Research
Verification and provenance in financial data
My current question is narrow: what evidence should sit behind a number before a system is allowed to call it verified?
The work treats provenance, field-level status, confidence, parser version, and observed error rates as product requirements. It separates what a source states from what an analyst infers, and it leaves unknown values unknown.
Capital Fields
Capital Fields
Active, 2026
Governance as code for private markets: independent research and engineering for records whose definitions, provenance, and point-in-time state must remain inspectable.
A Verification Standard for Filing Data
In progress
A working specification for source receipts, per-field status, confidence labels, versioned processing, and published error rates.
Study on public regulatory filings
Registered 2026
The design and analysis plan were timestamped before study-window analysis. The registration is under embargo; no study result is claimed here.