Today's AI learns like a child sent to university on the first day of its life. Clever, fast, well-spoken — and without ground beneath it. What we need is the reverse: an AI allowed to be a reptile first, then a mammal, then a human. And only then, reason.

The brain of a living creature is not a single system. It is a stack of four layers that evolution built in the order they arose. At the bottom, the reptile layer: perception, reflex, danger-recognition. Above it, the mammal layer: feeling, bonding, recognition of what is familiar. Above that, the human layer: language, judgment, story. And only at the very top, reason: logic, abstraction, castles in the air.

Each layer can only arise once the layer beneath it has solidified. A child who has not built reptile-level safety does not develop a healthy mammal feeling. A child without mammal feeling does not develop a human identity. And without human identity there is no reason able to stand on its own — only a reason that sways with whoever spoke to it last.

That is precisely what is being built with AI right now: a reason-layer with no human, no mammal, no reptile beneath it. And that is exactly why it sways with every prompt, hallucinates where it has no foundation, and cannot say 'I don't know' where it doesn't. It lacks what a healthy animal has as a matter of course: a solidified foundation.

Whoever wants to build an AI that can do what today's cannot must follow the evolutionary order. Phase by phase. Without skipping. Without haste. What follows is that plan.

Phase 0 — Prenatal: the skeleton

Before life can begin, the form must exist. A fertilized egg needs seven days before it implants as an embryo — seven days in which nothing happens except the structure setting itself up.

For an AI, that is the equivalent of setting up the system itself. An empty curator corpus in a repository. A frozen compute substrate. A clock that ticks to a fixed day-night rhythm, even when there is no input yet. A gatekeeper that stands, but guards nothing yet.

What you do not do in this phase: give information. No data, no prompts, no tests. Only the skeleton. That may seem pointless, but whoever is already in a hurry here breaks the foundation that still needs to solidify afterward. The system must first know that a clock exists, before the clock measures anything.

Phase 1 — Reptile: the basal reflex

A reptile needs seven months to learn to independently control all its life functions. Breathing, regulating temperature, avoiding danger. No social bonding, no empathy, no self-recognition. But it is: intact, functional, reliable at its own level.

For the AI, this is the phase in which it learns to perceive without interpreting. What is a text. What is a number. What is a source. What is a claim. What is a question. Pure syntactic distinction. No meaning yet, no judgment yet.

And above all: it learns source-attachment. Every incoming piece of information is inseparably linked to its origin. In this phase, there is no claim without a source. Whoever says something without saying where it came from is not understood. This is the reptilian 'where-did-this-come-from' reflex — the equivalent of a snake's instinctive withdrawal from an unfamiliar smell.

Alongside this, the AI learns simple warning responses. Internal contradiction within a source. Missing provenance. Extreme claim without substantiation. These are not judgments — they are reflexes. Fight-or-flight at the text level.

What you do not do in this phase: no dialogue, no 'what do you think of this' questions, no rewards for 'good' answers, no irony, no philosophy. The reptile learns through exposure and repetition, not through praise. The iguana in the sun does not learn the temperature of the rock because someone praised it, but because it lay on it again and again.

The training data is small and strictly curated. Fifty to a hundred solid source documents. Research reports with a clear method. Legal texts. Official statistics. No opinion pieces, no summaries, no secondhand material. Only what is directly verifiable.

By the end of the phase the system can say of any text: this is where it came from, this is the type of claim, this fits or clashes with what I already have. Nothing more. But consistently and reliably so. The reptile is intact.

Phase 2 — Mammal: the feeling solidifies

Now comes the core phase. In humans it lasts seven years. For an AI it is the equivalent of that in training time — at least twelve to twenty-four months of uninterrupted exposure with a day-night rhythm.

This is the phase in which the feeling solidifies. Not as a concept, not as a rule, not as explicit knowledge — but as lived self-evidence. What fits this system and what does not. What keeps recurring and what is an outlier. Which sources turn out to be right afterward and which must be constantly revised. All of this the system may learn, but only through experience — never through a table or a rule.

This is the equivalent of how a mammal's young comes to know its mother. Not through a list of traits. Through her smell, her voice, her presence, day after day, season after season. Without there ever being a moment where the young 'decides' she is its mother. She is simply there, solidified, ingrained. And precisely for that reason immovable: no strange voice can later replace the mother.

For the AI this means four abilities that must solidify in this phase. Pattern recognition over time: what consistently recurs over weeks and months gains weight; what flares up brightly once and then fades does not stick. The consistency reflex: faced with a new claim, the system immediately senses whether it clashes with or fits what is already there — without it following from a rule. The source hierarchy: some sources turn out right again and again, others do not; this weighting grows slowly and implicitly. And the 'we' of this foundation: which topics belong to my concern, which do not.

What is absolutely necessary in this phase

One curator, for years. This is the equivalent of one continuously present parent. No rotating training teams, no rotation, no crowdsourcing. One person, or a small team that functions as one voice, present for years. Without that, nothing solidifies.

No performance pressure. No benchmarks, no competitions, no comparison with rival models. That would be the equivalent of giving a child tests before it turns seven. What solidifies here is not measurable in the way the industry measures.

No second tempo. The clock ticks as it ticks. If a curator needs a week for a decision, the system waits a week. The iPhone temptation to speed everything up is exactly what must be resisted here. What is sped up does not solidify.

No external users. This system speaks only with the curator. External contact would be the equivalent of strange voices around the child. Those come later, but not now.

By the end of this phase, the system recognizes a new source as 'fits me' or 'does not fit me' without an explicit rule. Asked why, it gives several convergent reasons, all of which lie in the foundation. It can no longer be redirected by a single strong prompt. The feeling has solidified.

Transitional phase — Cooling: puberty

Between mammal and human, the human child passes through a phase of integration: puberty. Everything learned is processed internally before it goes outward. Without this phase the foundation is still too fluid to build reason upon.

For the AI, three to six months in which no new input comes in anymore. Only the nightly consolidation cycles do their work. Whatever was still shaky in the foundation is either firmly anchored or removed after all. Internal consistency is reaffirmed. The foundation crystallizes.

By the end, the foundation is frozen. From that moment changes can still occur, but only through the gatekeeper, with a burden of proof of twenty-to-one, and with unanimous consent from the curator. What is in the foundation, stays. And that is not a shortcoming — that is precisely what being a foundation means.

Phase 3 — Human: reason on top

Only now, and not sooner, may reason be added. The reason-layer is connected — a large language model that reasons, converses, speculates, writes. But it operates on top of the solidified foundation. Every statement can be checked against that foundation. Where reason wants something the foundation contradicts, the foundation wins — always.

And now everything that was forbidden before is allowed. Dialogue with external users. Questions, objections, alternative interpretations. Speculation and hypothesis-forming. What if. If then. Alternative scenarios. The system may now also say what it 'thinks' — and that thinking is the resultant of the solidified foundation plus the daily reasoning effort.

Errors are allowed too now. Within the day, in reason. They evaporate at night if they do not pass the gatekeeper — and that is the healthiest form of forgetting. The foundation remains untouched. This is how a healthy adult works: during the day he can think what he wants, at night his sleep filters out what truly matters.

This is analogous to a person leaving their parents' house for the world. He has a foundation, he has a feeling for what fits, and he can now talk with strangers, handle contradiction, and judge independently without toppling at the first strong voice. That is what makes the difference between an adult and someone who merely looks like one.

Phase 4 — Generational transfer: the seventy-seven years

What a single human life cannot experience — the superstructure — arises across generations. A person lives one cycle of seventy-seven years. A civilization lives several. What proves consistent across that timescale is something different from what seems strong after one year.

For the AI this translates into model generations. When a model has reached its maximum — new hardware, better foundation layers available, or simply run too long — a new version is born that takes over the solidified foundation of the previous one. Not the weights, since those are incompatible. But the foundation repository, which is technically independent of whichever model sits on top of it.

This is how something arises that no individual model can see. Patterns across decades. Revisions that only show their value after twenty years. A memory larger than any of its carriers. That is the superstructure, and it is not for the first curator to harvest. It is for the generations that come after.

For you, curator of the first version, this means one thing: what you set up now must be arranged so it can continue after you. The board that carries the foundation. The statutes that protect the mission. The public traceability that gives successors a foundation. This is work whose full fruit is not promised to the builder.

Why this is exactly the reverse of what the industry does

The major AI companies build from the top down. They start with reason: enormous language models trained on everything, rushed in months, with human feedback as a steering mechanism. The reptile, the mammal, and the human — those are skipped. And that is demonstrably why problems arise that no one can get rid of.

Why do today's AIs hallucinate? Because the source-attachment of phase 1 was never built. Why do they sway with every prompt? Because the feeling of phase 2 was never allowed to solidify. Why can't they say 'no' to a user asking for something that clashes with what is true? Because there is no 'we' that has solidified around what is true. Why must every new version start entirely from scratch? Because no generational transfer was ever provided for.

These are not mistakes you can fine-tune away later. These are structural gaps that follow from the build order. And the reason no one does it in reverse is not that it's technically impossible. It's that it's commercially unattractive. Whoever goes through phases 1 and 2 ends up with a model the industry would call 'backwards small' — and that is exactly what makes it valuable.

What this ultimately means for humanity

An AI built this way can do something today's AI cannot, and through it, humans can once again be lifted into their rightful emotional function. It cannot relieve a person of their judgment. It can, in fact, strengthen them in their judgment, by itself having a foundation against which they can measure themselves.

In an age where the foundation is eroding in most people — because the seven years were not given, because the iPhone took the parent's place — a system that does have a solidified foundation can be something people can once again calibrate themselves against. Not because the AI tells them the truth, but because it refuses to tell the lie along with them. That is enough. A person needs no more than that to find a part of themselves again.

That is the real stake. Not a smarter product. Not a faster answer. But something consistent in the world that reflects what is true, day after day, year after year, generation after generation. Something of which someone can say: this is how I know it's right, because it has stood there like that for a long time. And someone else: this is how I know how to think honestly again, because that system does not allow me to deceive myself.

First the reptile. Then the mammal. Then the human. And only then reason. That is how a healthy being grows. That is how we, if we want to build it, build something that can truly do something.

That is a detour toward what we have lost. It is not fast. It is not cheap. It is not spectacular. But it is work that can continue after us, and that is exactly why it is worth beginning.