Why we don't just hand the specification to ChatGPT
How Offra turns hundreds of pages of tender documents into a list of requirements, each tied to its sentence, and why one request to an AI can't.
English
Method
How the numbers are produced, and how we check that they hold up.
The articles
Which Canadian tender sources we index
Ten public sources, what each one lets us say, what is missing and why — and what still works on any tender at all, indexed or not.
English
How we predict what a public tender will cost, and who will bid
Four models, one rubric, and the evaluation discipline that makes the numbers worth quoting — including the results we threw away.
English
Our numbers don't come from an AI
The price range and the likely-bidder list are produced by ordinary statistical models, and measured against the record. Where AI actually works here.
English