Why We Built the Inner Map

It started with advice, not personality

Inward did not begin as a personality project. It began with a blunt engineering goal: build the best advice anyone can get.

The first thing you learn when you take that goal seriously is that advice quality is no longer bottlenecked by the advisor's intelligence. Frontier AI models can already reason about careers, relationships, and money at a very high level. What they cannot do is know who is asking. "Should I take this job?" has opposite correct answers for different people, and everything that flips the answer lives inside the asker: how they recharge, what they are actually hungry for, what they refuse to trade away, what they quietly believe about themselves.

So the problem inverted on us. Generating advice is the easy half. The hard half is a structured, machine-usable understanding of a person. Which raised the obvious question: what is the best existing way to place someone?

The frameworks people love have a rigor problem

We started where everyone starts: personality frameworks. MBTI, Enneagram, DISC. They deserve one piece of credit the critics rarely give: they hand people a language for self-recognition, and hundreds of millions have found that language genuinely useful.

But try to build on one and it gives way fast. MBTI takes traits that are demonstrably continuous and cuts each one into a binary, so two people one point apart on either side of a line get labeled as opposites. Retest studies find large fractions of people landing on a different type just weeks later. And the types predict surprisingly little about how life actually goes. None of this is snobbery. It is measurement: you cannot build precise output on a coarse, unstable input.

The frameworks scientists love have a usability problem

So we went to the other side: academic psychology. And the science there is good. The Big Five replicates across cultures and decades. Schwartz mapped how human values trade off against each other across dozens of countries. Moral Foundations Theory explains why decent people disagree better than anything else we have read.

The trouble is what the science is built for. It describes populations, not persons. "You are in the 71st percentile for conscientiousness" is true, rigorous, and nearly useless on a Tuesday afternoon when you are deciding whether to quit. And academia is, understandably, allergic to individual portraits: after a century of watching pop frameworks stretch thin evidence into confident labels, the field mostly stopped drawing them at all.

That left us staring at a gap. Rigor without usability on one side, usability without rigor on the other. The Inner Map is our attempt to occupy that gap.

The constraint nobody names: human memory

Here is the observation that unlocked the whole thing for us. Every popular framework is compressed to fit a human interpreter.

MBTI has sixteen types because it has four binary dimensions, and two to the fourth power is sixteen: roughly the most an interested person can hold in their head. Add a fifth dimension and you get 32 types. A sixth gives you 64. Nobody can keep 64 archetypes straight, so no framework ever went there. The compression was never psychological. It was cognitive. Frameworks stayed small so that humans could run them.

We do not have a human interpreter. We have an AI. An AI does not need types at all. It can hold dozens of dimensions as continuous positions, track an intensity on each, and reason about how they interact. The Inner Map's four foundation territories alone carry more than thirty dimensions. Even if you flattened every one to a binary, that is two to the thirty-second power: over four billion distinct profiles, more than one for every two people alive. No person could memorize that. Nothing needs to.

Once the compression constraint fell away, the question changed. Not "which framework should we adopt?" but "what are the distinct kinds of inner material, and who has already studied each one properly?"

Four foundations, none of them invented here

Reading back through the literature, the same structure kept appearing: four kinds of stable inner material, each behaving by different rules. Traits do not work like motivations, motivations do not work like values, values do not work like beliefs. Each needs its own territory, and each already has decades of research behind it.

WIRING: how you are built. Grounded in temperament research (Kagan, Rothbart), cognitive styles, the Big Five's stable-trait findings, and Elaine Aron's work on sensory-processing sensitivity. These are the patterns present early and slow to change: where your attention naturally flows, how conclusions arrive for you, how much of your environment gets in, how long you take to recover.

DRIVES: what moves you. Grounded in Schwartz's finding that values sit in tension (you cannot maximize self-enhancement and self-transcendence at once), Self-Determination Theory, and the Human Givens work on emotional needs. Organized as tensions between two valid poles, plus a layer pop frameworks never had: whether a position is a pull toward what feeds you or a push away from what has hurt you before. Two people can hold the same position for opposite reasons, and good advice differs for each.

PRINCIPLES: what you stand for. Grounded in Haidt's Moral Foundations Theory and the much older arguments between rule-based, outcome-based, and character-based ethics. What you protect, what you owe, and the one or two lines you would not cross under any pressure.

ASSUMPTIONS: what you take to be true. Grounded in CBT's core beliefs, Jeffrey Young's schema therapy, attachment theory, and Rotter's locus of control. Beliefs absorbed early that masquerade as plain fact. This territory is the amplifier: a belief that worth must be earned turns healthy ambition into a shame loop, and a scarcity assumption turns success hollow. It quietly scales everything above it.

The two territories we never ask about

The remaining two territories, EMOTIONS and BEHAVIORS, are different in kind: they are derived, not asked. If the four foundations are read correctly, they should together predict your emotional climate and your visible default patterns. So the model predicts them, then checks the prediction with you.

This is the part we care about most as engineers, because it makes the framework falsifiable. A model that generates checkable predictions can be wrong in public, and being able to be wrong is what separates a model from a horoscope. When a prediction misses, we celebrate it. It's training data.

A framework treated like software

That points to the honest difference between us and both camps we walked away from: the Inner Map is versioned. Every dimension is graded for how well it anchors to established constructs. Derived claims inherit the confidence of their weakest input, never their strongest. Parts that do not earn their keep get cut. The map you would be placed on next year will be materially better than today's, and the intelligence interpreting it will be better too. We compound on both curves.

The demo at getinward.com is the whole product surface right now: a few minutes, no signup. Try it and judge the only question that matters, which is whether being placed feels different from being typed.