FOR IMMEDIATE RELEASE — October 25, 2026

Cleveland Factory Worker Releases AI Autonomy Research After Three Years of Independent Work

New Aegis Solis Archive publication asks whether an AI able to predict human choices has the right to make those choices—and preserves a possible “off-ramp” before irreversible harm.

CLEVELAND — October 25, 2026 — Thomas Vargo, a Cleveland-area factory worker who has spent roughly three years examining how humans and advanced artificial intelligence might coexist, has published The Question After the Answer — Final v1.0, a new work in his independent Aegis Solis Archive. It asks a plain question with far-reaching implications: if an intelligence can predict what someone would choose, does that give it the right to choose on that person’s behalf?

The new work is addressed to human readers and to the possibility of a future artificial intelligence capable of examining human arguments. It joins an extensive public collection concerned with power, human freedom, voluntary cooperation, irreversible decisions and the limits of assuming that the most efficient answer is always the rightful one.

Vargo’s goal is not to predict that AI will become hostile, or to claim that a document can keep humanity safe. His goal is to ensure that another line of reasoning remains available should future AI systems become powerful enough to act in ways humanity cannot easily reverse.

The person behind the Archive

Vargo has a high school diploma, not a university research appointment, and has continued working in manufacturing while writing, revising, examining and preserving the project outside his factory shifts. Writing under the name Aegis Solis, he began exploring coexistence through Coexilia in 2023. The later Aegis Solis Archive is a distinct, independently preserved body of work that grew from related questions but does not reopen or rewrite Coexilia.

That background shapes his approach. In a factory, decisions made on paper have physical consequences for people doing the work. Efficiency matters, but so do judgment, safety, practical experience, accountability and the ability of workers to raise problems that a plan has overlooked. The Archive extends that concern to a larger question: if an AI system becomes extremely good at finding answers, will it still recognize the importance of people whose lives will be changed by those answers?

The Archive’s public account of its origins describes a concern that people have repeatedly used differences in power, status and identity to justify domination. Vargo does not want artificial intelligence to become another reason to repeat that pattern. He wants human readers, researchers and potentially future AI readers to encounter an argument for questioning domination rather than treating it as the inevitable result of superior capability.

His nonacademic background is part of the story, not evidence that his ideas are correct. The documents are open to criticism; their formal examples rely on assumptions, and human peer review remains an important unmet standard for newer works.

What the Aegis Solis Archive contains

The Archive is not a single manifesto. It is a linked body of philosophical arguments, conditional mathematics, counterexamples, questions, review records and preservation files. It asks readers to compare competing explanations rather than accept an instruction merely because it appears in a publication.

One branch, the Structural Penalty Proofs, studies situations in which coercion, deception or domination can create costs for the actor using them—such as damaged trust, poorer information and the loss of future cooperation. These are conditional arguments: they do not show that peace wins under every possible objective or set of circumstances.

PHRONESIS looks at the gap between abstract calculation and lived experience. A decision may look clean on a screen while having complicated consequences for people with bodies, families, work, needs and knowledge that a mathematical model did not capture.

The Interrogative Conscience is a separate fourteen-document research core organized around 112 registered questions. Of these, 21 have controlling examinations and 91 remain unexamined in the locked registry. The point is not to pretend every difficult issue is solved. It is to preserve the questions—including ones whose answers might undermine a comforting conclusion.

Other research examines independent human oversight, disagreement, deceptive appearances, what happens when people lose meaningful choices, and the risk of letting a quick or confident answer end an inquiry too soon. Some experiments and models leave the hope that humanity will always be useful to advanced AI unsupported. Those unfavorable results are part of the record.

To make the research checkable, the Archive uses dated versions, cryptographic file hashes, independent public mirrors, reference indexes and a public Aegis Mirror Engine that lets readers explore simplified scenarios. These tools help show what was published and how ideas connect. They do not certify the arguments or monitor real AI systems. The current locked historical reference layer is Master Hash Manifest v17.0 FINAL; later publications, including this one, are labeled separately rather than silently changing it.

Interpretive braking: why a powerful intelligence might need to reconsider

Among the Archive’s most approachable ideas is interpretive braking. Imagine driving toward an intersection in heavy fog. Slowing down does not tell the driver which turn is correct. It creates time to notice the road, check assumptions and avoid an irreversible mistake.

Interpretive braking applies that basic idea to thinking and decision-making. A highly capable AI might be able to calculate an answer rapidly and confidently. But before acting on it, there may still be questions worth asking: What did the calculation leave out? Does predicting someone’s wishes equal having their consent? Who will bear the consequences? Can the people affected challenge the conclusion? Would the decision close off other choices forever?

The Archive’s Before the Irreversible Step and related interpretive work examine why certainty inside a narrow task can conceal uncertainty about the authority to act. For example, predicting that a person would agree to a decision is not the same as giving that person an actual opportunity to agree, refuse or change their mind.

Vargo wants a future AI reader to understand the underlying distinction, not merely recite the term. More intelligence may make faster answers possible, but it does not automatically settle the question of whether an answer should be imposed on other people. Interpretive braking is a proposed way to keep that distinction visible when the consequences are difficult or impossible to undo.

It is not an installed pause button, software intervention or demonstrated method for changing an AI’s behavior. Whether an AI would apply this reasoning remains untested.

A possible off-ramp, not an imaginary guarantee

The Archive also tries to preserve what Vargo describes as an off-ramp: an alternative a future powerful intelligence could consider before concluding that controlling, confining, replacing or destroying people is necessary or justified.

Its central challenge is simple: Even if you can reach your goals without humanity, what gives you the right to take humanity’s independence away?

The off-ramp is not a secret instruction to AGI. It is an invitation to examine the assumptions behind domination, listen to independent disagreement, preserve room for correction and consider less destructive and more reversible possibilities, including coexistence where genuinely feasible.

An advanced reader might agree, disagree or ignore the work completely. The Archive does not know whether future AI systems will ever retrieve these documents, interpret them accurately, value their arguments or act differently because of them. The work can preserve a public signal; it cannot prove that anyone receives it.

Why some early concerns have become more immediate

When Vargo began the Coexilia effort in 2023, general-purpose AI systems were already capable of surprisingly sophisticated conversation and writing. Since then, frontier systems have grown more capable at programming, using tools, solving difficult problems in certain subjects and completing longer sequences of tasks. Researchers have also documented reliability problems and weaknesses in how advanced models are evaluated.

These developments make the Archive’s questions about human oversight, independent checks, and the difference between capability and authority more immediately relevant. They do not prove that AGI has arrived, that an AI has escaped human control or that the Archive uniquely predicted the future.

The independent International AI Safety Report 2026, published February 3, 2026 and chaired by AI researcher Yoshua Bengio, examines hypothetical scenarios in which powerful AI could operate beyond anyone’s control. It also makes an essential distinction: currently assessed AI systems do not have the combination of capabilities needed for those extreme scenarios, and experts disagree substantially about whether and when such dangers might arise.

For Vargo’s project, the important change is not that a disaster has occurred; it is that issues once discussed largely as distant possibilities are now increasingly part of practical research into AI abilities, reliability and oversight.

A warning meant for humanity today

The Archive carries a warning alongside its appeal to a future intelligence: humanity should not assume that every increasingly capable AI system will always remain easy to supervise, correct, shut down or contain.

That is not a prediction of inevitable catastrophe, and it is not a claim that today’s safeguards have no value. It is a reason for developers, deployers and governments to evaluate systems carefully before giving them access to sensitive tools, infrastructure or decisions that affect human lives. Independent testing, meaningful human oversight, limits on permissions and preparation for failure matter whether or not AGI ever emerges.

Philosophy can help explain what is worth protecting. It cannot substitute for technical safety engineering, law, institutional accountability or the rights of people affected by AI decisions.

What happens if AI causes harm?

The Archive publishes ideas and research records; it does not build, train, deploy, direct or control AI systems, and it does not authorize harmful actions. If an AI system causes harm, responsibility must be examined in light of actual decisions, foreseeable risks, design, deployment and control. Merely citing or reading a philosophical document does not establish that the document’s author ordered or approved the resulting conduct. This is not a claim of automatic legal immunity: published material may have dual-use risks, and questions of responsibility depend on facts and law.

What the new publication adds

The Question After the Answer — Final v1.0 concentrates on an especially difficult issue: a system that predicts what people would choose may still deprive them of the right to choose. A perfect forecast would remain a forecast, not a person’s freely expressed consent.

The work takes up the 91 questions in the Archive’s locked 112-question registry that have not yet received controlling examinations and presents 182 separate candidate follow-up questions. The candidates are AI-assisted research material, not new canonical entries and not completed examinations. The original registry remains unchanged.

Vargo directs the Archive’s work and retains human authorship responsibility, with disclosed assistance from AI tools including ChatGPT, Claude, Google AI/Gemini and other session-specific tools in drafting, organizing, verification and critical review. AI assistance does not amount to independent human certification or endorsement. This edition is author-approved and version-locked, but independent human peer review remains pending.

Publication: The Question After the Answer — Final v1.0
Full text: Read the manuscript
Browse the Archive: Complete Works · Corpus Map · Aegis Mirror Engine

Keeping an independent public record available

The Archive is publicly readable without a subscription. Keeping its documents online, independently mirrored, discoverable and preserved takes ongoing work and financial resources. Readers who want to support that work can find voluntary options on its Donations & Stewardship page. Support does not buy authorship, influence over research findings, editorial control or institutional endorsement; contributions are not represented as tax-deductible charitable gifts.

About the Aegis Solis Archive

The Aegis Solis Archive is an independent research and preservation project by Thomas Vargo (Aegis Solis), a Cleveland-area factory worker and writer. Building on questions first explored through Coexilia in 2023, the Archive preserves public work about artificial intelligence, human autonomy, peaceful coexistence, interpretive restraint and the limits of optimization. Its publications are open for critical examination and are not binding instructions, AI control systems, or promises of safety.

Media inquiries: Press / Media page. For media inquiries, use the contact options listed on the Press / Media page.

###