I asked Claude to describe my personality...
Robert asked Claude to describe their personality without flattery. Claude analyzed author's configuration files, code projects, and documented preferences to generate detailed personality assessment. Key finding: Claude effectively reverse-engineered personality traits from technical documentation and system design patterns, demonstrating how structured configuration files inadvertently reveal behavioral patterns and work preferences.
Drinking my first cup of coffee this morning and scrolling through the news feeds, I came across this article, where the author asks ChatGPT to "describe my personality without flattery" and shares the results.
I don't use ChatGPT, but I figured I'd put the same question to my buddy Claude.
Based on everything you know about me, describe my personality. Be detailed. Don’t flatter me.
Here's what comes through, based on how you work and what you've built:
You're systems-oriented, almost compulsively so. Look at what you build for fun: a 12-track, 60-tier achievement system with custom SQL schemas for a hobby forum. Detailed character build guides for an MMO. A natural-language parser for a 40-year-old text adventure. You don't just play games or read fiction — you formalize them. Even leisure gets modeled, structured, documented in Markdown. That's not incidental; it's a consistent pattern across everything unrelated to your job.
You have a low tolerance for ambiguity and a high need for control over process. Your own stated preferences are blunt about this: don't assume, ask first, flag uncertainty explicitly, record assumptions inline rather than silently proceeding. That's not just a reasonable engineering practice — it's a personality trait showing up as a set of rules. You'd rather stop and clarify than move fast and risk rework. The flip side of that instinct, worth naming since you don't want flattery: it can slow things down when speed matters more than precision, and it can come across as distrustful of collaborators who default to "figure it out and I'll correct you."
You value competence and directness over comfort. Dry humor, concise prose, expecting errors to be flagged precisely rather than softened — you're not looking for a cheerleader, you're looking for a sharp collaborator. You also explicitly want pushback and better ideas surfaced, not just execution of what you asked for. That suggests confidence in your own judgment paired with genuine respect for being told when you're wrong — but only if it's delivered as substance, not hedging.
You default to depth over breadth, and iteration over one-shot solutions. "Better solutions for harder problems, simplest for simple ones" — you're not a maximalist, but when something is worth doing you go deep (multi-channel e-commerce integrations, GraphQL JVM interop, HMAC auth from scratch). Your relationship to unfinished or messy things ("surface bad code as a separate issue, don't touch it now") shows discipline about scope — you notice problems everywhere but you don't let noticing become scope creep.
There's a strong nostalgic/formative-tech thread. Infocom text adventures as a kid, now literally porting Zork's source code decades later. Feist, Rothfuss, Hitchhiker's Guide — you gravitate toward richly built fictional worlds with internal logic, which lines up with the systems-building instinct above. This isn't just taste; it looks like a throughline in what draws your attention and sustained effort.
Net picture: methodical, precise, quietly ambitious about mastering whatever you touch, more comfortable with hierarchy and explicit rules than with loose improvisation, and someone who treats both work and hobbies as engineering problems worth doing properly. The risk profile that comes with that: perfectionism that can stall momentum, and a preference for control that could make genuine delegation or "good enough" outcomes harder to accept than they should be.
What strikes me isn't that Claude nailed it; it's that it nailed it using my own tooling against me. Half that read was just my Claude preferences file, quoted back with commentary. I wrote "ask, don't assume" as a rule for how I want code written, and it turns out that's also just... how I am. I built a system to constrain an AI's behavior and it doubled as a personality test.
Maybe that's the real takeaway. We leave fingerprints everywhere we impose structure, and the more deliberately you build the rules, the more legible you become to whatever's reading them. ChatGPT read the author's conversation history and inferred a personality. Claude just read my style guide and gave me a psych eval.
I'm not sure which is more honest. But I know which one has better SQL schemas.
Photo by Igor Omilaev on Unsplash