Notes on finance as code
How we build structured, traceable, AI-native financial models — from the team at Convexent.
Validation
A Financial Model You Can Test
Trust in a model should come from more than a polished workbook. Here is how Convexent uses deterministic validation to catch broken assumptions, inconsistent logic, and models that do not add up.
Engineering
Which Model Builds the Best Financial Model?
We held our whole finance-as-code system fixed and swapped only the generation model — Gemini, GPT, and Claude, judged the same way as the first post. The model matters, but the system matters more: the harness lifts every model, and which one you pick moves the needle far less than whether you run the system at all.
Engineering
Finance as Code
AI became exceptional at software engineering because code is structured, testable, and verifiable. We represent financial models the same way — and everything that makes AI good at coding carries over to finance.
More posts coming soon. For product updates, get in touch.