Replacing LLMs with Jev for decisions: pick an answer, skip the essay
The author argues that writing and reasoning should stay with LLMs, while judgments like whether an email is urgent should use Jev, which picks one answer from a given list with a confidence score.
You've been using the wrong tool for half your work. LLMs are extraordinary at what they were built for. Writing, reasoning, working through a problem with you. That's real and it isn't going anywhere. The mistake was using them for everything else. An LLM answers by writing an essay one word at a time. When the question is "is this email urgent," you don't get the right answer. Jev skips the essay. You give it a situation and the possible answers, it picks one with a confidence score. It can only choose from your list. Nobody sat down and decided every AI task should run through a chat window. It happened because chat was the only interface users had. So we built systems where the model writes a paragraph to answer a yes or no, and we put important decisions in the hands of the wrong tool. In the new world we'll use LLMs for writing and Jev for the decisions. LLMs don't think, you do.