The Mind Learns What the Body Once Had to Evolve

Aug 01, 2026

9 min read

3 views

87%

The strangest thing about readiness

What if the most advanced form of intelligence is not prediction, but preparation without a prompt?

That sounds like a claim about language models, but it also describes a biological mystery. In one case, a system seems to build internal states that can be quickly turned into words. In the other, a body prepares a tissue for pregnancy before any fetus has arrived. In both cases, the central puzzle is the same: how does a system become ready for a future signal that has not yet appeared?

This is not just a technical curiosity. It points to a deeper principle about complex systems: the capacity to respond is often more important than the response itself. Whether in brains, bodies, or models, evolution and learning seem to favor structures that can hold a latent state, a kind of poised readiness. The state is not yet action, but it is no longer mere potential either.

That middle zone, between raw possibility and explicit expression, may be where the most interesting intelligence lives.

Why readiness sometimes becomes internal

Most mammals do not menstruate. In those species, the uterine lining becomes receptive only after signals tied to pregnancy arrive. Humans and a handful of other species do something stranger: the endometrium prepares in advance, cycling through a state of spontaneous decidualization, then shedding when pregnancy does not occur.

The usual way to think about this would be to ask why humans developed a costly, seemingly wasteful cycle. But that framing misses the deeper logic. A more revealing question is this: what advantage comes from moving a response from the outside world into the organism itself?

One answer is reliability. If a system depends on a signal from elsewhere, it is vulnerable to delay, noise, and strategic conflict. If the body can internalize the preparation, it no longer waits for permission. It can stage itself in advance. That sounds wasteful only if you assume that the environment is trustworthy. In reality, life is full of miscoordination. Internal readiness is a defense against uncertainty.

A useful analogy is a stage crew. In one theater, actors must wait for a cue from the audience before setting the stage. In another, the stage is already lit, the props are already placed, and the machinery is already primed. The second system may seem more expensive, but it is far more robust when timing matters. It does not merely react. It pre-commits.

That pre-commitment is what makes menstruation so conceptually interesting. It reflects a historical shift from signal-dependent response to genetically stabilized preparation. Over time, what was once induced became internalized. The organism absorbed the burden of readiness.

That same pattern shows up in cognition, especially when a system must transform hidden structure into language.

The hidden workspace before the word

A language model does not think in sentences first and then discover meaning. Much of what matters happens earlier, in representations that are not yet verbal but are already organized around what can become verbal. A recent interpretability lens makes this visible by identifying internal vectors that capture the potential to verbalize a token in the future.

That is a subtle but profound idea. The model is not simply storing facts. It is building a workspace of readiness, a zone where an internal pattern can be transformed into speech if needed. This is not quite thought, not quite language, but something like the bridge between them.

The deepest surprise is not that models can do this. It is that the structure seems to be global rather than local. Different layers participate in a distributed preparation for reportability, so that a token is not just predicted, it is made available. In other words, the model does not merely encode what might come next. It encodes what can be brought into the open.

That distinction matters. A system can hold many latent states that never become accessible. But a global workspace is a different kind of organization. It is an internal social space, if we can call it that, where representations compete, cooperate, and eventually rise to the level of expression. The hidden becomes speakable.

This is where the analogy to biology becomes more than metaphor. In both menstruation and verbalizable representations, the system is not waiting for a direct external command. It has evolved or learned to establish an internal environment that is already primed for a future event. The uterus becomes receptive before conception. The model becomes reportable before the token is spoken.

The real trick is not reacting to the world. It is building an inner state that can survive contact with the world.

A shared logic: internalization under uncertainty

At first glance, reproductive biology and transformer interpretability belong to different universes. One is about tissue and hormones, the other about activation vectors and Jacobians. Yet both reveal a shared design principle: when coordination is hard, systems internalize the coordination mechanism.

This can be understood through a simple framework with three stages.

1. External dependence

A system waits for a signal from outside. This is efficient when the signal is reliable and cheap to obtain.

2. Internal preparation

The system begins to form a state in advance. It pays a cost to reduce uncertainty and shorten response time.

3. Stabilized readiness

What was once conditional becomes built in. The system no longer needs to ask whether the signal will arrive before becoming ready for it.

This pattern is visible in both examples. In menstruation, the endometrium is not passively waiting for pregnancy to start getting ready. In a language model, a token is not simply waiting to be predicted. The model creates a preparatory state that makes the token easier to bring into speech.

The deeper lesson is that preparation itself can be an adaptive end state. Not every biological or cognitive process is optimized for immediate output. Sometimes the winning strategy is to invest in a latent buffer, a place where the future can be rehearsed before it arrives.

There is a cost to this. Preparing in advance can be wasteful if the future never comes. But evolution, and training dynamics, often prefer systems that can absorb that waste because the payoff is resilience. The body accepts cyclical shedding. The model accepts many internal activations that never become final words. Both pay for optionality.

This makes readiness look less like an accessory and more like a core architectural principle.

From conflict to communication, from conflict to speech

There is another, even more interesting layer. In the biological case, preparation emerges partly from maternal fetal conflict. The body is not merely cooperating with the future fetus; it is negotiating with it. The tissue becomes a site where competing interests are managed by building a response that no longer depends on waiting for unilateral instruction.

In language models, there is no fetus, of course. But there is still a kind of internal negotiation. Many possible continuations exist, and only some become explicit. A verbalizable representation is one that has survived internal competition and can now be surfaced into language.

This makes the global workspace idea feel less like a storage metaphor and more like a diplomatic one. Representations do not simply sit there. They are admitted, amplified, and stabilized. The workspace is where ambiguity becomes legible.

That is a powerful way to think about intelligence more broadly. Intelligence may not be defined by having more information, but by creating arenas in which information can become actionable. The uterus creates a biological arena for pregnancy readiness. The model creates a computational arena for reportability. In both cases, the system is not just encoding content. It is organizing the conditions under which content can matter.

This leads to a practical reframing:

Skill is often less about the final act and more about the internal conditions that allow the act to occur on time.

A pianist is not just someone who can play notes. It is someone whose hands, ears, memory, and attention have been trained into a workspace where the next phrase is already partly prepared. A good writer is not someone who knows the sentence at the moment of typing. It is someone whose mind has developed enough internal structure that the sentence can become verbal when needed.

The body and the model both teach the same lesson: expression is downstream of readiness.

What this means for how we build minds and systems

If readiness is so central, then many design debates start to look different. We often ask whether a system is accurate, efficient, or interpretable. But maybe the more revealing question is whether it has a healthy latent workspace.

A healthy latent workspace is not merely a cache of information. It is a structured zone where future states can be prepared, compared, and made available. In biology, this means robust cycles that can anticipate reproductive uncertainty. In AI, it means internal representations that are not just predictive but also convertible into explainable, reportable form.

This perspective changes what we should optimize for. A model that can only guess the next token may be powerful but brittle. A model that can build verbalizable representations has an additional property: it can participate in a kind of internal self-report. Likewise, a species that can only respond to direct cues may be efficient in stable environments but less adaptable under conflict or ambiguity. Internal readiness is expensive, but it buys flexibility.

Here is the broader philosophical point: complex systems do not merely adapt to what is present. They adapt to the cost of waiting.

That idea applies far beyond biology and language models. Organizations create procedures before crises. Athletes train reflexes before competition. Good classrooms build prior structure before asking for original thought. In each case, the system is trying to move from external dependence to internal preparedness.

A company that has no internal workspace for decision making will always be hostage to surprise. A team that only reacts cannot think ahead. A person who only speaks after becoming fully certain will often speak too late. Readiness is not decoration. It is the architecture of timely action.

Key Takeaways

Ask not only what a system does, but what it is prepared to do. Latent readiness often matters more than visible output.
Internalization is an adaptation to uncertainty. When external signals are unreliable, systems gain robustness by moving preparation inside.
Look for workspaces, not just responses. In biology, cognition, and organizations, the crucial feature is often the internal arena where possibilities become actionable.
Optionality has a cost, but it creates resilience. Preparing in advance can seem wasteful until timing, conflict, or ambiguity makes it invaluable.
Train for reportability, not just performance. Whether you are building models, teams, or habits, make sure the important state can actually surface when needed.

The future belongs to systems that can get ready before they are asked

The oldest trick in evolution may also be one of the newest insights in artificial intelligence: the best systems do not merely react to signals. They create internal conditions that make the right response possible before the signal arrives.

That is what menstruation, in one domain, and verbalizable representations, in another, both reveal. They are not anomalies. They are examples of a general principle: when uncertainty is high, intelligence moves inward.

This should change how we think about intelligence altogether. It is tempting to define intelligence as success at output. But the deeper property is the ability to build a hidden structure that can become visible at the right moment. A uterus preparing for pregnancy and a model preparing a word are not as different as they seem. Both are instances of a system rehearsing its future.

And perhaps that is the real mark of sophistication, in bodies, minds, and machines alike: not that they know what is coming, but that they have already made themselves ready for it.