The artifact is no longer the proof of capability. The judgment behind it is.
When any candidate can prompt their way to something that looks like the work of someone with a decade of experience, the finished artifact stops being evidence of anything. A polished deck, a working app, a sharp memo — all of it is now cheap to produce and hard to attribute. The output used to be the proof. AI broke that logic.
Proof of Judgment moves the unit of evaluation from the output to the process behind it — not what you made, but the sequence of decisions that produced it. It reads four traces that a finished artifact hides: how the problem was framed, how the work was steered, what was rejected along the way, and what the person supplied that no model could. Judgment leaves fingerprints. This is a way to read them.
How the problem was framed before a tool was ever reached for. A vague ask gets a generic answer; a precise one shows the thinker already knew what mattered. The opening move reveals how someone defines a problem — the part the model cannot do for you.
The path from first draft to final. Nobody lands it on the first pass. The sequence of refinements shows how a person reads a result, names what's wrong with it, and steers toward better. This is where taste becomes visible and legible.
What was thrown away, and why. The directions not taken are as telling as the one that shipped. Knowing why an option was wrong is a stronger signal than landing on the right one by luck — it's the difference between judgment and a good roll.
What the person supplied that no model could: context, domain knowledge, the call on what “good” means for this specific situation. The part of the work that is structurally the human's — the layer that decides where the capability gets pointed.
The number isn't the point — it forces the evaluator to look at the process instead of reacting to the polish of the output. The scale runs from absent to exceptional; the total sorts the signal.
Proof of Judgment was first put to work as a hiring instrument at Assembled Intelligence — a toolkit built to evaluate how candidates think with AI, not what they handed in.
Enter the Brief →