Company
A mission, and a first proof.
Fairgrade exists to make collective intelligence trustworthy. The internet made it possible to gather the opinions of billions of people on any question; it did not make it possible to know which of those opinions to believe. Simple aggregation treats a careful judgment and a careless one as the same evidence, and so it produces bubbles, manias and mob rule as readily as it produces wisdom. The alternative most often proposed, a small group deciding what is true on everyone’s behalf, is worse.
There is a third way, and it is the one a courtroom already uses. Expert testimony carries weight in proportion to the witness’s credentials, record and freedom from interest, and the receiver decides. Fairgrade does this at the scale of a crowd, mathematically: every judgment is weighted by the demonstrated credibility of the person making it, the distribution of opinion is read as evidence in itself, and the two estimates, of the answer and of the judges, converge together. Where a single expert or a plain average would be wrong, a credibility-weighted crowd is usually right.
In September 2026 the leaders of the frontier AI laboratories publicly called for slowing the pace of their own models and for independent human evaluators to verify them. The people building the machines conceded that generation has outrun verification and that human judgment is the remedy. What they did not say is how to know which humans to trust. That is the question this company exists to answer.
When any answer can be produced for nothing, the scarce resource is trust, and trust is a human product. Our work is to measure it, so it can be relied upon.
Why education first
A corner of academia broken by inequity and inefficiency.
Grading is the first deployment for three reasons. It is a problem the founders have lived with every semester since 1996. It has a ground truth: an instructor’s own grade, against which the method can be checked, course after course. And it is a place where the failure of unweighted judgment is felt personally, by every student who has ever received a grade and wondered what it was for.
Second to the cost of tuition, the adversarial, time-intensive and tedious work of assigning grades is likely the largest source of frustration in a university, for faculty and students alike. Feedback that could be transformative is rarely given, because giving it well does not scale. Faculty rarely learn what their students actually understood. And now that generative AI can produce most gradable work, the one skill a course can still teach only by practice is the ability to evaluate.
“My husband and I have been full-time professors in top-rated business schools since 1996. We understand and love academia and affirm its central role in society. We also experience its deficiencies. The platform we have built draws on our joint forty years of teaching an enormous number of courses in many universities, and on an unshakable belief that it will significantly improve outcomes for universities, faculty and students. The problem we solve goes to the heart of the mission of education: how the learning of students is assessed.”
The method was tested where it was invented, across 60+ courses and 5,000+ students at NYU Stern. In 2020, Fairgrade won the $100,000 Rennert Prize at NYU’s $300K Entrepreneurs Challenge.
What we believe
Four commitments.
Human judgment is the centre.
Machines organise information; people decide what it means. A system that cannot say which humans to believe, and by how much, has not solved verification. It has moved it.
Theory before data.
The method is grounded in probability theory applied to collective intelligence. It works from a single class of reviews, and it can explain every weight it assigns. A model that needs a million examples and cannot explain itself is the wrong tool for a decision that matters.
AI is a second opinion.
Generative models are capable readers and useful ones. In Fairgrade they sit beside the human verdict as an optional check, labelled as such, for a person to weigh. They are never the source of trust.
Credibility is earned, in public.
The method's own credibility is being built the same way it builds a reviewer's: by being right, repeatedly, where the answer can be checked. Grading is that place. What follows will be earned the same way.
Who is behind this
A method, a platform, and the people who built both.
Professor of Management and Organizations at NYU Stern. Her courses were the proving ground for the method, over a decade and sixty-plus classes.
Professor of Finance at NYU Stern, formerly Chicago Booth and MIT Sloan. Principal architect of the method and the mathematics behind it.
Yash Goel
Met the Marcianos as their student. Built the Fairgrade product and the platform that carries the method into institutions.
The Marcianos started it. Many others joined along the way, and many more built it: the engineers, designers and researchers who have worked on the platform over the years, the faculty who ran it in their courses before it had a name, and the five thousand students whose reviews taught the method what a careful judgment looks like. Fairgrade is the work of all of them.
The Marciano Method is described in a working paper by its inventors. Grading is its first application; the engine generalizes to any question where many observers of uneven reliability meet a hard problem.
Ready to put credibility-weighted judgment to work?
Start for free in your classroom, no credit card required, or talk to us about a pilot in your field.
Are you a student? Join a class →