Wake

Wake is a world model for organizational action. Drop one action a company could take — an acquisition, a layoff, a product change, an internal proposal — into a graph of ~200 real entities (executives, competitors, journalists, customer cohorts, platforms, regulators), where every node is its own Gemini 3.5 Flash reasoner. Then watch the consequences cascade in real time: you can read who is reasoning and why, one rationale at a time, as the press turns, communities go hostile, and a private grievance leaks when a node's public face drifts too far from what it privately thinks. It doesn't stop at one run. Wake runs the same action many times with the world perturbed, collapses the futures into outcome regimes, and surfaces the single pivotal variable that most decides which one you land in. You can take the same action and see how it plays out under different framings, and click any outcome to trace the exact causal chain back to your decision, cited event by event. The world isn't fixed either — describe a scenario in plain language and Wake assembles the real cast and builds a runnable world around it. The thesis: as AI takes over execution and humans become the judgment layer, every decision-maker needs to ask "what happens if I do X?" before doing it. Robotics is building world models for physical action; Wake builds one for organizational, textual action. It is not an oracle. It makes the consequence space legible, so you can watch your decision before you make it. Gemini 3.5 Flash is what makes hundreds of parallel reasoners cheap enough to actually run.