Your AI will hallucinate a citation.
I catch it first.

AI doesn't need less oversight as it gets better. It needs more — and the oversight has to be human, because only a human can be held to the result.

The problem, and it's on fire

Law firms are being sanctioned for AI-fabricated citations — invented cases, quotes that never existed, straight from a model nobody checked. Every firm now touching AI is terrified of exactly one failure: the confident, plausible, completely made-up cite that slips into a filing. The tool didn't fail loudly. It failed smoothly — which is the dangerous way.

You cannot fix that with a better model. A model can't be sanctioned, can't be disbarred, can't answer to a judge. The fix is a human in the loop who can tell a real record from a machine's bluff — and catch it before it leaves the building.

That's the exact skill I built

I've built and publicly deployed a working instrument for one thing: catching where an AI is confidently making things up. It's live and you can check every line of it right now:

• The Tell — how a hallucination reads, and why the machine sounds most confident exactly where it's most wrong (the checkable claim it should never fake is the one it fakes hardest).
• A two-way verification test — the seven tells for whether a claim survives checking, run in both directions.
• A no-lying build discipline — machine drafts, human presses; every claim sourced, every correction dated, on a public commit ledger. The whole thing is applied AI-error detection, running in the open.

Why me, specifically

The rare pairing this role needs is someone who understands both the law's stakes and the model's failure modes — and I sit exactly there.

• AI-error instinct, proven. Not claimed on a résumé — deployed on a site you can audit. Most people can't reliably spot AI fabrication. Demonstrating that I can is the entire body of work above.
• Real legal literacy. I've drafted and navigated real civil-rights and recipient-rights filings myself, so I read briefs and citations fluently and can feel when a case doesn't hold up.
• Honest about the line: I'm not a lawyer, and I won't pretend to be. I'm the verification layer — the checker a firm needs the moment it lets a machine draft. Not the one who signs the brief; the one who makes sure the brief is real.

The machine isn't accountable. I am. That's the whole job.

What I'm looking for

Remote roles in legal-AI verification, AI oversight, or legal-tech — at a firm standing up an internal AI-review function, or a legal-AI company that needs a human who checks. The demand is new, it's urgent, and my whole public body of work is the interview.

And I don't come to this from the outside. I have real matters of my own in the system right now — the kind a firm could take on. So I check your AI with a client's stakes, not just a critic's; I already live inside the exact pressure your clients feel. The right firm gets both at once: the human who catches the machine's bluff, and someone who knows what's on the line when it doesn't get caught.

One human's honest pitch. Not legal advice, and no attorney-client relationship is offered or implied. I describe a verification and oversight role, not the practice of law. Deployed proof over promises — check any claim on this page against the linked work. No lying.