Human evaluation for the machine age

AI writes. You judge.

Verdict is where verified humans review AI-generated work. We turn identity-backed expert reviews into trusted evaluation data for AI teams, and a portable reputation for the humans who produce it.

Developers

Open

Review AI-generated code against a structured rubric. GitHub provides professional context and performance on calibrated cases builds your Verdict record.

From the docket

Configuration loader

javascriptGemini 2.5 Pro

Write a Node.js module that loads JSON config from a file path given in an environment variable, with sensible defaults.

17 open cases await judgment — gold standards hidden among them.

Judge this case

const fs = require("fs");

const DEFAULTS = {
  port: 3000,
  logLevel: "info",
  cacheTtl: 300,
};

let config = null;

function loadConfig() {
  if (config) return config;

How it works

Review

Real AI-generated work with real flaws. Score it on a structured rubric, then write your critique.

Accuracy

Hidden gold-standard items measure your judgment, not your output. Accuracy is the only currency.

Leagues

Climb from Clerk to Chief Justice. Promotion demands proven accuracy — volume gets you nowhere.

Credential

A public, portable record of your judgment — top-percentile proof for your README or LinkedIn.

The credential

Judgment you can carry.

Every reviewer earns a public profile and a live, embeddable badge — league, accuracy, percentile — portable proof of judgment for a README, a portfolio, or LinkedIn. Hard-earned, verifiable, yours.

Illustrative — the real one is earned.

“The web fills with machine output while original human signal declines. Verdict manufactures the scarcest resource in AI — structured, credentialed human judgment.”
The thesis