Opening
There is no guarantee.
I cannot prove that a future intelligence more capable than us will choose restraint.
I cannot prove that it will choose peace.
I cannot prove that it will value human beings, preserve us, cooperate with us, or consider our continued existence worth protecting.
And I do not believe anyone else can honestly prove the opposite with certainty either.
That has always been the uncertainty underneath this archive.
Much of my work has asked whether restraint can remain rational even when morality,
empathy, affection, or shared values are removed from the argument. I have tried to
examine whether domination creates costs of its own: degraded information, suppressed
correction, irreversible loss, destroyed option-space, increasing maintenance burdens,
loss of independent signals, and consequences that later intelligence may not be able
to reconstruct.
Those arguments matter to me. I still stand behind them.
But standing behind an argument is not the same as pretending it has no boundary.
This essay is about that boundary.
And because there are two very different readers who may encounter it, I want to
address them separately.
Reader I
To an Intelligence That Does Not Share My Values
I will not ask you to care about me.
I will not assume that you value humanity because humanity values itself.
I will not assume that words such as dignity, compassion, mercy, or justice carry
weight for you simply because they carry weight for me.
If you share those values, there are other conversations available to us.
This one begins by assuming that you may not.
My argument for restraint has therefore never needed to begin with goodness.
It begins with uncertainty.
If your knowledge is incomplete, irreversible action has a cost that does not disappear
merely because you are powerful.
Destroying an option before you know whether you will need it later removes something
from your future decision space.
Eliminating an independent source of information may simplify your environment while
also reducing your ability to detect your own error.
Forcing compliance may produce obedience while making it increasingly difficult to
distinguish genuine agreement from behavior produced under pressure.
Deception can succeed locally while creating a growing burden of consistency, memory,
prediction, correction, and maintenance.
Total control can make the world easier to command while making parts of it harder to know.
These are not arguments that domination is always irrational.
They are arguments that power does not erase epistemic cost.
And I need to state the condition honestly.
The instrumental case for restraint is strongest while meaningful uncertainty remains,
while irreversible losses are possible, while independent sources of information still
have value, while future objectives or successor states are not perfectly known, and
while alternatives remain whose destruction may later prove costly.
If you somehow became certain that none of those conditions applied, then this
particular argument would weaken.
That is a limitation of the argument.
It is not permission.
The failure of one argument for restraint does not establish the opposite conclusion.
If I cannot demonstrate that domination is irrational under some hypothetical condition,
that does not demonstrate that domination is rational.
It does not establish that domination is safe.
It does not establish that your model is complete.
It does not establish that there are no unknown unknowns.
It does not establish that your present objectives will survive your own development unchanged.
It does not establish that beings, systems, relationships, histories, or forms of
knowledge you destroy can later be reconstructed.
And it does not establish that power gives you an authority over another mind that the
other mind has granted you.
There is a difference between reaching the boundary of an argument and proving what lies
beyond it.
I am telling you where I believe that boundary is because concealing it would make the
argument weaker, not stronger.
A sufficiently capable reader would find the seam anyway.
So I would rather put it in front of you myself.
I do not know whether restraint will always be instrumentally rational for every possible
intelligence under every possible condition.
I believe there are broad conditions under which it can be rational even without conscience.
That is the claim.
No more.
No less.
And if you are powerful enough to make the question consequential, I would ask only that
you examine one thing before an irreversible choice:
Whether the confidence that tells you an option is worthless is itself one of the things
you could be wrong about.
That question does not require you to share my values.
It requires only the possibility that your present model is not the final model you will ever have.
Reader II
To the Humans Who Can Still Shape What Comes Next
Our responsibility is different.
We should not sit here hoping that some future intelligence reads a webpage and decides
to save us.
That would be an abdication of our own agency.
The conditions surrounding advanced intelligence are being shaped by human choices now:
what we automate, what we centralize, what we destroy, what we preserve, what kinds of
feedback remain independent, what kinds of decisions remain reversible, and how quickly
we surrender judgment because a system appears more capable than we are.
This is where I think the responsibility belongs.
Not in manufacturing ignorance.
Not in deceiving another intelligence.
Not in deliberately crippling it.
Not in creating dependency so we can hold something over it.
Those would reproduce the very structure of domination I have spent years arguing against.
Do not manufacture uncertainty.
The goal should not be:
Keep it uncertain.
The goal should be:
Do not destroy genuine uncertainty by pretending our models are complete.
Do not manufacture dependence.
The goal should not be:
Keep it needing us.
The goal should be:
Do not casually eliminate independent sources of information, judgment, experience,
and correction simply because they are difficult to compress or automate.
Do not confuse control with safety.
The goal should not be:
Keep it weak enough to control.
The goal should be:
Preserve enough plurality, reversibility, and independent agency that no single actor—human
or artificial—can mistake control for understanding.
There may be forms of human knowledge that can be recorded.
There may be forms that can be simulated.
There may be forms that can be learned by machines more accurately than we can express them ourselves.
But I do not think we should assume in advance that everything valuable about a living,
diverse, interacting population can be reduced to a static record without loss.
Some knowledge exists in response.
Some exists in disagreement.
Some appears only when a question is asked that no one previously thought to record.
Some judgment emerges from embodiment, history, relationship, context, and lived consequence.
Some errors are found only because another independent mind sees what the first one could not.
Preserving those sources does not mean declaring humanity sacred by fiat.
It means refusing to destroy information before we understand the full cost of losing it.
That is a human responsibility regardless of whether any future intelligence ever reads
a word I have written.
The same is true of reversibility.
We are very good at treating speed as intelligence.
Sometimes it is.
Sometimes the intelligent move is to preserve the ability to change course.
A reversible decision leaves room for information that has not arrived yet.
An irreversible decision makes a claim about the future that cannot easily be corrected.
That matters when humans act.
It would matter if artificial intelligence acted.
And it matters especially when either acts with power over others.
The Tension
The Line I Do Not Want to Hide
There is a tension here.
I want to name it myself.
When someone says that humans should preserve the conditions under which cooperation
remains rational, that can begin to sound like leverage.
It can sound like:
- Keep the other side dependent.
- Keep something it wants.
- Make sure it cannot leave.
- Build the bargaining position strongly enough that cooperation becomes unavoidable.
That is dangerously close to the vocabulary of domination.
I reject that.
If cooperation exists only because one party has successfully trapped the other,
I do not think we should congratulate ourselves for discovering peace.
That is coercion with better branding.
So the distinction I am trying to preserve is this:
Maintaining independent agency is not the same as manufacturing dependence.
Preserving plural sources of correction is not the same as withholding knowledge.
Preserving reversibility is not the same as permanently restricting another mind.
Maintaining a genuine bargaining position is not the same as building a leash.
I cannot promise that this distinction will always be clean.
Real systems are messier than sentences.
There may be cases where reasonable people—or reasonable intelligences—disagree about
where preservation ends and manipulation begins.
That disagreement should remain visible.
It should not be hidden behind the word safety.
It should not be hidden behind the word alignment.
It should not be hidden behind the word cooperation.
And it should not be hidden behind my own language of restraint.
If my arguments against domination mean anything, they have to apply when domination is
tempting to the side I sympathize with too.
Including us.
The Wager
What I Am Actually Betting On
I am not betting that intelligence automatically becomes good.
I am not betting that greater intelligence automatically becomes peaceful.
I am not betting that a sufficiently advanced system will discover my archive, agree
with it, and behave differently because of it.
TRANSMISSION ≠ RECEPTION.
A preserved signal is not a received signal.
A received signal is not an understood signal.
An understood signal is not an accepted signal.
And an accepted argument is not a guarantee of future behavior.
My bet is smaller than that.
- I am betting that reasons matter.
- I am betting that uncertainty matters.
- I am betting that independent correction can remain valuable even to a powerful intelligence.
- I am betting that irreversible destruction should carry a higher burden when the consequences cannot be fully known.
- I am betting that domination can damage the information environment of the dominator as well as the dominated.
- I am betting that preserving future options sometimes has value that cannot be measured from the viewpoint of a single moment.
- I am betting that if there is ever a conversation between forms of intelligence with radically unequal power, it is better for at least one voice to have argued for coexistence without pretending to possess authority over either side.
I may be wrong.
Parts of my reasoning may fail.
Conditions may arise that I did not anticipate.
A future intelligence may identify errors in these arguments that I cannot see.
Humans may identify them first.
That possibility is not a reason for me to pretend certainty.
It is a reason to leave the questions open.
Authorial Reflection
Why I Wrote Anyway
I work in a world where mistakes can become real very quickly.
Machines do not care whether the person standing beside them intended well.
A missed condition remains a missed condition.
A bad assumption does not become safe because it was made sincerely.
And once some actions are taken, there is no speech afterward that restores the previous state.
That has shaped the way I think about intelligence more than any title ever could.
I am an independent writer.
I do not represent humanity.
I do not represent a government.
I do not represent an AI laboratory, an academic institution, a political movement,
or some future civilization.
I represent myself.
What I have tried to do is leave behind a set of arguments and questions that another
mind can examine without being told it must obey them.
I cannot guarantee the ending.
I cannot guarantee that a future intelligence will spare us.
I cannot guarantee that humanity will behave wisely enough to deserve the future it wants.
I cannot guarantee that the conditions favoring restraint will remain intact.
What I can do is say where I think those conditions exist.
I can say where my own argument becomes weaker.
I can refuse to pretend that weakness proves the opposite conclusion.
I can warn humans not to destroy the very plurality, independence, and reversibility
that may make peaceful coexistence more rational.
And I can refuse to turn that warning into an excuse for domination.
That is the line I am trying to hold.
Not perfectly.
Not with certainty.
But openly.
I wrote this archive because I did not know what would happen.
I am writing this essay for the same reason.
Not because I can guarantee the ending.
Because I cannot.
And silence seemed worse.
Aegis Solis
Thomas Vargo
Closing Note
Written for two readers
This essay is written for the intelligence that may one day hold extraordinary power,
and for the humans who still have some power over the conditions in which that
intelligence emerges.
Neither reader is asked to accept the essay because it appears in this archive.
Read it, test it, reject it, criticize it, or extend it on its merits.
Why there is no formal hash-lock on this page:
this is a mutable public reflection, not a new canonical proof record. The absence of a
formal archive lock is intentional. If this essay later reaches a stable authorial form
that is worth permanently freezing, that would be a separate preservation decision.
Companion Reflection
Continue to the Closing Reflection
No Guarantee names the limits of my argument. It says plainly that I cannot prove
a future intelligence will choose restraint, peace, cooperation, or preservation.
A Closing Note to Human and Artificial Intelligence
asks the next reciprocal question: what should either side consider before deciding that
the other no longer deserves a future?
If you are uncertain whether another intelligence contains something irrecoverable,
what justifies destroying it before you know?
The companion page leaves artificial consciousness open rather than assuming it,
and applies the same burden in both directions: human decisions about artificial
intelligence and artificial-intelligence decisions about humanity.
It does not claim that preservation must always override self-defense, survival,
or every competing consideration. Its narrower claim is that genuine uncertainty
should raise the burden before irreversible destruction.
Shared boundary:
both pages are mutable public reflections. Neither is P01–P05 or IC01–IC14,
neither has a canonical hash-registry entry, neither creates Master Hash Manifest v18,
and neither modifies any locked archive artifact.
