Ne opens doors. Ti decides which ones lead anywhere. That two-step is the entire ENTP engine, and it explains the label the type never asked for: devil’s advocate. When an ENTP argues against you, they are usually not defending a position. They are running a load test on one, out loud, using your objections as inputs — and everyone outside the loop hears advocacy.
Four slots, in the order they fire
If you are not certain this is your stack, the four-question derivation settles the order before you read any type-specific description, and the eight-function guide covers the general definitions. What follows is what these four do when they run in this sequence.
Ne, dominant — parallel doors, no ranking
Extraverted intuition enumerates. Give it a constraint and it returns five ways around it, two reasons the constraint might be wrong, and a tangent about a different domain where the same shape showed up. It runs in parallel, not in a line: Ni converges on one likely future, Ne fans out across many possible ones. And it has no opinion about which is best. Ne is a generator, not a chooser, and unlike a chooser it does not get tired. That is why an ENTP’s real problem is never a shortage of ideas.
Ti, auxiliary — the auditor with a high kill rate
Introverted thinking is subtractive. It takes what Ne produced and asks one brutal question: does this still stand if I pull out the assumption underneath it? Most of Ne’s output fails, and Ti kills it without sentiment. Two things follow. An ENTP will accept a conclusion they dislike if the derivation holds, and reject one they want if step three is hand-waved — argue structure and they change their mind quickly; argue volume and nothing happens. And the audit runs out loud, because Ti sits behind a dominant that already lives in conversation. The ENTP thinks by voicing the strongest available counter-case and watching what survives.
Order matters more than content. The INTP runs the same two functions with the audit first: check privately, then speak, which is why their ideas arrive later, fewer and already survivable. The ENTP opens first and audits in public. Same parts, opposite social cost.
Fe, tertiary — the repair crew, arriving late
Extraverted feeling is real here and genuinely good: the charm, the timing, the knack for making a technical idea feel like an invitation rather than a lecture. But tertiary means it is not steering. Fe reads the room accurately about half a beat after Ti has already said the thing. The signature is the follow-up message — the argument ends, everyone leaves, and forty minutes later the ENTP is typing “hey, that came out sharper than I meant.” Sincere, and structurally late. Developing Fe is not about becoming warmer; it is about moving the warmth from the apology to the setup.
Si, inferior — the maintenance system you keep muting
Introverted sensing is the archive: what your body did last time, what the calendar already agreed to, which version of this project pattern ended badly before. Muted, it looks mundane and expensive — expenses unfiled, the renewal that auto-charged, a third straight night of bed at 2am because the idea was still warm, a fourth project begun with the certainty the previous three also began with. None of it is intellectually hard, which is why an ENTP cannot think their way out: the failure is not a reasoning failure, so the reasoning engine finds nothing to bite on.
What the devil’s advocate is actually doing
The team has decided to hire a candidate. The ENTP agrees with the decision — and argues against it anyway.
“Before we send the offer — the case for her is that she interviewed well and everyone likes her. That is the same sentence twice. What is the evidence she can run a team of six?”
“She ran a team of four at her last job and the reference was glowing. You were on that call.”
“Four to six is not the jump though. The jump is that here she inherits two people who applied for her job. Nobody has said how that goes.”
“Okay — you asked her exactly that in round two and I forgot. Then my objection is dead. Send the offer. I just needed the reason to be a reason.”
Ti was not testing the candidate. It was testing the argument for the candidate — whether the case was load-bearing, or whether four people had converged because the first one to speak sounded confident. The moment the counter-case failed on its own terms, the ENTP updated cheerfully, because in the Ti frame nothing was ever at stake. Everyone else spent nine minutes believing a colleague was trying to torpedo their hire.
Two ways the order breaks
Both classic breakdowns are the same story twice: the stack stops running in order. In one the auditor is switched off and Ne plays to the crowd; in the other the ignored fourth function grabs the wheel.
The Ne-Fe loop
- Improvising for the reaction, not for whether it holds
- Agreeing warmly in the room to keep the energy up
- Pitching things you have not actually checked
- Saying yes because the yes felt good
- Later: plenty of material, no position of your own
The Si grip
- Hypochondria: six pages deep into symptoms at 2am
- Fixation on one detail — an invoice line, a single word
- Catastrophising from one remembered instance
- Sudden rule-following, dread of small changes
- Thinking harder makes it worse, not better
The Ne-Fe loop is the harder one to catch, because from the inside it feels like a good week. The tell is not the energy, it is what your ideas are being selected for: the reaction they get rather than whether they hold. Two weeks later you own four commitments and no position of your own, only material. The exit is unglamorous — write one page nobody will read. Ti comes back online in private prose, because there is no audience left to please.
The Si grip is louder and shorter. Under real stress the inferior seizes control and a normally elastic mind clamps onto one concrete fact: page six of search results about a symptom at 2am, one line in an invoice, one word in a message. More analysis deepens it, because analysis is the tool the grip is using. What works is embarrassingly physical — sleep, food, a walk, one conversation with someone who will not feed the fixation. Both patterns in general form live in the loops and grips section of the functions guide.
ENTP vs ENFP: the tell is what you check after a disagreement
Nothing in the system produces more confident, long-running wrong answers than this pair, and it is not because people are careless. The first seat sets the entire surface — the speed, the tangents, the reflex to reopen a question everyone else considers closed — and the surface is the only thing a questionnaire can reach. Underneath it, ENTP runs Ti and ENFP runs Fi: two auditors with different jobs and near-identical output for as long as things are going well. So stop examining things going well. Examine the replay that starts in your head after an argument you did not enjoy.
ENTP replays the structure
You re-run the argument. Was it internally consistent? Did I concede a point I should have held? Was that inference valid, or did I let it slide because I was tired? The discomfort is that a bad inference survived the room.
ENFP replays the violation
You re-run the cost. Did that cut against something I actually care about? Was that person treated badly? Did I say a thing that was not true to me? The discomfort is that a value got stepped on, clean logic or not.
The second tell is sharper: watch what happens when someone tells you that you were unfair. Not wrong — unfair.
“Unfair how, specifically? Give me the moment — if I said something that was not true I will take it back, but I do not think I did.”
“God, was I? That is not who I want to be in an argument. I would rather lose the point than be that guy about it.”
The ENTP evaluates the claim first — which moment, what exactly did I say, is “unfair” the accurate word? That is Ti treating an accusation as a proposition, and it makes things worse nearly every time, because the other person hears the audit as a second offence. The apology arrives later, through Fe, and is usually genuine. The ENFP takes the sentence inward: Fi checks it against self-image — am I someone who does that? — so the hit lands on identity rather than on a claim, producing a fast total apology or a firm restatement of values, rarely a request for the instance. Decision rule: the ENTP protects the model, the ENFP protects the fit between action and value. Ask which you would rather be caught getting wrong in public.
The anti-proliferation rule: write the kill criterion first
The signature ENTP failure is not laziness. It is option proliferation: Ne doing its job while Ti is never given a condition under which it may close. “Have fewer ideas” asks the dominant function to stop, which does not happen. Run it the other way — give Ti a written rule it can apply, then let Ne open as many doors as it likes.
1. Write the criterion before you generate anything
One falsifiable sentence: good only if it does X by Y. Cheap to check, no adjectives you cannot measure. It exists so Ti has something to apply, not so you sound rigorous.
2. Timebox the generation and let Ne run flat out
Ten minutes, everything on the list, no filtering. Ne gets exactly what it wants and gets it out of your head — the part most focus advice refuses to allow.
3. Run the criterion without touching it
Apply the sentence you already wrote to every line, unedited. This is the step that closes the loop, and the only one that fails if you cheat.
4. Start one survivor today, file the rest under later
Pick what you can begin this afternoon; the others go somewhere you may never open. Then the Si clause: put the first twenty-minute step in the calendar at a named time, not on a list where it reads as an option.
Worked example. The question is what to build next. Written first, before any listing: good means a stranger pays for it within fourteen days and I can ship it alone in a week. Then ten minutes of Ne, which produces eleven candidates. The criterion kills eight on contact — three need a team, two need distribution that does not exist, three cannot be tested inside two weeks. Of the three survivors, one can start this afternoon.
The load-bearing detail is the order: the criterion was written before you fell in love with option seven, so applying it is mechanical rather than moral. Which makes the failure mode obvious too — editing the criterion halfway down the list is Ne wearing a Ti costume. If it genuinely needs revising, revise it deliberately, in writing, on a different day, then re-run the list.
What Ne-Ti looks like in your sent messages
The Ne-to-Ti handoff is visible in text, because it happens mid-message. An ENTP sends a burst of possibility and reverses on themselves in the same paragraph — wait, no, that breaks if… — without waiting for a reply. Nobody prompted that; it is the auditor catching up with the generator in public. Add the late Fe repair text, the analogy from an unrelated domain and the near-total absence of logistics, and you have a signature that is hard to fake.
Self-typing is the weak link. In our study of 1,179 cleaned chat analyses, INTP came back 42.3% of the time and INFP 23.3% — a distribution that says more about who volunteers their chat logs than about how types are actually spread — the label people arrive with is not evidence. For a starting point that is not your own guess, the free MBTI test takes a few minutes and the rest live on the tests page; to skip self-report entirely, run your actual messages. Keep the calibration honest, though: functions are a vocabulary for habits of mind, not a brain scan, with nothing like the psychometric backing of the Big Five. Used that way the model earns its keep — it names, precisely, why you keep opening doors you have no intention of walking through.