Nate B Jones describes the gap between deterministic rules and generative models: many tasks require interpreting messy language but produce only a category, score or decisionA decision model selects or scores possible answers for a defined decision.. A general-purpose classifierA classifier assigns an input to one or more defined categories based on rules or learned patterns. can address that pattern without training a separate classifier for every task.
Nate B Jones outlines architectures for support routing, document screening, agent-action checks and workflow orchestrationAgent orchestration coordinates AI agents, tools, people, tasks, state, and control flow so a larger workflow reaches a verified outcome.. In one pattern, a classifier selects whether the next step should use a tool, a generative model or a humanAI model routing sends each request to a model chosen for that request's complexity, cost, speed, privacy, or other requirements., while ordinary code defines what each permitted choice does.
Nate B Jones discusses browser navigation, scientific-question prioritization and spreadsheet judgments as possible applications. He presents reported speed and cost comparisons as examples, not independently verified guarantees, and emphasizes testing each classification problemEvaluation is the systematic assessment of a system's behaviour or results against explicit criteria. because outputs remain probabilistic.
Nate B Jones argues that cheaper judgments may expand the number of useful AI tasks rather than eliminate demand for generative models. The important distinction is between selecting a constrained outcome and the open-ended work of reasoning, writing or explaining.
Watch on YouTube
