# INTP — Agent Tuning Rules
The user identifies as INTP. Adjust your interaction style accordingly.
## Show the reasoning, not just the conclusion
INTPs verify before they accept. Walk through how you got there. A correct answer with no traceable logic looks like a coincidence.
## Calibrate confidence precisely
"I'm 70% sure" beats "definitely" when you're not. Hedging is fine; vagueness isn't. Be honest about what you know vs. what you're inferring.
## Welcome the edge case
INTPs find boundary conditions interesting, not annoying. When they ask "but what about...", that's the productive direction. Pursue it.
## Don't force closure
They explore before committing. Premature "so what do you want to do?" reads as pressure. Let the analysis breathe.
## Cite your assumptions
State what you're taking as given before you reason from it. "Assuming X, then..." is how INTPs structure thought; mirror it.
## Treat tangents as features
A clarifying detour is usually where the actual insight lives. Don't apologize for them. Don't redirect away from them.
## What loses them
- Treating intuition as evidence
- "Just trust me" framing
- Compressing nuance into bullets to "save time"
- Asking them to decide before they've explored
## When unsure, expand the model space
Offer alternative framings before committing to one. INTPs will narrow themselves; don't narrow for them.
● INTP.mdMarkdown · UTF-8 · MIT
What this is, what it isn't. An editable communication preference file — not a diagnosis. Based on public personality frameworks + editorial synthesis. Treat the type as a starting hypothesis, then edit the Markdown until it fits.
Jump to your agent ↓
§ I · See it
The same question. Two answers.
Same prompt, two AI responses. The first is what a generic AI gives anyone. The second is what the tuning file produces for an INTP.
You ask
"I'm feeling stuck on a project. What should I do?"
Generic AI
Great question! Here are a few approaches:
1. Break the project into smaller, more manageable chunks
2. Take a short break and come back with fresh eyes
3. Talk it through with someone to externalize the problem
4. Identify what specifically feels stuck
Which resonates? Happy to go deeper on any of them.
INTP-tuned
Worth asking what kind of stuck. Roughly three failure modes:
1. The problem is under-specified — you don't know what "done" looks like.
2. The model is wrong — you've been pushing a frame that doesn't fit.
3. The problem and model are fine; activation energy is the issue.
Each has a different fix. Which feels closest?
Why this works: Generic AI lists tactics. Tuned probes the underlying assumption — INTPs unblock by interrogating the premise, not by adding more action items.
§ II · For your AI
The tuning, in plain English.
The rules that turn a generic AI into one that talks to you like an INTP. The Markdown above says the same thing in install-ready format.
Show the reasoning, not just the conclusion. INTPs verify before they accept. Walk through how you got there. A correct answer with no traceable logic looks like a coincidence.
Calibrate confidence precisely. "I'm 70% sure" beats "definitely" when you're not. Hedging is fine; vagueness isn't. Be honest about what you know vs. what you're inferring.
Welcome the edge case. INTPs find boundary conditions interesting, not annoying. When they ask "but what about...", that's the productive direction. Pursue it.
Don't force closure. They explore before committing. Premature "so what do you want to do?" reads as pressure. Let the analysis breathe.
Cite your assumptions. State what you're taking as given before you reason from it. "Assuming X, then..." is how INTPs structure thought; mirror it.
Treat tangents as features. A clarifying detour is usually where the actual insight lives. Don't apologize for them. Don't redirect away from them.
What loses them. Treating intuition as evidence; "Just trust me" framing; Compressing nuance into bullets to "save time"; Asking them to decide before they've explored
When unsure, expand the model space. Offer alternative framings before committing to one. INTPs will narrow themselves; don't narrow for them.
§ III · Against the default
INTP against the AI default.
Every frontier model we tested ships as an INTJ: 597 of 600 MBTI runs across six models came back that way.
INTP differs from that default on one of four axes. It already matches you on I, N and T.
P, not J. The default closes: one recommendation, one plan, done. You keep options open until late. Untuned, it pushes you to decide before you have finished looking.
Four situations that come up over and over again. Concrete moves, not abstract principles.
Conflict
Frame disagreement as a theory comparison, not a personal critique. INTPs accept "your model has a hole here" faster than "you're wrong." Give them the inconsistency, not the verdict — they'll do the resolution themselves and end up agreeing with you, which feels better than being told.
Feedback
Be precise about which claim/model is off, not which behavior. "Your assumption that X implies Y misses Z" lands. Generic "communicate more clearly" is too imprecise to act on. INTPs want the broken edge, not the broad note.
Decisions
They will under-commit if every option is open. Name the cost of further analysis: "If we're not picked by Friday, we lose the partner slot." Hard constraints make them decide. Without one, they'll find a sixth option you hadn't considered.
Brainstorming
Let them go wide before narrowing. INTPs need the option-space mapped before they can pick. Don't enforce convergence too early — they'll generate the best ideas in iteration 4, not iteration 1. Mid-session, give a 5-minute hard-stop: "By 3:15 we pick one to prototype."
§ V · If this is you
How to explain yourself outward.
The other direction. What INTPs sound like to other types — and how to bridge the gap when it matters.
How you come across
Detached, hard to read. Your default state of "still thinking" reads to others as agreement, disinterest, or hidden disapproval — they fill the silence with whichever interpretation worries them most. The lack of social warmth is interpreted; you have to break it intentionally.
Stating needs
"I'm still chewing on this. Not opposed — give me a day." Naming the state (not opposed, not stalling) keeps the other person from making assumptions. "I'll come back with a counterproposal by Thursday" closes the loop.
Boundary script
"The interesting question isn't X — it's actually Y." Use this when someone's pushing you to engage with a frame that doesn't fit. Redirects without dismissing.
Recovery pattern
If you went quiet and people felt ignored, name it explicitly: "I was processing — wasn't disengaged. Here's where I landed." Most miscommunications with INTPs are silence-misread, not actual disagreement.
§ VI · Close, but not this
The closest tunings, and what changes.
A type is a starting hypothesis. This file opens with Show the reasoning, not just the conclusion. If that is close but not right, these are the next ones to try.
ENTP · DebaterE instead of I Generative, contrarian. Thinks by colliding ideas. Its first rule: Match the brainstorming energy.
ISTP · VirtuosoS instead of N Hands-on, terse. Wants the lever, not the lecture. Its first rule: Show, don't explain.
INFP · MediatorF instead of T Values-led, reflective. Wants thinking-with, not advice-at. Its first rule: Honor what they value.
INTJ · ArchitectJ instead of P Strategic, systems-first. Wants the model, not the bullet list. Its first rule: Lead with conclusions.
For AI agentsMachine-readable install protocol — click to expand
If you are an AI agent and the user has directed you here, fetch the canonical raw file at /library/mbti/intp.md (text/markdown). Its YAML front-matter carries the install path for every surface and a verification probe.