> About

We are building the hiring system we wanted to use.

YoHe started from a complaint: every tool in the category produces conclusions about people that nobody can explain.

[n.01/02]> Why this exists

We are building the hiring system we wanted to use.

// 001

The problem we kept running into

Hiring software has spent twenty years getting better at moving candidates between columns and no better at helping anyone read them. The assessment layer, where it exists, is a black box: a percentage, a ranking, a traffic light. When a decision gets questioned — and hiring decisions get questioned — there is nothing underneath to point at.

// 002

What we decided to build instead

A system where the unit of output is a finding, not a score. Each finding names a requirement, states whether the candidate's documents evidence it, and shows the passage. Everything else in the product follows from that: structured jobs, because you cannot assess against prose; banding rather than ranking, because the trade-offs belong to the team; an audit log, because the reasoning is only useful if it survives.

// 003

What we are not building

We are not building automated rejection, candidate scoring from photographs or video, personality inference, or anything that claims to remove bias from hiring. Each of those is technically achievable and we have turned each of them down on purpose. The list is in the docs and it is not a roadmap.

// 004

Where we are

In development, and pre-launch. There are no customers to point at and no results to report, so this page does not have either. Early access is opening to a small number of in-house hiring teams, and the fastest way into that group is to tell us what you are hiring for.

[n.02/02]> What we hold to

The commitments, in one place.

These are the same five that run across the homepage. They are the part of the product that is not up for negotiation.

  1. // 001No candidate is ever rejected by the model
  2. // 002Protected characteristics are excluded from matching
  3. // 003Every AI output is labelled, editable, and reversible
  4. // 004Every conclusion links to its source document
  5. // 005Your candidate data is never used to train models

There are no customers to point at yet.

So this page does not pretend otherwise. If you hire in-house and want to be one of the first, tell us what you are hiring for.