I got tired of AI tools that make up numbers, so I built one that refuses to
A few posts on here lately have stuck with me. A searcher who found a six-figure obligation only after closing, because the seller and the buyer had read the same clause two completely different ways. Someone asking how to write a seller-note metric an owner can't quietly game by piling on expenses. Someone else asking the question that should scare all of us: "what did you discover post-close that was a surprise?" Different deals, same fear — the thing the documents didn't tell you until it was too late. That's the exact problem I've been building Verity for. One rule underneath it: it never invents a number. Every figure links back to the page it came from. If the documents don't support a claim, it flags the gap and tells you what to ask for — it doesn't fill it. And where the seller's or broker's framing doesn't match what the numbers actually say, it puts the contradiction in front of you instead of smoothing it over. You can see how it works here — every figure sourced, gaps flagged, nothing invented: redacted I'm looking for a few searchers to kick the tires and give me feedback — free Operator-level access, no trial clock. The only thing I want back is honest feedback, including where it falls short. If you've got a live CIM you're working through, I'd like you to run it and tell me where it's wrong. Reply or DM.