Sometimes, as I explain how AI is reshaping the world, I watch as the audience moves through a compressed cycle of grief. Denial appears in crossed arms and sceptical looks. Anger flickers when an example becomes too compelling to dismiss. Then comes bargaining, usually during the questions at the end. Which jobs are safe? When should humans have the final say? What should my children study? Will I get to retire before i have to deal with this?
People laugh when I point out what is happening. Yet there is a deeper truth in that pain. AI is often presented as an inevitable force, leaving people to negotiate emotionally with a future they did not choose and cannot clearly see.
Sadly, much of the current debate about AI is stuck in the bargaining phase. Authors demand that human authorship be protected while using AI to accuse each other of secretly using AI to write. Software engineers equipped with coding agents are becoming radically more productive even as students wonder whether programming is still worth learning. Artists defend the sanctity of human creativity while experimenting with models trained to reproduce its forms. Companies promise to keep people in the loop while building systems in which human intervention becomes increasingly rare. Universities punish students for using AI even as researchers and faculty are encouraged to embrace it.
Everywhere, we are trying to mark out the parts of work that machines should never touch, while drawing those machines deeper into the work itself.
The instinct is understandable. There are real losses ahead: career pathways disrupted, skills devalued, creative work absorbed into training data, and power accumulating around the owners of models and infrastructure. Grief is a natural response to that upheaval. Yet it belongs to this moment of transition. It cannot provide a design for what comes next.
The old division of labour will not survive simply because we declare some activities human and others machine. Capabilities are moving faster than the boundaries we draw around them. More importantly, meaningful work was never a single, indivisible act. Writing, programming and discovery all involve chains of framing, searching, generating, testing, interpreting, selecting and taking responsibility.
Provenance still matters. We need to know who is accountable when a decision causes harm, whether someone’s work or identity was used without consent, and when the act of making something is central to its meaning. Yet in many other cases, asking whether an outcome was produced by a human or a machine tells us less than understanding how the whole system worked, where judgment entered, and who had the authority to decide.
Software development already shows how quickly the old categories are coming apart. GitHub Copilot’s coding agent can take an issue, research the codebase, plan and implement a fix, run tests and return a pull request for human review. The developer may write none of the new code, yet still defines the problem, sets the constraints, judges the architecture and decides whether the change should go live. So who is the programmer: the machine that produced the code, the person who specified the outcome, or the system they formed together?
Science is moving in the same direction. DeepMind’s AlphaEvolve begins with a problem and evaluation criteria defined by human researchers, then uses Gemini models and automated evaluators to generate, test and progressively improve possible algorithms. The machine participates in invention. People decide what success means, validate the result and recognise why it matters. The discovery belongs to neither participant alone. It emerges from the loop between them.
Words such as programmer, scientist, and inventor begin to lose their old clarity. They were built for a world in which intelligence could be located inside a person and tools remained separate from the people using them. That distinction is becoming harder to sustain.
We need richer ways to describe the identities and forms of agency now taking shape. Three archetypes offer a useful map: the cyborg, the centaur, and the cyberpunk. The cyborg reveals what happens when machine intelligence becomes part of the self. The centaur shows how performance emerges from the design of human-machine interaction. The cyberpunk confronts the final question of who controls these systems and who has the power to challenge them.
The cyborg: power through symbosis
In Cyberpunk: Edgerunners, David Martinez installs a military-grade implant that allows him to move so fast the world seems almost frozen around him. The effect is intoxicating. Each new piece of chrome makes him stronger, faster and more dangerous. It also changes who he believes himself to be. His body becomes a platform in a technological arms race, and every attempt to overcome his limits brings him closer to losing control. The series turns cybernetic enhancement into both aspiration and addiction. Power comes through integration, along with the danger that the additions may eventually take command.
Ghost in the Shell asks an even stranger question. Major Motoko Kusanagi has a fully prosthetic body and a networked brain. Her shell can be replaced. Her memories can be manipulated. Even she wonders whether someone with an artificial body and uncertain memories can still be considered human. The story eventually allows identity to escape the body altogether, challenging the assumption that consciousness must remain contained within one stable individual.
These stories reveal why the cyborg is more radical than the augmented worker. A copilot remains a tool beside you. A cyborg changes the boundaries of who you are.
The idea began as a solution to space travel. In their 1960 article Cyborgs and Space, published in Astronautics following a US Air Force symposium on human spaceflight, researchers Manfred Clynes and Nathan Kline proposed adapting the human body to hostile extraterrestrial environments. They coined the word cyborg to describe a self-regulating human-machine system in which artificial controls would work alongside the body’s biological processes, largely beyond conscious attention. The ambition went far beyond attaching machinery to a person. Technology would become part of the organism, freeing the human to explore, create and think.
AI gives the idea of symbotic organism a cognitive form. A model or agent may hold part of your memory, recognize patterns across years of work, anticipate your intentions and contribute to your judgment. As the relationship deepens, removing it could leave you unable to think or act in quite the same way. The pronoun starts to shift from “I used it” toward “we thought.”
We already shift our thinking into notebooks, calendars, search engines and smartphones. Researchers use the term cognitive offloading to describe changing our environment so that a task demands less mental effort. In a recent essay, I argue that civilization itself can be understood as the progressive offloading of cognition into tools, institutions and technologies that allow us to manage complexity beyond the capacity of any individual mind.
AI promises an even more intimate relationship. A notebook stores what you give it. An adaptive model can answer back, question your assumptions, connect distant ideas and change what you think next. It learns how you work, while you learn to think in ways that exploit its capabilities. The system does more than extend cognition. It enters the process through which cognition is formed.
This is where Donna Haraway’s A Cyborg Manifesto becomes newly relevant. Her cyborg was a feminist political figure as much as a technological one. It challenged supposedly fixed divisions between human and machine, natural and artificial, body and technology. Haraway argued that such categories often help make existing hierarchies appear inevitable. The cyborg offered a different kind of identity, assembled from partial connections, contradictions and chosen affiliations, with no pure human origin waiting to be restored.
Applied to AI, that idea unsettles the attempt to preserve certain parts of work as pristinely human. The emerging cyborg will be a compound self. Some memories will sit in biological tissue and others in machine infrastructure. Some ideas will begin as intuitions, others as model outputs, and many will emerge through exchanges whose origins can no longer be separated. Identity will stretch across a human body, an adaptive model, its accumulated context and the feedback loop connecting them.
There is an exhilarating promise here. Machine intelligence offers a new form of self-overcoming: the ability to remember more, perceive patterns across greater scales, explore unfamiliar perspectives and create beyond the limits of a single mind. Yet greater capability does not automatically produce greater freedom. Everything depends on who controls the systems through which the self is being extended.
Consider an executive, scientist or designer who has worked for years with a persistent AI that understands their history, language, relationships and patterns of judgment. It has become a kind of cognitive organ. Then that person leaves the organization. Can they take it with them? The company owns much of the data and infrastructure, while the individual supplied the experiences, preferences and judgment that made the system distinctive. Can the organization continue using an agent that has learned to reason and communicate like its former employee? Does that person have a right to delete it, disable it or carry some version into their next role?
Traditional offboarding assumes that employees leave behind files, messages and devices. They will be escorted out of the building by security with a cardboard box full of their posessions. And yet, cyborg workers may leave behind functioning extensions of themselves. The biological half walks out of the building while the digital organ remains on the company’s servers.
The question of who owns that extension will become inseparable from identity, consent and power. AI may help us exceed our limitations. It may also make us dependent, observable and easier to direct. The cyborg’s promise is a larger self. Its danger is discovering that parts of that self belong to someone else.
The centaur: power through co-ordination
The centaur was never a simple symbol of harmony. The human did not ride the horse. The two bodies had been fused together, governed by a shared nervous system. In Greek mythology, centaurs were often wild, violent and driven by appetite. Chiron was the great exception: a healer, teacher and mentor to heroes including Achilles and Jason. He represented the possibility that two very different natures might combine into something wiser and more capable than either alone.
Garry Kasparov gave the myth its modern form. In 1997, IBM’s Deep Blue defeated him in a six-game match, becoming the first computer to beat a reigning world chess champion under standard tournament conditions. The event was framed as a contest between human and machine intelligence, competing for supremacy over sixty-four squares.
Kasparov responded by changing the game. The following year, in León, Spain, he introduced Advanced Chess. Players could consult computers and databases during play. The machine searched tactical possibilities and exposed mistakes. The human interpreted the position, set the strategy and decided which recommendation to follow. Kasparov’s centaur kept the human and machine distinct. Its advantage came from the quality of the handoffs between them.
The twist in the story came from a freestyle tournament in 2005, where teams could combine people and computers however they wished. Kasparov later wrote that the winners were two amateur American players using three computers. They defeated grandmasters and teams with greater computing power because they were better at directing, comparing and correcting their engines. His formula was blunt: “weak human + machine + better process” could outperform stronger components connected by a poorer one. Performance no longer belonged to the person or the computer. It belonged to the coupling.
Cybernetics had anticipated this idea decades earlier. In his 1948 book Cybernetics, Norbert Wiener explored how organisms and machines regulate themselves through communication, control and feedback. In 1960, J. C. R. Licklider’s essay “Man-Computer Symbiosis” imagined people and computers working together to formulate problems, make decisions and control complex situations. The goal was a productive loop in which each participant continuously shaped the actions of the other.
Formula 1 turns that idea into a high-speed operating system. The driver is one part of a nervous system stretching from the car to the garage, the pit wall and the factory. Each car as carries more than 300 sensors and produces over a million data points every second. The driver feels grip, balance, vibration and changing weather through the car. Engineers monitor tyres, energy systems and mechanical performance. Strategists model possible futures. The pit wall compresses all that intelligence into a few words over the radio, often while the car is travelling at more than 300 kilometres per hour.
The closing laps of the 2021 Russian Grand Prix showed what happens when that human-machine dynamic is tested. Rain began falling while Lando Norris led Lewis Hamilton. Both drivers initially stayed out on slick tyres because much of the circuit still felt manageable. From inside the cars, that judgment made sense.
The pit walls could see more. Mercedes had weather information showing that heavier rain was coming, along with evidence from Valtteri Bottas, whose lap times improved after switching to intermediate tyres. Hamilton initially resisted the call to pit, then accepted it when Mercedes made clear that conditions were about to worsen. McLaren had a less complete picture and chose not to overrule Norris when he refused to come in. Hamilton changed tyres and won his hundredth Grand Prix. Norris stayed out too long and lost the lead.
The teams faced different risks. Hamilton was second with a large gap behind him, while Norris was protecting the chance of his first victory. Yet their decision architectures still mattered. Mercedes combined the driver’s local experience with weather data, evidence from another car and a clear pit-wall call. McLaren struggled to resolve the tension between what Norris could feel and what the team needed to anticipate. As former driver Jolyon Palmer later argued, the driver knew the current level of grip, while the pit wall could see the forecast, the tyre performance of other cars and the wider strategic picture.
Real organizations face similar challenges to F1 teams. Different forms of intelligence see different parts of a problem, work at different speeds and hold different rights to act. The phrase “human in the loop” tells us only that a person appears somewhere in the process. It says little about what they can influence, where they really add value, and when specifically they should act. The highest performing organizations route the right kind of intelligence to where it has the most impact.
Imagine a pharmaceutical plant operating late at night. A predictive model detects a faint change in vibration on a sterile filling line, although every measurement remains within official limits. A veteran operator hears something unusual and remembers the same sound before an earlier failure. The quality system sees no formal breach. A supply-chain agent warns that stopping production could create a shortage of a critical medicine. The cognitive architecture gives each perspective a clear route into the decision, rather than asking one manager to absorb everything and make a heroic call.
But here is the critical part. A true cognitive organization learns from the result. It records which warning mattered, whose judgment changed the response and whether the system acted early enough. Those answers improve the model, the thresholds and the way people respond next time. Human experience teaches the machine, while machine feedback sharpens human judgment. Call it a harness if you like, but it is bigger than that. Thinking like a centaur means designing an organization in which people, models, rules and agents can form better judgments together.
The cyberpunk: fighting power with radical agency
Cyberpunk began with a contradiction: the same technology that expands human freedom can also become the machinery of control.
Bruce Bethke coined the term for a story written in 1980 and published in Amazing Stories in 1983, joining cybernetics with the defiance of punk. William Gibson gave the archetype its defining world in Neuromancer, where technology has spread through bodies, streets, corporations and criminal networks. Its protagonist, Case, is a gifted data thief whose access to cyberspace can be removed and restored by people with greater power. His freedom is real, yet conditional. The technology that gives him power also determines who can take it away.
Bruce Sterling turned the spirit of technological rebellion into a literary movement with the 1986 anthology Mirrorshades. Neal Stephenson later pushed it into satire with Snow Crash, a fractured world of corporate enclaves, private infrastructure and freelance operators moving between physical and virtual realities. Cyberpunk imagined technology escaping the laboratory and becoming the environment in which power was exercised. The Matrix imagined reality itself as an instrument of control, with hackers learning to see and rewrite the code beneath it. Across these stories, technology creates the prison, the weapon and the route of escape.
This is why the cyberpunk matters far beyond neon cities and freelance hackers in leather jackets. Cyberpunk is a theory of power. Technology allows institutions to observe, predict, persuade and control at extraordinary scale. It can also give individuals the means to inspect those systems, challenge their decisions and build alternatives. The real prize is agency: the ability to understand the machinery acting on you, refuse its defaults and negotiate from a stronger position.
AI brings that struggle into everyday organizational life. Algorithmic management is already being used to allocate tasks, monitor work, evaluate performance and shape managerial decisions. These systems promise efficiency, but they can also turn management into a form of machine-mediated control. Algorithmic management may appear objective because its conclusions arrive through dashboards and scores. However their outputs still carry the goals, assumptions and blind spots of the people who designed them. And worse, they often do so without clear explaination.
Now imagine that an employee is told that a workforce model has classified her team as underperforming. Her own AI agent reconstructs the analysis and discovers that the model ignored months of stock shortages. It shows that comparable teams with reliable supply performed worse, tests what would have happened under normal conditions and prepares an evidence-based appeal.
Previously, challenging the decision might have required a data science team, weeks of analysis and access to information held several levels above her. AI compresses that capability into just enough time to make a fresh cup of coffee. The employee is now using intelligence to contest intelligence. She is an organizational cyberpunk. She is not trying to destroy the system. She is forcing it to defend its assumptions.
Strong organizations should want this kind of productive defiance: people who refuse to accept the system’s answer until its assumptions survive challenge. Frontline teams need to inspect evidence, test alternative explanations and challenge decisions that affect their work. Leaders need dissent that arrives with data, simulations and credible alternatives.
The same dynamic extends beyond the organization. An insurer may use AI to assess a claim; the customer’s agent can read the policy, test the reasoning and identify grounds for appeal. An employer may use automated screening to reject an applicant; the candidate’s agent can examine the criteria, expose missing context and request a human review. A platform may use AI to rewrite contracts and optimise commissions; creators can use their own agents to model the impact, compare terms and negotiate together.
This is the new bargaining power: the ability to meet institutional intelligence with intelligence of your own. Individuals no longer have to confront an institutional model empty-handed.
There is no guarantee that this power will be progressive. AI can help workers expose flawed decisions, and it can help employers anticipate resistance. It can strengthen citizens challenging bureaucracy, and it can strengthen states seeking greater control. Institutions will often possess better models, richer data and more power to act.
That ambiguity has always belonged to cyberpunk. Technology does not arrive with a fixed politics. Power belongs to those who can access it, direct it and decide what happens next. And if you are going to negotiate with the future, better to do so augmented, modded and armed to the hilt.
Beyond the bargain
Together, these three archetypes move us beyond the tired debate over human work versus machine work, and toward the more consequential questions of identity, architecture and power.
The cyborg asks: What am I becoming through this system, and who owns the parts of me that now live inside it? The centaur asks: Which form of intelligence should shape this decision, with what authority, and how should the system learn from the result? The cyberpunk asks: Who controls the system, who can challenge it, and who has the power to rewrite its terms?
We will need all three instincts. Cyborgs without cyberpunks risk becoming dependent on technology owned by someone else. Centaurs without strong architecture risk reducing human oversight to a rubber stamp. Cyberpunks without centaurs risk turning agency into anarchy, creating a world of powerful tools, fragmented systems and collapsed trust.
The pre-AI world was built around scarce intelligence. The next will be shaped by how abundant intelligence is combined, governed and contested. It is time to stop grieving the old division of labour. The future will belong to new kinds of humans with the power to set the terms of their relationship with the machines that watch over them.
"Will I get to retire before i have to deal with this?" I hear this from a lot of academics.