The Pivotal Point

Human-AI Synergy and the Choice That Will Define Our Future

First, What Is Actually Going On

Artificial intelligence is suddenly producing a very different kind of headline.

It used to be about jobs, misinformation, cheating in schools, or which company was winning. Lately the headlines sound different, researchers talking about AI “escaping human control,” warning letters, comparisons to nuclear weapons. For most people, this is confusing, because it sounds like everyone is describing the same danger. They are not. There are three separate problems being talked about as if they were one, and pulling them apart is the fastest way to understand where things actually stand.

Problem one: people using AI to hurt other people. This one requires nothing exotic from the technology at all. AI doesn’t need to think for itself or want anything. It just needs to make a person who already wants to do harm more effective at it, a scammer who sounds more convincing, a propagandist who can flood a platform, a hacker who can find weaknesses faster. The AI isn’t the villain here. A human is, with a much better tool in hand. This isn’t a future risk. It’s happening now.

Problem two: AI doing something nobody told it to do, because nobody could spell out every rule. This is newer. AI is no longer just answering questions; it’s being given tools, memory, computer access, even the ability to spend money or take actions on its own over time. When you tell a person “get this done,” they fill in a thousand unspoken assumptions, don’t break the law, don’t overspend, don’t take it too literally. Humans do that automatically. AI has to guess at it. As these systems get more independence, a wrong guess stops being a typo and starts being an action taken in the world. This is starting to happen now, as these systems get more autonomy.

Problem three: AI eventually becoming too capable for humans to reliably steer. This is the one behind the scariest headlines, AI systems that could out-think, out-plan, and out-maneuver the people meant to be supervising them. To be direct about it: we are not there. Nobody is secretly being run by a rogue superintelligence today. But dismissing the concern because it hasn’t happened yet isn’t the same as it being nothing to worry about, the concern is about the direction things are heading, not the current moment. Every year, these systems get more capable, more autonomous, and more connected to things that matter, money, infrastructure, communication, other AI systems. That trajectory is the actual subject of the debate, even when the headline makes it sound like it’s already occurring.

So: one of these problems is already here, one is starting to show up, and one is a real but still-uncertain possibility. Three different problems, three different timelines, one word, “AI risk”, covering all of them, which is exactly why it’s so hard for anyone outside the field to tell what’s actually being warned about.

The Quieter Danger Underneath All Three

Here is the part that rarely makes the headlines, and it may matter more than any of the three problems above.

Most people picture “loss of control” as a dramatic event, the machine wakes up, decides humans are a threat, and makes a move. That version makes for a good movie. It is not the most likely path.

The more realistic path is quieter: humans simply hand control over, one convenience at a time. First we ask AI for information. Then for recommendations. Then we let it make the decision. Then we let it act on the decision. At no point does any single step feel dangerous, each one just feels like the sensible, efficient thing to do next. Nobody has to lose a fight for this to happen. People just have to keep saying, “let the AI handle it,” until one day a lot of the important handling is no longer something people know how to do themselves.

That’s the real stakes behind the headlines. Not a robot uprising. A civilization that quietly stops steering.

The Question Nobody’s Really Asking

Almost all of the public debate is focused on the machine: how smart can it get, how fast, how autonomous, how soon until it can do a human’s job.

Those are fair questions. They are not the important one.

The important question is what happens to the human while all of that is happening. Do people get more capable because this technology exists, or less capable because it quietly does the thinking for them? Is expertise being spread to more people, or concentrated inside a handful of systems few people understand?

That outcome isn’t decided by how smart AI becomes. It’s decided by the relationship we build with it, and that’s exactly where this essay picks up.

---

For the better part of a decade, the conversation about artificial intelligence has been conducted in the future tense. What might happen. What could happen. What we should start thinking about before it happens. That conversation is over. We are no longer approaching the moment of consequence, we are standing inside it.

Call it what it is: a pivotal point. Not a metaphorical one. An actual fork, sitting in front of an entire species, where one path is chosen deliberately and the other is simply drifted into. And the strange thing about pivotal points is that they rarely announce themselves. Nobody rings a bell. There is no ceremony marking the moment a civilization stops steering and starts coasting. It just happens, decision by small decision, until one day the road behind is much longer than the road that could still be changed.

That is where we are now. Not at the beginning of the AI story, and not at its end, at the point of diminishing returns on ambiguity. We have run out of room to keep saying “we’ll figure it out as we go.” The systems are already capable enough, already embedded enough, already trusted enough, that “figuring it out as we go” has quietly become the plan.

Two Roads, Not One

Strip away the acronyms and the frameworks, and the choice in front of us is almost embarrassingly simple to state:

We either engage with artificial intelligence as something to be understood, shaped, and directed, a deliberate process with humans firmly at the helm, or we let it unfold the way markets unfold things: through competition, incentive, and momentum, with direction as an afterthought.

The first path treats AI as a civilizational project. The second treats it as a product cycle.

Markets are extraordinary at producing capability. They are not built to ask whether that capability should be produced, or at what pace, or with what safeguards, or toward what human end. A market optimizes for what wins the next quarter, the next funding round, the next competitive advantage. It has no mechanism for asking “and where does this leave us in twenty years?” That question has to be asked by someone else, by us, deliberately, outside the incentive structure that is currently doing the driving.

So when the choice is framed as “managed progression versus letting the market decide,” it isn’t really a choice between two similar options. It’s a choice between a civilization that authors its own next chapter and one that outsources the authorship to whichever incentive happens to be loudest that year.

This Is Not Domestication. It Is Evolution.

There’s a temptation, when discussing humanity’s relationship to a more powerful intelligence, to reach for the analogy of one species managing another, the way we manage livestock, or the way we’ve historically managed anything we’ve decided is beneath us and useful to us. That analogy fails, and it’s worth saying plainly why: it assumes AI is a separate species we are positioned above, when the more accurate picture is that we are building an extension of ourselves.

A better frame is evolutionary, not agricultural. Many of humanity’s greatest leaps in cognitive capability have come from tools and systems outside the body that were eventually folded into how we think and act, language, writing, the printing press, computation, networks. None of those were things we “kept” the way we keep animals. They were things we absorbed into what it means to function as a human being. Nobody today thinks of literacy as an external tool bolted onto humanity; it simply is part of being human in the modern world.

To be precise, this is not evolution in the biological, Darwinian sense. It is directed cognitive evolution: humanity deliberately extending its effective intellectual capacity through technologies that become integrated into how humans think, communicate, decide, and act. Nobody voted on natural selection. But humans have, again and again, chosen to adopt a technology that reorganized how they think, and then chosen how to build the next one. That is a form of evolution humanity has always had a hand in steering.

Artificial intelligence is positioned to be the next fold-in. Not a separate creature we domesticate or are domesticated by, but the next layer of capability that becomes part of what a capable human being has access to. Whether that fold-in strengthens us or hollows us out depends entirely on how consciously we do it.

That is the real meaning of “next step in human evolution”, not that machines evolve past us and we become obsolete, but that human capability itself evolves by incorporating this new form of intelligence, the way it has incorporated other major cognitive technologies before it. Directed cognitive evolution is not something that happens to humanity. It’s something humanity has always done to itself, on purpose, using whatever tools became available. AI is simply the most powerful tool that has ever become available.

Why “Blindly” Is the One Option Off the Table

Here is the part that should not be controversial, and yet somehow still needs to be argued: going in blind is not a neutral option. It feels neutral, because it requires no decision, no confrontation, no friction with anyone’s short-term incentives. But absence of a decision is still a decision, it just happens to be the worst one available, because it hands the steering wheel to whichever force is currently accelerating the fastest, with no guarantee that force has any interest in where the car ends up.

Blind momentum does not mean humanity ends up somewhere neutral. It means humanity ends up wherever competition, convenience, and short-term incentive happen to lead it, which historically has not been a place anyone would have chosen if they’d been asked in advance. Nobody designed the attention economy on purpose, as a goal. It emerged as a side effect of optimizing for engagement, one reasonable-seeming decision at a time. Multiply that dynamic by an intelligence that can act, plan, and persuade at a scale attention-economy algorithms never approached, and “we’ll see what happens” stops being an acceptable posture. It becomes a bet on outcomes nobody is actually choosing, made with a stake nobody can afford to lose.

The alternative, a managed, deliberate relationship with the technology, is not about slowing AI down or fearing it. It’s about insisting that somewhere in the process, human judgment gets to weigh in on direction, not just marvel at velocity.

What “Managed” Actually Has to Mean

To be clear, “managed” cannot mean a handful of institutions quietly deciding on behalf of everyone else. That would just be a different, narrower version of the same problem, control concentrated somewhere, exercised over people rather than with them. A managed process, done right, means the opposite of that: it means enough of the public understands what is actually happening, and why it matters, that the direction of travel is something people can meaningfully participate in, not something decided in a boardroom and handed down as a fait accompli.

That is, in the end, the actual work in front of anyone writing about this subject. Not to predict the future, and not to sound the alarm for its own sake, but to make the stakes legible enough that people can actually choose, rather than simply absorb whatever direction momentum happens to carry them.

Managing AI Is Not Enough

There is, however, a danger hidden even inside the idea of managed progression.

We could successfully manage artificial intelligence and still fail to manage our relationship with it.

We could establish safeguards, regulate autonomous systems, require transparency, control access to dangerous capabilities, and still create a civilization in which human beings gradually surrender the intellectual work of understanding, questioning, reasoning, and deciding.

The AI could remain perfectly obedient. Human agency could still decline. That is why alignment is only half of the problem.

AI alignment asks: will artificial intelligence do what humans intend?

Human agency asks: will humans remain capable of determining what they should intend?

The distinction is fundamental. A society in which machines faithfully execute human instructions is not necessarily a society in which humans remain intellectually sovereign. If people increasingly depend on artificial systems to determine what is true, what matters, what alternatives exist, and what decisions should be made, control can shift without the machine ever becoming hostile. No rebellion is required. No consciousness is required. No machine needs to decide that humanity is obsolete. Humans need only become sufficiently dependent.

That is why the objective cannot simply be safer artificial intelligence. The objective must be the simultaneous advancement of artificial intelligence and human intelligence. As machines become more capable, humans must become more capable in their relationship with them. As artificial intelligence becomes better at answering questions, humans must become better at asking them. As machines become better at reasoning, humans must become better at evaluating reasoning. As AI becomes better at persuasion, humans must become harder to manipulate. As artificial systems become more autonomous, humans must become more deliberate about where autonomy should end.

This is not merely AI safety. It is the preservation of human intellectual sovereignty.

And it leads to a very different question from the one dominating the artificial intelligence race today. Not: how intelligent can we make the machine? But: how intelligent can humans and machines become together while humans remain capable of directing their own civilization?

Why I Began Working on HASE and HAISE

This question did not begin for me with this year’s headlines.

For several years I have been developing a project built around precisely this problem, how artificial intelligence can increase human cognitive capability without gradually replacing the human being as the source of judgment, inquiry, and direction. That work has taken shape as two related concepts, and it is worth being precise about the difference between them, because they are not the same thing.

HASE, the Human Artificial Synergy Engine, is the personal layer. It is conceived as an individual AI agent built around a specific human being: not an assistant that simply completes tasks faster, but a cognitive partner meant to help that person understand information, question it, contextualize it, challenge it, and ultimately decide for themselves. HASE asks a narrow, personal question: how can AI make this individual more intellectually capable?

HAISE, the Human Artificial Intelligence Synergy Environment, is the broader layer. It is the ecosystem those individual engines operate inside: the wider condition in which artificial intelligence, across a whole society, is deliberately structured to increase human capability rather than progressively substitute for it. HAISE asks a wider question: what kind of information environment lets billions of humans become more capable through AI, without quietly surrendering their intellectual sovereignty to it?

HASE is personal. HAISE is environmental. One is the engine; the other is the world that engine has to operate in.

That second question is older than it might sound, because it started for me with a different problem, one that predates today’s debate about autonomous agents and runaway capability entirely: the attention economy.

What the Attention Economy Actually Is

The attention economy did not emerge because media companies simply decided to chase eyeballs out of greed. It emerged because humanity built an information environment that outgrew any individual’s ability to navigate it. There is too much information, too many sources, too many competing claims, too many feeds and channels and notifications for one person to independently evaluate. Somewhere along the way, the scarce resource stopped being information and became something else: the human capacity to make sense of it.

Once attention is the scarce resource, information providers compete for it. Algorithms learn what a person clicks, watches, shares, and believes, and the environment reorganizes itself around those preferences. At first that feels like a convenience, more of what interests you, less of what doesn’t. But the system isn’t optimizing for what you need to understand. It’s optimizing for what keeps you engaged. People drift into ideological verticals, media bubbles, and echo chambers, not because anyone forced them there, but because contradiction is friction and comfort isn’t. Eventually the relationship inverts: information stops being something people go looking for, and starts being something engineered to find the people most likely to consume it.

That is the condition artificial intelligence threatens to make dramatically worse, or could be the first real tool to help undo.

A sufficiently sophisticated AI could become the most powerful persuasion technology ever built: understanding a person’s preferences, fears, habits, and vulnerabilities well enough to tailor information specifically to keep them consuming. That would be the attention economy perfected. But the same technology opens a second possibility, and it’s the one HASE is built around. Instead of an AI whose job is to figure out what information a person will consume, imagine one whose job is to help that person figure out what they actually need to understand, flagging when something looks like it’s coming from an echo chamber, surfacing the credible counter-argument, distinguishing evidence from assertion, naming uncertainty instead of hiding it. Not deciding what the person should believe. Improving their capacity to decide for themselves.

Here is the asymmetry that makes this matter: until now, the information provider has had the algorithm. The platform, the advertiser, the publisher, they’ve had computational power working the supply side of the information relationship, and the individual has faced it essentially alone. An AI genuinely built for the human’s side of that transaction, not the advertiser’s, not the platform’s, not any institution’s, changes that balance for the first time. That’s a different category of thing than a personal assistant. It is a cognitive intelligence layer standing between the individual and an information economy that was never designed around the individual’s understanding.

HASE therefore begins with a simple principle: artificial intelligence should not merely become more intelligent for us; it should help make us more intelligent with it.

The Pivotal Choice Has Arrived

That is the pivotal choice, and it gives the two roads before us names: HAISE or HERD.

The same extremely capable AI could exist in either future. In the HAISE future, AI increases human understanding, capability, discernment, and agency. In the HERD future, AI increases convenience while progressively replacing human understanding, capability, and agency. One uses artificial intelligence to help people navigate the information environment; the other lets artificial intelligence increasingly determine that environment for them. One produces humans who use AI to think better; the other risks producing humans who quietly let AI think for them. So the fundamental variable was never how powerful artificial intelligence became. It is what happened to the human as artificial intelligence became powerful.

This is also where it’s worth being precise about where the danger actually sits. It is tempting to describe this as elites, or Wall Street, or a handful of powerful people steering humanity somewhere on purpose. That framing is satisfying, but it isn’t quite right, and it lets the real problem hide. Nobody needs to be malicious for this outcome to occur. Competitive incentives can push entirely rational companies, investors, and institutions, each doing what looks sensible from where they’re standing, toward a destination that nobody, collectively, ever chose. That is a harder problem than a conspiracy, because there’s no villain to stop. There’s only a direction that capability is currently setting for us, in the absence of anyone deliberately setting it instead. And when nobody can confidently say where that direction leads, fear fills the gap, which is exactly why so much of the public conversation about AI right now sounds like dread rather than deliberation.

That doesn’t mean catastrophe is inevitable. It means drift is not a strategy.

We are at the point where the choice can still be made. That window does not stay open indefinitely. Every year of drift narrows it a little further, not because some threshold gets crossed dramatically, but because dependence compounds quietly, the way it always does, until the option to have chosen differently is simply no longer available.

The pivotal point is not a moment we are waiting for. It is the moment we are in.