-rate being framed as a full-year figure. A regression suite of twenty golden queries runs on every update, so if a load-bearing fact silently changes, a red check catches it.