Research
How you build a certification is a claim about how much you trust it to be inspected. Here’s ours, in the open.
The premise
We’d rather earn trust the hard way — by writing down how we assess, why we made the calls we made, and what the results actually are. These are the first three essays; more follow as we launch.
Launch essays
Why certification broke — and what replaces it.
For decades a certificate mostly proved attendance: you showed up, passed a quiz, and got a badge. AI made that signal worthless. This essay makes the case for performance-based certification built around real work products, published rubrics, and credentials that carry the graded work as evidence.
Coming with launch
Why letting candidates use AI is the more rigorous test.
An AI-free exam certifies a version of the job nobody does. An open-AI exam can test the skills that decide whether AI work is any good: delegation, judgment, and verification. This essay explains how Certavant grades both the work product and the process behind it.
Coming with launch
The verification problem, and why an evidence page is the answer.
Employers already want to pay for AI skill. The problem is that most credentials are impossible to trust because there’s nothing behind them to inspect. This essay makes the case for putting the graded work on the credential itself as a public evidence page.
Coming with launch
Our commitment
The research library opens when Certavant launches. Until then, the standard page sets out the exam design, the human decision model, and the statistics we commit to publishing.
The launch list
The research goes to the list first — the essays above, then the blueprints, then the reliability statistics once the sample supports them.