Most prompt engineering advice pushes in one direction: add more. More context, more constraints, more examples, more edge case coverage. Prompts get longer, outputs improve slightly, hallucinations persist anyway.
A technique gaining traction among practitioners flips this: instead of defining everything yourself, let the AI define its own working vocabulary before it answers. The method involves adding something like "use Aristotelian first principles reasoning - before you proceed, break every undefined term down to its atomic meaning" to the front of your prompt.
The idea draws on a real philosophical method. Aristotle's approach to any problem started with defining terms precisely before reasoning from them. When applied to prompting, this forces the model to make its assumptions explicit - what it thinks "tone" means in your context, what "concise" means, what "professional" implies - rather than silently defaulting to whatever interpretation its training data happens to favor.
What Actually Changes
In practice, the model produces a working definitions block before generating the actual output. That block often reveals where your prompt was ambiguous. If you ask for a "professional summary" and the model's first-principles breakdown treats "professional" as meaning stiff corporate-formal, you can correct that assumption before it writes 400 words in the wrong direction.
This is different from exhaustively padding your prompt with explicit definitions upfront. That approach works, but it requires you to anticipate every ambiguity in advance. Letting the model surface its own assumptions catches the ones you didn't know to specify.
Results aren't universally better. Claude and ChatGPT both vary in how they execute this instruction, and for simple, clear-cut tasks it adds unnecessary overhead. For complex writing tasks, structured analysis, or anything where tone and framing matter significantly, forcing the model to commit to its interpretations before generating tends to reduce the "technically correct but wrong" outputs that longer prompts alone don't fix.
The addition takes about five seconds to type. For complex or high-stakes briefs, it's worth one test before dismissing it.