A worked example of first-principles prompting: the same question asked four ways. Named, it returns the familiar answer; asked for first principles by name, it reasons better and lands in the same place; given a vague outcome, it returns theory; given a checkable outcome, it returns a design in which every choice traces to a clause. … readThe Only Free Variable

Your AI reasons by analogy because it is built from precedent. To get first-principles reasoning, name the outcome and leave the thing unnamed: write a checkable outcome, ask for the parts, sort law from custom, rebuild, then let the analogy back in as the auditor. … readThe Name Is the Analogy