Your job description, checked against what actually happened.
Paste the posting. We break it into claims and, for each one, show you what your team’s own Slack, docs, tickets and meeting notes say — including where they say something else, and including, out loud, where they say nothing at all.
We do not tell you what your role is. Every verdict that is not “silent” carries a sentence a real person really wrote, verified character by character against the artifact it came from — if the quote is not there, the verdict is deleted rather than softened. You can check any line on this page by opening Slack.
Two model calls over the whole corpus, about a minute. Nothing is compared until you press this.
- Contradicted
- Your own material shows something incompatible with the line as written — someone with standing saying the opposite, a thing the line assumes exists that visibly does not, or a question the posting settles that your team is still openly arguing about. Always quoted, always attributed.
- Silent
- Nothing in the corpus touches it. This is a result, not a miss: either the work is not happening, or it happens somewhere we cannot see. We say which rooms we were looking in so you can tell the difference yourself.
- Supported
- The corpus shows it happening, specifically, in a passage we can point at. The least interesting third of the page, which is why it is at the bottom.
The claims are extracted by a call that has never seen your corpus, so the denominator is not picked to flatter the numerator; each quoted line is then verified as really being in the document you pasted. A verdict biased toward silent is the point — a false contradiction tells a hiring manager their own posting is at odds with their own team on evidence that will not survive being looked at, and the first one of those takes every true verdict next to it down with it. The corpus behind the demo is realistic and synthetic; it is not Lexhav’s real Slack, and we have never had access to it.