Part IV: of the Pre-Institutional Intelligence Research Program | Simulation, Deep Learning, and the Separation of Training from Execution

Parts I through III of this research program traced a single cognitive sequence within the individual: how intelligence begins before institutional structure exists (Part I), how a Training State builds a repertoire of prepared possibilities before any decision is made (Part II), and how a Decision Selection Layer converts that repertoire into one committed action under uncertainty (Part III). This fourth paper asks what happens when the same sequence is stretched beyond a single mind. Deep learning offers a precise engineering vocabulary for the individual architecture already proposed in Part II — a training phase, conducted offline and at low cost, that is structurally separated from an inference phase in which the trained system acts on live data. This paper argues that Part II's Training State → Decision State architecture is the direct structural bridge to that vocabulary, and proposes its extension to the institutional level: a Simulation Infrastructure that hosts a Collective Training State, producing an Institutional Prepared Space from which Real Execution draws. Rather than displacing the earlier finding that simulated experience cannot substitute for reality — established on separate philosophical grounds in the companion research program on communicative compensation — this paper reassigns simulation a different function: not the compensation of missing experience, but the deliberate construction of institutional preparedness prior to real-world engagement.