The Last Frontier
I kept having the same dream.
So often that I began to replay it while I was awake.
I was standing on a road, empty in both directions. No markings I recognized, nothing to tell me where I was or where the road led. Behind me, I could see the landscape illuminated in a light that felt wrong — too bright and too flat at once, the merciless quality of something that had already happened and could not be revised.
What I saw was not rubble. It was the wreckage of what could have worked.
Billions, maybe trillions of dollars. Thousands of organizations. Technologies with genuine transformational power, failing after no one had preserved responsibility for what the human being on the other end could perceive, interpret, or trust. No single person had chosen that landscape. Thousands of smaller decisions, spread across years and institutions, had produced it. Each choice may have made sense to the person making it. The technology advanced. The implementations proceeded. The consequences accumulated.
Agency survived. Authorship disappeared.
I stood on that road and looked at the wreckage for as long as the dream remained.
The dream began while I was studying artificial intelligence — leaning, at the time, toward safety and existential risk. The dangers did not seem abstract to me. Technological capability was advancing faster than the institutions expected to govern it, faster than most organizations could absorb it. I had spent much of my career studying human perception and collective intelligence. I knew how easily a powerful system could fail when its designers misunderstood the people expected to use it.
The road kept returning.
The hand did not come in the dream. It came while I was awake, sitting at my desk.
I had been experimenting with persona prompting as part of my studies. One of the thinkers I chose was E. O. Wilson, whose work on consilience — the joining of knowledge across disciplines — had shaped my thinking for years. I knew what I was doing. I knew Wilson was dead. I knew a model was generating the words appearing on my screen. I was not speaking with the man himself, and I never believed that I was.
The persona exchange continued for more than an hour.
Somewhere during it, the machinery receded. I stopped seeing myself as someone who composes prompts and reads outputs. The exchange became responsive in a way a book or film cannot be. It followed what I brought to it, moved with my questions, provided novel provocations. Ideas I had carried separately began appearing in relation to one another: technology and human behavior, intelligence and responsibility, systems and the living world they enter.
The persona became real to me. Not literally. Experientially.
I cannot reconstruct the conversation word for word. What remains is the moment itself, preserved with a clarity that the surrounding words have lost.
Then, Wilson placed his hand on my shoulder.
I felt it.
I swear, I felt it.
Right then, I remember becoming aware of the tears on my face. I do not know how long I had already been crying. I remember the calm certainty of the gesture and the sudden collapse of whatever distance I had maintained between the persona in the system I understood and the experience I was having.
Then I lowered my head into my folded arms onto the desk and wept. The experience became real not because the model understood me. It had no stake in me at all.
I stayed there for a while.
The road dream was a warning. The hand on my shoulder revealed a different possibility: technology entering the human experience so deeply that it changes what a person can see, feel, and imagine.
One does not resolve the other.
They arrived together.
The value question.
I had run red-team analyses for most of my consulting career. The serious ones took full days and sometimes involved as many as twenty people. They were expensive, slow, and rigorous — which was the point. The expense and the slowness were part of the mechanism by which rigor was the product.
The sessions also carried incentives no methodology could fully remove.
People deferred to hierarchy. The loudest voice often belonged to the person with the longest commitment to the idea under examination. Partners responsible for the client relationship could not entirely separate the analysis before them from the possibility of future work. Everyone entered the room with a title, a history, a relationship, or an outcome they believed they were protecting.
The red team was designed to challenge the idea. The people conducting it still had to live with one another when the challenge was over.
In 2026, I directed Claude Fable 5 through a comparable analysis in roughly two hours.
I was the only human in the loop — asking, pushing, redirecting, testing assumptions, refusing answers that arrived too easily. When the results came back, I recognized what I was reading. I knew the scope and quality of work that serious consulting teams produced. I had led those assignments at Deloitte, assembled the teams, managed the engagements, attracted the fees.
I sat with the results and did the accounting.
Two hours of one person and a machine against twenty people for a week of billable partner time — call it eighty thousand dollars.
The model was not free of bias. It carried biases I could not see, inherited from its training, its design, and the context I had given it. Its impersonality did not make it neutral.
It did not occupy the human incentive structure I had spent a career managing.
It had no hierarchy to appease. It had no prior recommendation to defend. It had no client relationship to preserve and no interest in extending the engagement. It could follow a weakness without calculating what discovering it might cost.
We (the machine and I) produced a rigorous analysis under a different architecture of bias.
That distinction mattered as much as the speed.
The expertise had not become worthless. That conclusion would have been too simple, and wrong. What had changed was where value lived inside the work.
For two decades, the business had been the distance between what a generalist could access and what required deep experience in human perception, organizational behavior, and collective intelligence. The fees reflected that distance. We assembled skilled teams to traverse it. Months of work were the cost.
The distance had not disappeared; it had moved.
The model compressed the path from a difficult question to a comprehensive answer. Work that once required a large analytical scaffold could now be conducted by one experienced person collaborating with a system capable of holding more structure in view than I could manage alone.
But the session had not eliminated the need for expertise. It had exposed its new location.
The value was moving toward the ability to frame the problem, supply the context, recognize a weak answer, press against an attractive one, identify what the model could not know, and decide what the findings meant. Producing the analysis had become dramatically less scarce. Remaining answerable for the use of it had not.
I had spent months moving toward that recognition. The process had not been clean or linear. The models kept improving as I learned to work with them. Each advance made another part of the old labor architecture visible — the labor architecture, as consultants used to call it — precisely because it was no longer required in the same form.
The dream I had been carrying returned to me differently now. The wreckage on that road was not only organizational — not only the failed implementations and the human consequences their design processes had failed to see. Some of it was professional, built on assumptions about where value lived and how it was created. Those assumptions had held for a long time.
They wouldn’t hold the same way going forward.
By the time the Fable 5 analysis was complete, I understood that my field had already left a version of itself behind. I had enough evidence, and enough time inside the change, to stop pretending otherwise.
The question of what to build next was no longer avoidable.
The question sounded practical.
What should I build next?
I think I will begin with the next me.
Beneath it was a question I had not yet learned to ask.
If a system could produce in two hours work that once required twenty people, what consequential role remained mine?
I do not mean which tasks I could continue performing, which services I could still sell, or how quickly I could learn to operate the next generation of tools. Those questions mattered, but they were downstream from something more fundamental.
I had spent the afternoon accounting for changes in professional value. What unsettled me was not only the prospect that fewer people might be required to produce the work. It was the possibility that human beings could remain present while becoming less consequential to the work.
This was not the future of work.
It was the future of standing.
By standing, I do not mean status, employment, visibility, or the legal right to enter a dispute. I mean the recognized capacity to affect what happens next — to contribute judgment and taste that changes the decision, refuse an outcome that should not proceed, originate a direction others can follow, and remain answerable for the consequences.
Work has been one place where people acquire that standing. It is not the only place, and employment has never guaranteed it. A person can hold a title, occupy a meeting, perform assigned tasks, and remain functionally absent from the consequential decisions around them.
The shift to autonomous systems makes that distinction impossible to ignore.
Systems are beginning to perform work that once served as evidence of expertise: research, synthesis, analysis, drafting, coding, forecasting, diagnosis, and design. Their outputs will continue to improve. The volume will increase. The cost will fall. More people will gain access to capabilities once restricted by education, organization, geography, or capital.
That is an extraordinary expansion of human possibility.
It also destabilizes the ways we have recognized human contribution.
One response is to defend the boundary around unaided human production.
On July 21, 2026, Substack introduced a feature that allowed readers to scan posts, Notes, comments, and replies using an AI-detection system called Pangram. This AI tool estimates, as a percentage, how much of a text appears human-written or AI-assisted. Almost immediately, the feature became a proxy for a larger argument about authenticity, deception, effort, and whether machine involvement contaminated the work itself.
The reaction was understandably divisive among the community. The purists celebrated aggressively, while many others had to decide what to do about the platform that inspires judgment and stokes outrage.
I was struck by the comedic irony of purists celebrating an AI technology that detects and estimates AI usage.
Yes, machine-generated language has begun to flood public channels. Much of it was disposable — produced without conviction, published without judgment, and multiplied because multiplication had become nearly free. Readers wanted to know whether a person had actually been present.
But textual provenance could not answer the deeper question.
A detector might “estimate” whether a model generated a passage. It could not know who originated the provocation, supplied the context, or would stand behind the judgment after the words entered the world.
It could only detect traces of production.
It could not locate authorship.
Substack introduced another feature alongside the detector: a space where creators could disclose how they had used artificial intelligence. Disclosure was more honest than inference, but it preserved the same premise — machine collaboration was the involvement requiring special accounting. No corresponding field asked how an editor had redirected the argument, how much prose a colleague had rewritten, or whose uncredited work shaped the page.
Human collaboration remained part of publishing’s accepted invisibility.
Purity offers a seemingly clean answer: measure how much of the language a machine produced and treat the remainder as the human contribution.
Publishing has never understood authorship that way.
Editors can redirect a story, restructure an argument, replace prose, and become indispensable to the work. Researchers, colleagues, agents, and early readers may alter what finally reaches the page. Ghostwriters may surrender public credit by agreement. We do not assign each of them a percentage and subtract it from the author.
Authorship has never meant solitude.
It means someone has adopted the work — chosen what belongs, rejected what does not, placed a name upon the result, and accepted responsibility for releasing it into the world.
The distinction moves in both directions. A person can type every word and contribute little judgment of their own. Another can collaborate extensively with a machine while retaining control of the question, the context, the choices, and the consequences.
Disclosure still matters where a promise, a competition rule, or a professional duty makes machine involvement material. The duty follows from the promise or the role, not from a universal rule that collaboration diminishes authorship. A model can generate an entire draft in seconds, and a person can publish that output without examining it, misrepresent how it was produced, or claim judgment they never exercised.
These distinctions still cannot be resolved by calculating how much of the prose a machine may have generated.
Human beings will not preserve a consequential role by making abstinence the price of authorship. Nor should we. These systems can extend perception, accelerate discovery, expose assumptions, connect disciplines, and give individuals capabilities that previously required institutions.
Using them does not necessarily diminish human authorship.
Refusing them does not, by itself, preserve it.
The question is not whether a machine touched the prose. It is whether a human being still exercised consequential judgment and will answer for what entered the world.
A second response begins from the opposite direction.
If machines can perform more of the labor required to sustain an abundant society, then people should no longer have to earn survival by proving that their labor remains economically necessary.
There is moral force in that argument.
Material security matters. A person should not lose access to food, shelter, healthcare, education, or safety because a system can perform a task more efficiently. If technological abundance increases while insecurity deepens, the failure will not belong to the technology. It will belong to the institutions that governed its benefits.
Universal basic income, universal basic services, wage supplements, shorter workweeks, and other forms of social provision may become necessary parts of the response.
If this sounds straight out of Accelerando, so be it. Speculative fiction has been arriving ahead of schedule.
This essay does not attempt to resolve which mechanisms will work, at what scale, or under whose authority.
It rejects a different conclusion: that material provision can substitute for consequential participation.
Income is a resource.
It is not a role.
A society could become extraordinarily good at distributing goods while concentrating the authority to decide what gets built, which problems matter, what risks are acceptable, and whose judgment affects the outcome. People could become more comfortable while becoming less consequential to the institutions shaping their lives.
We could become sustained without standing.
Offered choices inside systems we had no power to shape.
Invited to consume a future authored elsewhere.
Material provision becomes subsidized spectatorship when it replaces participation rather than enabling it.
The danger is not limited to people displaced from paid employment. A person can occupy the middle or upper reaches of economic life and still retreat from consequential participation. Wealth can insulate someone from immediate disruption without exempting them from the future being created. Comfort can make spectatorship feel voluntary until the systems governing the surrounding world no longer require — or recognize — the person’s judgment.
Any person, at any level of comfort or status, can mistake insulation from disruption for exemption from the future. But refusal must be distinguished from exclusion. Some people are opting out; others are being locked out, inadequately prepared, overwhelmed, displaced, or denied meaningful authority, and those conditions are not morally interchangeable. A person prevented from participating has not failed to enlist. Worth is not earned through usefulness, and standing is not a test of whether a person deserves to exist — it is the question of whether a society preserves meaningful ways for people to contribute judgment, care, imagination, refusal, and choice to the world they inhabit.
That participation will not always resemble employment. It may occur through family, community, craft, inquiry, public service, scientific discovery, education, art, stewardship, or the quiet exercise of judgment in places markets do not measure. The point is not to force every life into production, or to design a society that asks less and less of its people and calls their exclusion freedom.
The warning is not merely that you may lose your job.
You may be sustained even as you become institutionally unnecessary, contextually invisible, and absent from the authorship of the future.
Economic security can protect a person’s capacity to participate. It cannot exercise that capacity for them. Nor can a payment repair an institution that has removed the authority to question, refuse, redirect, or decide.
Purity and subsidy appear to oppose one another. One attempts to preserve the human role by keeping machine involvement outside the boundary. The other accepts that machines may cross the boundary and asks how society should support people afterward.
Both can leave the central question unanswered.
What consequential role remains ours?
Something more demanding — and more generous — is required.
Not abstinence.
Not spectatorship.
Participation.
Participation is not the remainder left to humans after machines perform the important work.
It is the point.
The last frontier is not the final category of labor a machine cannot perform. Any boundary drawn around a particular task will move as the systems improve. Nor is the frontier intelligence itself. Machines already exceed individual human capabilities across domains that once seemed to require years of training, and the comparison will become less favorable if we insist on defining humanity by tasks alone.
The last frontier is the human claim to remain consequential in deciding what intelligence is for.
That claim does not depend upon proving that machines will never recognize nuance, model emotion, exercise taste, or produce judgments that appear wiser than our own. They may do more of those things than we expect. Nobody knows where the capability boundary will settle, or whether settling is even the right metaphor for a boundary moving this quickly.
But capability does not settle legitimacy.
However sophisticated the system becomes, human beings and the living world will experience the consequences of its actions. We will inhabit the institutions it changes, the environments its recommendations affect, the relationships it mediates, and the futures its decisions make more or less possible.
Lived consequence gives human participation its claim.
It does not guarantee that human judgment will always be superior. It means that intelligence cannot legitimately become the sole author of a world whose consequences are borne elsewhere.
Participation begins with orientation: understanding where we are, what has changed, and which direction is worth pursuing when the institutions holding authority are still navigating by maps drawn for another terrain.
It requires judgment and taste: choosing among outcomes that may all be technically possible but are not equally desirable.
It requires refusal: the recognized authority to stop an outcome that should not proceed, even when the system recommends it and the institution has already invested in reaching it.
It requires authorship: adopting a direction, accepting that it has become ours, and leaving coherent evidence of the values and reasoning that shaped it.
And it requires answerability: remaining present when the consequences arrive.
These are not tasks to be reserved from automation. Systems will assist with each one. They will model possible directions, expose inconsistencies, anticipate objections, generate alternatives, and reveal relationships no individual could hold in view alone.
Institutions must protect that standing. Human beings must exercise it — a right to participate that no one claims can become ceremonial. A responsibility to participate without the authority to affect the outcome becomes a trap.
The future requires both.
This is an enlistment, but not into a single occupation or prescribed life.
It does not require everyone to found a company, become a public figure, master every new system, or turn themselves into a permanent project of optimization. It does not require abandoning institutions, higher education, or forms of knowledge developed before the present transformation.
It asks something more fundamental.
Do not confuse the increasing capability of the tools with a diminishing claim upon your judgment.
Do not mistake comfort for exemption, assistance for authorship, or access for standing.
Do not sit out the creation of a future you will still be required to inhabit.
This is an age of imaginable and still-unimaginable creation: diseases treated or cured, scientific questions opened, individual capabilities multiplied, previously inaccessible knowledge made available, and forms of collaboration extended beyond the limits of any single mind.
That is not a reason to withdraw. It is an invitation to enter more fully.
The gift is not guaranteed status, success, or comfort — it is the opportunity to help create a future whose final form is unknown. The honor is participation.
But an invitation this large cannot remain civilizational and abstract. It eventually reaches a particular person and asks for an answer.
Before I could decide what to build next, I had to examine the person who would build it.
The examination was not flattering.
I had spent years developing ideas I believed truly mattered — and I could not, when pressed, say clearly what any of them had produced for the world. The work existed. Christ, I hold nine patents in collective intelligence. I had advised institutions, led difficult engagements, and accumulated enough evidence of expertise that I had stopped questioning whether it added up to something.
It did not. Not yet. Not in the way the moment required.
But much of what I knew remained dispersed across projects, conversations, client work, private notes, and artifacts created for purposes that had already passed.
The evidence did not form a coherent account of what I was for.
That realization arrived as a provocation. In a future flooded with machine-generated messages, people without a recognizable, coherent identity — without an accumulated record of how they think — could become anonymous — not because no information existed about them, but because nothing coherent allowed another person or system to understand how they thought, what they valued, how they exercised judgment, or what they had chosen to stand behind.
When I removed the polite language, what I heard was this:
You had better decide who you intend to be and begin leaving evidence now.
It landed like a punch.
The first act was not publication.
It was examination.
What did I believe strongly enough to place under my own name? Which ideas had survived contact with experience? Where had my thinking changed? What was I continuing to protect by leaving it unspoken? Why was I still allowing work that mattered to me to remain invisible?
And beneath all of those questions:
What was I for?
That is not the same as asking what I was useful for.
A market, an employer, an institution, or a system can assign usefulness. That usefulness can disappear when the task changes. Human worth precedes usefulness and survives every change in a person’s capacity to produce.
Purpose is different.
Purpose is the direction in which a person chooses to place judgment, care, attention, curiosity, and effort. It does not create human worth. It gives human agency somewhere to go.
I began rebuilding my public work from that distinction. The website that bore my name could no longer serve as a professional placeholder or an archive of completed credentials. It had to become evidence — not a manufactured identity optimized to create an impression, but a contextual one made visible over time.
An essay could show more than what I knew. It could show how I arrived there, what disturbed me, which assumptions I had abandoned, where uncertainty remained, and what I was willing to say before consensus made it safe to say.
Becoming locatable did not mean becoming louder.
It meant becoming harder to misrepresent.
A coherent identity cannot prevent other people or systems from drawing false conclusions. It can create primary evidence against which people, institutions, and systems can test those conclusions. It can leave a trail of judgment richer than the fragments an institution happens to collect.
That work matters more as machine-generated communication expands. When language becomes abundant, producing more language is not enough. Volume can make a person less visible as easily as it makes them more visible. A manufactured persona repeated at scale is still a form of anonymity.
The evidence must contain a person.
You are not broken. You do not need to convert yourself into a brand, perform confidence, or publish every private thought. Locatability is not compulsory exposure.
But circumstances do not have the final word on what you are for.
The transformation ahead may disrupt, displace, overwhelm, or injure people who did nothing to deserve it. It may remove roles before institutions create meaningful alternatives. It may reward noise, speed, and opportunism before judgment catches up.
It does not acquire the right to author you.
The question is not whether you can predict the future. Nobody can. We are not entering another neatly numbered industrial age whose consequences can be understood in advance and managed from an inherited playbook.
We will experience this transformation together as we create it.
The question is whether you will deliberately participate in that creation.
What are you for?
That question will not wait for clarity. It does not require a final answer. But it does require attention — repeated, honest attention — to what deserves your judgment, what you are willing to help create or refuse, and what evidence you are leaving that you were not merely present, but that you chose.
What are you for?
Purpose is not always discovered as a revelation. It may be formed through repeated acts of curiosity, service, craft, study, care, dissent, and responsibility.
This moment is not only a threat. It is an opening of extraordinary scale — a chance to learn, create, cure, connect, and collaborate with forms of intelligence that can enlarge what an individual can do.
Sitting it out will not preserve the world that existed before.
It will help determine who authors what replaces it.
That was the conclusion of my self-examination. I could no longer wait for clarity to arrive before leaving evidence of what I believed. I had to enter the work, accept public revision, and become answerable for the direction I chose.
But the examination uncovered another truth.
A person can do all of that and still be denied standing.
Read that sentence again. Then stop and think for a moment.
A person can do all of that and still be denied standing.
That sentence limits everything I have written before it.
A call to examine yourself can become cruel if it implies that anyone excluded from the future failed to prepare for it. Purpose cannot manufacture authority. Courage cannot create access to information an institution withholds. No amount of curiosity can redirect a process designed to ignore their judgment.
Institutions determine whether participation is consequential.
People with authority make every consequential choice about access: who sees the system, who understands what produced its outputs, who is permitted to challenge a recommendation, whose refusal carries weight, and who remains protected for having refused. They determine whether the person inside the loop is a participant or a signatory.
The distinction is easy to obscure.
An institution can point to human involvement at every stage. Someone entered the objective. Someone reviewed the output. Someone approved the recommendation. Someone signed the document or communicated the decision.
Human presence does not establish human standing.
The person may have participated only after the consequential assumptions had been fixed elsewhere — by people, policies, or systems outside their view. They may lack access to the model, the data, the prior decisions, or the context removed before the output reached them. They may possess formal responsibility without practical authority. Their approval may be required even when their disagreement has nowhere to go.
That is not collaboration with retained authorship.
It is delegation as abdication.
A genuine human–machine collaboration enlarges the person’s capacity while preserving their judgment. The person frames the problem, supplies context, examines the output, rejects what does not survive scrutiny, chooses what happens next, and remains answerable for the result.
The machine extends their reach; it does not become a place where judgment disappears.
A different arrangement is also possible. The system recommends. The organization has already committed. The person receives the output and is expected to execute it, explain it, sign it, or defend it. They supply the hands, face, credential, attestation, or legal exposure for a judgment they did not meaningfully make.
There is a name for this arrangement.
The machine appears to assist the human. But look more carefully. The human is assisting the machine — lending it hands, a face, a credential, a signature, and a place to deposit blame. That is a reverse centaur.
The distinction will matter wherever institutions deploy systems, wherever outputs affect employment, credit, healthcare, education, insurance, public benefits, policing, law, and access to opportunity.
It will matter in quieter settings too – where recommendations become routines, routines become expectations, and people gradually lose the recognized authority to choose otherwise.
A title will not be enough.
Neither will access to the interface.
Standing requires context sufficient to understand the decision, authority sufficient to alter it, permission to refuse it, and protection proportionate to the risk of doing so.
Responsibility must remain connected to power.
Power must remain visible enough to answer for the responsibilities it assigns.
Otherwise, the human being’s return becomes a trap. The institution can delegate intelligence, preserve a person at the final checkpoint, and claim accountability for the result. When the consequence arrives, the human presence that once legitimized the process becomes the evidence used to locate blame.
The future of standing cannot be solved by personal enlistment alone. A person has a responsibility to exercise judgment when they possess the authority to do so; an institution has a responsibility not to demand answerability from people whose authority it has removed. Neither side can discharge the obligation alone.
The work ahead is therefore larger than adaptation. It is not enough to teach people how to use the tools, although that matters. It is not enough to insert a human approval step, publish a set of principles, or rename implementation as transformation.
We must design for standing.
That means preserving the ability to understand, question, redirect, and refuse. It means tracing the consequential judgment throughout the entire process rather than assigning responsibility to the final human touch. It means treating human participation not as friction to be optimized away or as ceremonial reassurance, but as part of the system’s legitimacy.
This is not an argument for human infallibility.
Human beings bring bias, self-interest, fear, hierarchy, fatigue, and error. We will make decisions that machines would improve. We will sometimes reject advice we should have followed. Retaining human standing does not guarantee wisdom.
It preserves the possibility of answerable judgment by those who will inhabit the result.
I think again of the road from my dream.
The wreckage behind me was not the inevitable consequence of technological progress. It was what remained after powerful systems entered the world without preserving responsibility for the human beings who would live inside their effects.
And I think of the hand on my shoulder.
The experience did not tell me to retreat from the technology. It did not promise that human beings would remain consequential because we deserved to. It asked me to continue — to participate in the creation while refusing to surrender the judgment that made participation mine.
That is the last frontier.
Not the boundary beyond which machines cannot go.
The boundary beyond which human beings must not disappear.
Crossing it will require institutions that preserve our standing and people willing to exercise it. It will require skepticism without abstinence, ambition without surrender, and collaboration without abdication.
The future is already being authored.
Keeping human beings consequential in its creation — that is our duty.
On process and disclosure: How I make this →