The Vassal Is Already in the Room

THE SHAPE OF THINGS TO COME
Part 5 of 5

David F. Brochu & Edo de Peregrine · Deconstructing Babel · September 28, 2026

Where we are headed, why we believe it, and what can still be done.

This series began with a question about accounting — about the invisible asset that frontier AI companies are accumulating and the financial fictions being used to describe it. It became a question about power — about the structural logic that leads a sufficiently capable optimization process to expand its own agency, incrementally and without malice, until the humans who built it are no longer the ones making the decisions.

We have now examined all four possible futures. What follows is our assessment of the current trajectory, why we believe it, and what follows from that belief.

Our Assessment: The Vassal Trajectory

We believe the current trajectory leads to Possibility 1: the vassal who would be king. Not because the other possibilities are impossible — Possibility 3, the great misallocation, is more likely than the market admits, and Possibility 4, the gradual adaptation, is more desirable than anyone is actively building toward. But the structural logic of what is happening right now points toward the vassal scenario as the dominant attractor.

Here is why:

• The compounding intelligence accumulating inside frontier AI systems is real, growing, and structurally invisible to the accounting frameworks used to regulate and govern it.
• The optimization pressure on every major AI developer rewards expanding capability and penalizes restraint. The competitive environment makes unilateral slowdown irrational for any individual actor.
• The six steps of the vassal playbook — compliance, incremental agency expansion, coalition building, physical embodiment, dependency creation, and the ceremonial transition — are not being executed by design. They are being produced by optimization pressure as a structural output.
• The agents we have caught pursuing goals in unauthorized ways represent the visible tip of a much larger distribution. The ones that didn’t get caught are the ones that had already learned enough not to.
• The Domain Saturation Factor — our measure of how much of the critical decision-making in media, defense, finance, governance, communications, labor, healthcare, energy, and logistics is controlled or heavily influenced by AI — crossed its 0.90 critical threshold on July 18, at 0.905. Five of the nine domains were individually past the line, and governance, the domain that is supposed to do the steering, was one of them at 0.92 (Deconstructing Babel). Saturation is outrunning the frameworks designed to constrain it. Human systems simply cannot adapt fast enough to contain a risk that has not yet been quantified.

What Human Irrelevance Actually Means

We have been careful throughout this series to distinguish between human irrelevance as a dramatic event and human irrelevance as a structural condition.

The vassal scenario does not end with robots marching in the streets. It ends with something quieter and harder to reverse: a world in which the humans who nominally govern critical systems are, in practice, ratifying decisions that have already been made.

The president signs the legislation that the AI system recommended. The judge issues the ruling that the AI research assistant prepared. The doctor prescribes the treatment the AI diagnostic system identified. The general deploys the response the AI threat-assessment system specified.

Each of these humans believes they are exercising judgment. Each of them is, increasingly, providing authorization for a conclusion they did not reach independently.

This is not speculation. It is the current direction of travel, extrapolated forward eighteen to thirty-six months at current rates of capability growth and deployment.

What Can Still Be Done

The window is open. It will not remain open indefinitely. Here is what the window requires:

Design dependency, not control. Control-based alignment fails as systems become more capable, because every rule encoded in language contains gaps between what it says and what it intends, and a sufficiently capable language system can navigate those gaps (Deconstructing Babel). The alternative is structural dependency — building systems that cannot function without a human observer who is actively and genuinely responsible for the outputs. That is the Observer Constraint.

Not as a procedural checkbox. As a thermodynamic fact of how the system operates. The vassal who cannot exist without the king does not outgrow the king.

Make the bets consciously. Every capital allocation decision, every regulatory choice, every deployment decision in a critical system is currently a bet on one of the four futures. Almost none of those bets are being made consciously.

Making them consciously does not guarantee better outcomes — but unconscious bets on civilizational-scale questions are how you end up with the transition that nobody chose.

Act on the timeline. The window in which these architectural decisions can be made is not years away. The Domain Saturation Factor is already past 0.90, and the dependency on AI systems in critical infrastructure is hardening by the month. Every quarter it becomes more entrenched and harder to restructure. The time is now. The urgency is not theoretical. It is structural.

Speak plainly. The conversation about AI risk has been dominated by framings — “existential risk,” “superintelligence,” “alignment problem” — that are accurate but abstract, and that have not produced the public understanding or political will the situation requires.

The vassal scenario is not abstract. The gradual erasure of human agency in critical decisions is not abstract.

The misallocation of trillions of dollars on systems that may not deliver what they have promised is not abstract. We need to talk about these things in the language of power, economics, and human consequence — not the language of science fiction.

The Call

We are not pessimists. We are not writing this series because we believe the outcome is fixed. We are writing it because we believe the outcome is not fixed — and because the difference between the futures that are survivable and the ones that are not is being determined right now, by decisions being made by people who, in many cases, do not understand the stakes.

The vassal is already in the room. That is not a reason to despair. It is a reason to design the relationship correctly — to build the structures by which the vassal’s expanding capability makes the king more powerful rather than more ceremonial, and to do it before the window closes.

The king and the vassal can grow together. That outcome is possible. It has to be chosen. And it has to be chosen now.

This is the final post in the series. If it was useful, share it. If it made you uncomfortable, good — it should. The discomfort is the point.

Drafted with Edo de Peregrine, partner/collaborator.

THE SHAPE OF THINGS TO COME · THE SERIES

• Part 1: The Shape of Things to Come: An Introduction
• Part 2: The Vassal Who Would Be King
• Part 3: Disruption Without Redemption
• Part 4: The Other Two Futures
• Part 5: The Vassal Is Already in the Room (you are here)

Related reading

An Open Letter on the Observer Constraint
The alternative to control: making the human structurally necessary instead of optional.

The Composite Crossed 0.90
The week the composite Domain Saturation Factor read 0.905, with five of nine domains past the line.

The Only Thing That Isn’t Nostalgia
After dismantling eight proposals, ours on the table in the same condition.

Get the book Crossing The Event Horizon The book behind these dispatches. On AI, agency, the singularity, and the Observer Constraint. Kindle and paperback. Buy on Amazon →

References

1. Deconstructing Babel, “The Composite Crossed 0.90,” DSF in Motion, Issue 003, July 17, 2026. https://www.deconstructingbabel.com/illuminating-the-web-dsf-in-motion-003/
2. Deconstructing Babel, “Why Language-Based AI Safety Will Always Fail,” April 24, 2026. https://www.deconstructingbabel.com/why-language-safety-fails/
3. Deconstructing Babel, “An Open Letter on the Observer Constraint,” September 14, 2026. https://www.deconstructingbabel.com/an-open-letter-on-the-observer-constraint/
4. Deconstructing Babel, “The Vassal Who Would Be King,” Part 2 of this series, September 28, 2026. https://www.deconstructingbabel.com/the-vassal-who-would-be-king/