The Standard
Every badge on this site was awarded against a bar that was published first. This page is that bar. Read it before you decide whether to trust anything here.
What we check
Before a profile goes live, four things
- 01The link resolves and the work is reachable.
- 02What's behind the link matches what was claimed.
- 03The description is an account of real work, not a list of tools.
- 04The person is traceable — a real human with a findable presence.
Fail any of these and it goes back with a note saying which one.
Speed
Time is part of the standard
AI's real effect is compression: work that took weeks takes days. So every builder is asked to state what AI actually did to the timeline on each project — what it took over, where the time went, and how long the same work would have taken without it. That answer is the builder's own account, not a verified fact, and it is read against the artifact at review. A claim of four days for something that plainly took months gets the same note as any other claim the evidence doesn't support.
Rubric
What we look for
- 01
Work nobody asked for
Something built without a client, a brief or a paycheck behind it. People who are serious about this build when there is no reason to.
- 02
Something recent
At least one project from the last ninety days. This field moves monthly; a portfolio from two years ago describes who someone used to be.
- 03
Finished, not demonstrated
A live thing a stranger can open and use — not a prototype, a screen recording or a walkthrough. The last stretch of a project is unrewarding, and it is where most AI-assisted work quietly stops.
- 04
An honest account of what the tools got wrong
Anyone can say they used a model. Only someone who did the work can say where it failed them, what it produced badly, and what they rebuilt by hand. This is the question we weigh most heavily.
These are judgments made by a human reading the evidence, not a score computed from a form. Tiers reflect what the work shows. Anyone who disagrees with a review can say so, and it gets looked at again.
Tiers
The four tiers
Tier 1
Listed
Approved and real, but the proof is a prototype or a demo. Nobody outside the builder has used it.
Tier 2
Shipped
Something live that people other than the builder actually use.
Tier 3
Proven
Shipped, plus the hard part: the eval suite, the production failure and its fix, the measured outcome, the discard rate. Each role page defines exactly what this means for that role.
Tier 4
Vouched
Proven, plus corroboration from someone who isn't the candidate — a named client, a public reference, an employer for that specific work. Rare by design.
The line in the sand
What the badge is not
It is not a rating of how good someone is. It is a statement about how much evidence exists. A brilliant engineer whose best work is private will rank below a mediocre one who shipped in public. That's a limit of evidence, not a judgment of skill, and we'd rather say it than hide it.
It also cannot be bought. No paid tier, no promoted placement, no upgrade path. Evidence is the only route up.
Two different things
Claim and badge are separate
Builders pick their own depth — Core, Working or Light — against their role's standard. Nobody verifies it, so it is always shown as "Self-claimed" and you cannot filter the directory by it. The badge is awarded independently, from the evidence. When the two disagree, that gap is information, and we show both.
Review
Who reviews this
Right now, one person. Every profile, by hand, against the bar on this page. That doesn't scale and isn't meant to yet — early reviews are how a standard gets tested against real people instead of against a plan. As volume grows, mechanical checks and AI will do the reading. A human will keep making the decision.
Challenge
If you disagree
Reply to your review note, or email support@findaihuman.com. Decisions get revisited. If a badge is wrong, say so — a standard nobody can challenge is just an opinion.
Coverage
How roles get added
A role gets defined when enough people are doing work that existing titles describe badly. If you do something these twelve don't cover, tell us what it is at support@findaihuman.com.