How BootFile works — and what it isn’t.

BootFile is an eight-archetype framework for how people prefer to reason with an AI. The archetypes are: the Surgeon, the Architect, the Sparring Partner, the Translator, the Co-Pilot, the Librarian, the Closer, and the Maker.

We developed this taxonomy internally. It is not adapted from a published psychometric instrument, and we’re not going to dress it up that way. It is a working model designed to capture differences that matter for AI interaction specifically — not general personality.

The quiz is eight questions. Each question is designed to surface a preference that maps to one or more archetypes. The mapping is deterministic: each answer carries weighted points across the eight archetypes, and your top score is your primary archetype.

When two archetypes are close (within 3 points), the second is shown as a secondary tendency. When three are tightly grouped (within 5 points), the third is surfaced as context, not displayed prominently. The thresholds are ours, set by what we’ve seen produce useful results — not derived from external statistical validation.

Your archetype determines tone, defaults, and the shape of the nine sections. Your quiz answers personalize the content within those sections. Two people with the same primary archetype will get visibly different files if their other answers differ.

The nine sections: First Message, About Me, How I Think, How to Reason With Me, Communication Rules, Format Preferences, Failure Detection, Never Do This, and Quick Commands. You can see a full sample at bootfile.ai/sample.

BootFile is not a clinically validated psychometric instrument. It is not Big Five, not MBTI, not DiSC, and not adapted from peer-reviewed research. It will not tell you anything reliable about your mental health, relationships, or career fit.

It is also self-report. Like any self-report assessment, it’s shaped by how you saw yourself the day you answered the quiz. Retaking it later can yield a different result, especially if you’ve gained clarity about how you actually want an AI to talk to you.

Honest answer: we observe two things. First, the file produces noticeably different AI behavior — that part is mechanical, it’s just custom instructions. Second, people who retake the quiz often re-land on the same archetype, which is the minimum bar for a working classification.

We have not run statistical reliability or validity studies. If we ever do, we’ll publish the methodology and the data. Until then, treat the archetype as a useful frame, not a diagnosis.

• Eight archetypes are a compression. Real cognition doesn’t cleanly bucket. Your secondary archetype matters; so do the ones at the bottom.
• Self-report bias applies. The quiz reflects how you describe yourself, not how you actually behave under pressure.
• The framework is English-language and likely culturally indexed to the kind of work people use AI for today: knowledge work, creative work, technical work. Other domains may need different framing.
• We iterate on quiz questions and content. The archetype names and taxonomy are stable; the surrounding language evolves.

Native AI memory works inside one platform. People use multiple platforms. A BootFile is portable — same cognitive profile, every tool — and the archetype framing was the cleanest abstraction we found for what to put in those custom instructions.

The eight archetypes aren’t arbitrary. They’re the patterns we kept seeing in how people we trust prefer to be reasoned with — the ones who want the answer first, the ones who want the system view, the ones who want to be challenged. We named them and wrote down the rules for each.