A first-of-its-kind format in which the panelists are the AI tools themselves, and the only human in the exchange is the one asking the questions. A moderator poses the same tax fact patterns to a lineup of general-purpose models and legal- and tax-specific platforms, each answering in its own synthesized voice with its response on screen — and then puts each model's answer to the others for response. The interest is less in whether any model is right than in what happens when they disagree: whether a model defends its reading under challenge, revises it, or abandons a defensible position simply because another model pushed. Where they converge may mark settled law, or may mark nothing more than shared training data; where they split maps the questions practitioners still contest. Proof, for now, that the singularity can wait — and a closer look at the privilege, confidentiality, and unauthorized-practice questions every firm now faces.