A post speculates that decision models could become a category for new AI labs or branded offerings from other AI companies. The author argues that their different economics may give frontier labs less incentive to invest compute—especially if the data differs substantially, which the author says they are unsure about. The post also predicts that software companies could easily fine-tune or customize these models. But it expects frontier labs to return to this area for small models, making newer labs’ prospects depend on how quickly they hire customer-facing staff to gather data and launch customizations. The author still anticipates demand for decision models from a different family to check or steer an LLM’s work.