Yes — and it lands closer to iterFact than to almost any other product, because the failure in that Fast Company piece (Harry McCracken, "AI-assisted journalism needs disclosure") is basically iterFact's job description. The Rosenbaum disaster isn't "AI is bad." It's that AI-generated assertions got published with no provenance — a quote attributed to Kara Swisher that traced back to nothing. McCracken's defense is structural, not moral: he never pastes chatbot output into a draft, and every claim has to trace to a source before it can appear. That's not a writing habit, it's an architecture rule. And it's the exact line iterFact already draws between the deterministic assembler (real numbers from FL DOR, COT data, whatever the engine actually queried) and the AI planner (structure, narrative, framing). So the lesson is sharp: the deterministic split only protects you if it's enforced at the prose layer. Engines pulling real parcel data are safe — that's ground truth. The risk lives anywhere the planner writes a sentence asserting a fact the engine didn't pull. A submarket brief where the prose says "rents rose 12% YoY" but no engine computed that number is a Rosenbaum quote in a different costume — except automated, at scale, with the iterFact name on it. Which is the part that should actually get your attention: the product is named Fact. Rosenbaum's book was about the future of truth and it shipped five fabrications. iterFact eats that irony at 10x. One high-profile artifact with an unsourced number, sent by one of Justin's agents to a client, is exactly the incident the name invites people to go looking for. That's an asymmetric downside — cheap to defend against now, expensive to survive once. Two moves fall out of it, both additive to what's already being built: 1. Provenance as a visible feature. Every artifact carries an auto-generated methodology/sources footer — what data fed it, which county, pulled when, computed vs. narrated. For the real estate vertical that's a trust signal and a liability shield. It's also the anti-generic-LLM wedge: anyone can ask Claude for a report; iterFact's is the one that shows its work. 2. A fact-critic pass. The critic already catches design violations — shadows, font weights, uppercase. The natural extension is a stage that flags any numeric or factual claim in the prose that doesn't trace to a field the engine actually returned. That's the pass that catches the Swisher quote before it ships. The positioning line to take from it: iterFact is the AI document tool that can't lie about your data, because it isn't allowed to make any up.