Dealing With AGI: Tracing the Singularity Back to Today

Most essays on artificial general intelligence work forward. They start at the present, layer assumption on assumption, and arrive—after a great deal of throat-clearing—at a guarded prediction. The result is honest, and almost completely useless for figuring out what to actually do this year. The piece you are reading takes the opposite approach. It starts at the singularity and traces the path backwards—through the human work that gets restructured first, the institutions that adapt or fail, the professions that are no longer there in the same form, and the responsibilities the model builders inherit on the way. Done seriously, the exercise produces a far sharper instruction set for the present than any forward extrapolation. It also makes it possible to plan, build, and live—rather than watch.

This essay is the written companion to dealingwithagi.com—an immersive piece I am building to make the same argument visible: a Maslow-by-Maslow reverse-traversal from the singularity back to today, with scenes designed to be experienced, clipped, and argued with. The site is the long version. This is the manifesto.

The premise — start at the end

Imagine, for a moment, that the singularity has happened—on whatever timeline you find honest. Pick twenty years, pick fifty, pick any horizon where general intelligence at or above human level is broadly available, embodied, and integrated into infrastructure. The end state is not the story. The story is the order in which we get there, because that order tells us which categories of human work are restructured first, what gets reorganized next, what the late-stage settlement looks like, and— critically—where the responsibility lies at each step.

Maslow's hierarchy of needs is, for all its limitations, the most useful map for that traversal.^[1] Built for one purpose; surprisingly load-bearing for another. The hierarchy tells us where humans are most and least defended against substitution: the higher-order needs—esteem, self-actualization, meaning—are the slowest to be replaced and the first to be misunderstood. The lower-order needs—physical, economic, safety—are the fastest to be restructured by anything that compresses the cost of capable work. The traversal goes top-down from the singularity, layer by layer, back to today.

Layer one — self-actualization (the slowest to dissolve, the fastest to be misunderstood)

At the top of the hierarchy live the things humans do for reasons that do not reduce to a function: creating, exploring, deciding what is worth caring about. These categories survive AGI in the literal sense—the model can write the poem, design the building, win the contest—but the human practice of them does not disappear. It transforms. The painter does not stop painting; the painter stops being paid for technical execution and starts being paid (when paid at all) for the framing of why this painting now. The novelist stops being paid for the prose and starts being paid for the sentence that nobody else could have written. The work narrows to the part that is irreducibly the human's, and the rest is delegated. This is not a tragedy; it is what the high-leverage version of these professions always looked like, finally made visible by the disappearance of the busywork around them.

What the model builders owe at this layer is restraint: refuse to ship the version that pretends to replace human meaning rather than the one that augments it. That is a values question, not a capability question. The clearest public attempt to write down a normative spec for that restraint is the kind of constitutional framing Anthropic has tried to publish, with all its limits.^[2] The companion warning is the dynamic in Defensive Agents: The Weaponization of Adoption—the model builders are not neutral, and pretending otherwise is the most dangerous form of naïveté.

Layer two — esteem (status, recognition, the social fabric of work)

Below self-actualization sits the layer where most of modern professional life actually happens: status, recognition, the experience of being valuable to a community of peers. AGI hits this layer hard. When a general system can write the brief, code the system, design the campaign, draft the contract, and run the analysis, the recognition economy that organized human work for two hundred years has to be rebuilt. The credentials thin. The proof shifts. The visible body of work matters more, not less, because it is the only thing that distinguishes one operator from another when the technical floor is shared.

This is the layer where the structural argument in The Individual Contributor Age applies in full, and where the hiring rebuild in The Battle of Mindsets becomes load-bearing for society, not just for one team. Professions that mostly conferred status through credential—rather than through demonstrable, recurring, falsifiable work—lose their grip first. Professions that always organized themselves around a public, visible body of work do not. The asymmetry is not subtle, and it favors a particular kind of operator we should be deliberately training more of.

Layer three — belonging (the institutions that hold us together)

Below esteem sits belonging—the institutions that organize human life: workplaces, schools, churches, unions, professional bodies, neighborhoods. These do not collapse because of AGI; they reorganize. The workplace is the first to feel it, then the school, then the credentialing institution, then the professional society. Each one has to decide what it is actually for, now that the easiest thing about it—the certification of who is competent—is being substituted by the general system.

The honest answer at this layer is that most of these institutions will survive, and most of them will be very different. The school that defined itself by content delivery will struggle; the school that defined itself by mentorship, community, and the formation of judgment under pressure will compound. The workplace organized around the protection of headcount will collapse; the one organized around decentralized delivery and real governance will accelerate. The professional body that gatekept access to a credential will lose its purpose; the one that holds the standards after the credential—ethics, ongoing practice, the conduct of the profession in the public eye—will be more important than ever.

Layer four — safety (the economic and physical floor)

Below belonging sits the safety layer—income, healthcare, predictability of life. This is the layer where AGI's effects are felt most directly by the people with the least margin to absorb them, and where the political and economic responses will be most contested. The honest part of this conversation is that the labor displacement is real, it is broad, and it is happening unevenly. The dishonest part is to pretend it can be handled by retraining alone. The professions that disappear at this layer disappear functionally even when the title persists: the title may still be held, but the work it describes has been delegated to the system.

The model builders owe a great deal at this layer—both in terms of how they release capability and in terms of the public conversation they participate in honestly. The institutional response owes even more. The least useful contribution any serious person can make is the comforting story that "the new jobs are coming"—they are, in some categories, and not in others, and the difference matters at household scale. The most useful contribution is to be specific about which categories are restructured first, to name the timelines candidly, and to support the institutions that have to absorb the change.

Layer five — physiological (food, water, shelter, the unglamorous infrastructure)

At the bottom of the hierarchy sits the unglamorous layer that almost never makes it into AGI conversations: the food supply, the water systems, the energy grid, the housing stock, the medical supply chain. This is where the singularity should help most and where the institutional gap is widest. Cheap, general capability deployed against these problems is unambiguously good—better drug discovery, better agricultural productivity, better climate modeling, faster materials science, more reliable infrastructure. The question is not whether the capability arrives. The question is whether the institutional surface—governments, NGOs, regulators, public-research bodies—can absorb it fast enough to convert it into outcomes, or whether the gains stay locked inside private optimization.

This is the layer where the strongest pro-AGI argument lives, and it should not be abandoned to either the maximalists or the doomers. It deserves serious operators willing to spend their compressed time on compounding goods.

The role of the model builders

Tracing backward from the singularity makes one thing legible: the model builders are not in a position of neutrality at any layer. At the top, they are choosing what kind of substitution they ship. At esteem, they are choosing which credentials they bypass and which they reinforce. At belonging, they are deciding what institutions are easy or hard to build on top of their systems. At safety, they are deciding what they release, how, to whom, with what guardrails. At the physiological layer, they are deciding what problems they direct compute toward and which they leave to slower public processes.

The reasonable response to that asymmetric responsibility is not to demand the model builders be perfect. It is to demand they be specific, public, and accountable—and to build the institutional and individual capacity that does not require their virtue. The argument for hedging away from any single provider, picked up in Hedging with Open Source, applies to the entire social fabric, not just to enterprise procurement.

The professions that are not there in the same form

Said honestly, the traversal produces a list. At the technical-execution layer, the professions whose defining work was producing something a general system can produce—routine code, routine prose, routine design, routine analysis, routine legal review, routine accounting, routine customer support—are already in functional restructuring. They do not vanish. They consolidate. The remaining humans in them operate at a higher level of judgment, with the system underneath them, and at smaller headcounts. The humans displaced are real, in real numbers, with real households.

At the orchestration layer, new professions emerge: operators who own end-to-end workflows, skill-library maintainers, evaluation engineers, alignment auditors, agent supervisors, the people who specify the problems precisely enough that the system can solve them at scale. The earliest version of this work is what I describe in The Fractional Digital Employee; the mature version is recognizable and respectable career territory, but it is not the same career your training program prepared you for.

The professions that compound

Three families of work compound under the traversal, and they are the right place to be standing now if you have the choice:

• Judgment under irreducible uncertainty. Strategy, ethics, regulation, governance, the shaping of institutions, the conduct of difficult negotiations. The model accelerates the analysis; it does not absolve anyone of the call.
• Craft tied to embodied presence. The teacher who actually mentors, the clinician who actually listens, the artist who is in the room, the leader whose authority is earned by being accountable to people who can see them.
• System-builders. The operators who build the durable infrastructure—skill libraries, knowledge layers, brand systems, public commons—that the rest of the economy will run on top of. The thesis for that work is in Why I Back Agentic Infrastructure.

What dealing with AGI looks like in practice

Five concrete moves a serious person can make right now, derived from the backward traversal:

• Compound visible work. Publish. Ship. Maintain a public surface that proves your judgment over time. The case is in A Public /writing; the principle generalizes beyond writers.
• Pick a layer of the hierarchy and serve it deliberately. The physiological-layer builder, the safety-layer institution, the belonging-layer educator, the esteem-layer operator, the self-actualization-layer artist. Pick one. Be useful at it. Refuse to pretend you can serve all five.
• Build the in-house capability you need. The argument for not waiting on someone else to give you the stack is in The Individual Contributor Age. The argument for not waiting on a single model provider is in Hedging with Open Source.
• Support the institutions whose mission still matters. The school that mentors, the professional body that holds standards, the union that protects the floor, the open-source project that holds the commons. Their work gets harder, not easier, and they are still where the legitimacy of the next decade comes from.
• Hold the model builders publicly accountable. Not as enemies. Not as gods. As consequential institutions whose decisions deserve the same scrutiny we apply to any other organization whose product shapes daily life at this scale. The asymmetric responsibility argument from The Agentic Enterprise applies more broadly than to procurement.

A preview of the immersive piece

dealingwithagi.com takes this traversal and makes it inhabitable. The site walks the visitor backwards from the singularity, scene by scene, through each layer of Maslow—what the world looks like at the top, what the model builders chose, which professions stayed, which were restructured, what the institutions did or failed to do, and how it all traces back to a set of decisions being made right now. It is built to be experienced once at length and then clipped, argued with, and forwarded. It is not a forecast. It is a frame.

The site is in build. If the manifesto resonates, the immersive piece is the one I want you to see when it ships, and the one I want you to send to whoever in your life is still pretending this is a speculative conversation.

Working forward from the present, AGI sounds like an argument. Working backward from the eventual state, it becomes an instruction set: which work to do, which institutions to defend, which professions to restructure on purpose, which model-builder behaviors to demand publicly, which compounding moves to make before the rest of the field catches up. The traversal does not make the future easier. It makes the present clearer. That is the most useful thing an essay on this subject can do.

Related: Why I Back Agentic Infrastructure, The Agentic Enterprise, and The Individual Contributor Age.