On Personification of Things and “Thingification” of People
The communist hyper-states vs ultra-“nationalist” enclaves of the AGI future
Disclaimer
This is an essay about hypothetical consequences of AGI arriving as soon as possible based on our observations on the state of global policy making, ideologies driving the technologists and human nature in general. It’s a deliberately uncomfortable text. We just took today’s trends to their logical extremes, because that’s where their internal contradictions show up.
For now, we refrain from professing solutions to the posed problem in this piece, instead letting the readers marinate in these hypotheticals and process the patterns around them through this lens.
(Dis)Comforting Truth
Before diving into our extreme case-building of what “AGI around the corner” may practically imply on a macro scale, let us confess why we treat them only as hypotheticals. Our base scenario is much more prosaic: the researchers stop dictating where the market allocates capital soon enough to postpone AGI’s arrival for another 5-10 years or maybe more, which gives the society and institutions more time to gradually transition into the age of (almost) complete automation. The reason is simple and has always been along the same lines — protecting shareholder value. The “business people” will take back the reins, as they usually do, unless among AI leaders there are those who’re savvy enough to evolve their products as fast as necessary to fund the next round of their frontier aspirations.
But so far it very much resembles the movie we’ve seen before with crypto: the nerds become so arrogant in pursuit of meta-goals (thinking their revenge has truly come this time) that they forget to sell the dream to outsiders by widening the touchpoint area with constantly improving products. The models are indeed improving, although not as fast as the benchmarks seem to suggest, but the product stays the same – those marginal improvements become less and less visible to regular users. That’s where the “business” people come in, in our view, to polish the UX and juice the margins to finally return the capital. But there’s a whole parallel debate about it already happening and we prefer letting others contribute to it more than we could ever offer.
So instead, let’s get back to (our) edge case that is, truth be told, still the base scenario for the rest of the market. What if Dario/Sam in the West or Yang/Liang in the East get to AGI before their investors and creditors start “protecting” the business? To paraphrase one of these guys’ newly mythologized investor meeting:
“Products are side quests on the road to AGI”.
(Dis)Comforting Hallucination
So, let’s suppose AGI is about to arrive in 2-3 years (which has been said for the last 2-3 years already) and starts solving all of humanity’s big problems. We’ll try to extrapolate what it leads to using methods of induction and study of history.
Let’s start with the belief itself, which many analysts have equated to a religious one - that AGI is very much attainable here and now. Objectively, it is grounded in reason: scaling laws and other optimizations do work. However, we are also witnessing the slow but clear acceptance by the “non-believers” that AI can be very knowledgeable in certain areas, consistently improve on benchmarks and solve previously unsolved math and (hopefully soon) similar scientific problems - in the abstract, platonic, so to speak, space. It is indeed a powerful tool, but real goal setting, autonomous planning around it, picking the right problems to get there with and inference from limited feedback — which is exactly what made humans eventually progress to this stage — is currently lacking in AI and seems more of a fundamental (a.k.a. algo) problem, rather than resource limitation (scaling/optimization) issue. Realistically, one needs to be a professional/expert in their field of application to truly leverage these models sustainably today. But for a 100x engineer, 10x lawyer or 5x content creator these things indeed are indispensable already.
Ok, so we’ve established that what we have today is not AGI. So, what is AGI? Here is our rough definition: a recursively self-improving intelligence system that no longer requires human planning or oversight to work 24/7 on humanity’s aggregated problems both in the digital (idealistic) and physical spaces by serving specific interests of individual users. By this definition (but not in principle) its alignment with the interests of its user base, rather than e.g. creators/operators, is implied by the logic that a system conscious of and consistent with its purpose to serve humans (rather than itself) would opt to serve as wide of a group as possible; another implication of our definition is existence of persistent memory, which would allow such a system to take the initiative in proposing new solutions by ingesting data on an on-going basis.
It is clear that what we have right now is nowhere near this level of consciousness, and hopefully the critics won’t blame us for expecting a little too much from an “AI God”. But the majority of AI researchers, who are also the people most one-shotted by their own algorithms, seem to have more lopsided expectations with information processing capability being the primary factor. Thus “AGI around the corner” is still being pushed, which forces us to consider it more of a hallucination than reality, albeit a reality-distorting one unless scaling limits get finally hit (whether for chip, electricity or capital constraint reasons). So, to paraphrase the same guy we quoted earlier:
“Hallucination is a product problem [a.k.a. a side quest] — we will address it, but it is not a priority.”
Core Thesis of the Essay
Ok, so the “AGI” that arrives in 2-3 years’ time or sooner may not fit our definition, but as determined by AI researchers, passes that hype test. Then whatever they end up calling AGI, without a breakthrough in algorithms, will at first be a really messy, but still effective, set of agents and automatons mostly dealing with mundane and mechanistic labor, but at tremendous speeds — which is exactly what will produce the productivity gain. That thing will become good enough at 70-80% of human service jobs (that are actually more mundane and mechanistic than people make it out to be) as well as science that requires big data pattern matching more than insight (genomics, cellular biology, chemistry — including industrial applications) — enough to usher in the real acceleration. There is no need for embodied AI to replace the current physical capabilities beyond the contours already visible for it to also add to this boost: full production line automation and end-to-end completion of simple but voluminous tasks like autonomous driving, broader logistics, including higher load delivery, basic civil engineering and construction, inspection, etc.
So, actually let us agree with the researchers and just shift our definition of AGI from the substantive one to the one that deals with its first order effect instead — extreme productivity gain. This way, if whatever is called AGI turns out even much better and closer to our original definition, then it just adds a larger multiple to the current baseline productivity boost.
Now, to the thesis itself: we see an extreme scenario barbell if even this limited/“hallucinated” definition of AGI arrives really fast and there is no sufficient economic theory to prepare us for it. Humanity may fall back on old ideas with a renewed appeal given automation and eradication of the so called disutility of labor: communism (for more autocratic states)/radical socialism (developed democracies) - which will expand control the regions under their instances of AGI. This development will prompt those with means and some autonomy to move out of those countries and create ultra-"nationalist" enclaves, with the idea of a “nation” being used loosely here - meaning groups of interest, wealth level, ideology and, potentially, religion and race. Those enclaves, presumably AI-rich to the level of sovereignty, may be more free to explore the design space of Übermenschen, otherwise unexplored by the communist side given ideological conflict with the natural equality of all under the AI’s embodiment of the collective’s supremacy.
The bottom line is, acceleration, even if selective, that boosts economic output to the extent that human labor definition and expectations shift drastically across multiple industries, will disrupt the existing economic and social orders, as both have been unbalanced enough already. The only natural reaction to counter-balance its effects would be going into extremes at other ends, which is exactly what produces this fall back onto the already resurging ideological alternatives to cope with the change. The absence of new ideas around human cohabitation and governance creates a longing that will be filled with the old ones.
Socio-Economic Factors
As just noted, none of those political shifts begin with the arrival of AGI, the thing just acts as accelerant to social conditions already present. Let’s establish them:
- A K-shaped economy of “haves” and “have nots”, to borrow Ray Dalio’s verbiage. The top 10% of US earners now account for 49.2% of all consumer spending — the highest share since the data collection began in 1989, up from roughly 35% in the early 1990s — while the spending of the bottom 80% has merely kept pace with inflation since the pandemic.
Figure 1. The K-shaped Economy. Sources: Federal Reserve Distributional Financial Accounts; Moody’s Analytics
- Racial, religious and national clashes across developed western countries, prompted by mass migration from the less developed Global South to the North — another “haves vs have nots” configuration, in which the local have-nots blame the incoming ones for loss of opportunity, while local haves blame the incoming have-nots for making their cities/towns less habitable.
- The lowest recorded trust towards governments in the West. 17% of Americans trust Washington (lawmakers + executive branch), down from 73% when the question was first asked in 1958; and across the OECD countries that usually rank higher on this metric more people now report low or no trust in government — 43% against the 40% who report high or moderately high trust, according to the OECD Trust Survey published in June 2026. The dynamic is growingly negative even in historically well-to-do countries : 49% in Finland vs 61% in 2021 and 39% in the Netherlands vs 49% in 2021 - with only 35% in Germany, 27% in the UK and 22% in France.
Figure 2. The Trust Collapse. Sources: Pew Research Center (US, 1958–2025); OECD Survey on Drivers of Trust in Public Institutions, 2026 results (fieldwork 2025); UK shown at 2023, the latest published country note.
- A huge and growing divide between financial wealth and actual wealth in the economy. Astronomical paper valuations, especially — the private AI labs, or the world’s first trillionaire minted by the SpaceX IPO this June — set against working people’s stagnant wages and buying power. US equity values now sit at 2.4 times corporate net assets, close to double the historical average. The financial riches are openly made with the support from the Administration in the US, while the same Administration is responsible for “punishing” the regular folks with the price spikes of e.g. gas and imported goods.
All of this creates a classical setup for the precondition of social instability — where the haves are mostly the financial and knowledge economy class, and the have-nots are anyone from a “useless” social-studies-diploma-holder-turned-barista to a welfare-exploiting immigrant. Now add even 5% of currently employed junior and low-to-mid-level urban area labor to this mix and it becomes a much bigger problem.
The broad automation of both the manufacturing and services portions of the economy inevitably displaces a big chunk of the working class for at least a few years, while the alternative “creative job” routes fit neither the skillset nor the expectations of their majority. Lower income cohorts in developed economies do not have the savings to sit that out. In the United States, only 46% of adults hold three months of expenses, 24% hold nothing at all, a third could not cover even a single month, and 59% could not absorb a $1,000 emergency without borrowing. Across the OECD roughly 36% of people are not income-poor but hold too little liquid wealth to keep themselves above the poverty line for three months, and a further 11% are already both income- and asset-poor (OECD, Society at a Glance). In the EU the more forgiving question — could you cover one unexpected expense — still gets a “no” from 30% of the population. Without a clear transition strategy that doesn’t just involve government handouts that never end up producing any sustainable growth, we may face a few years’ gap for a big chunk of labor to get re-qualified in developed economies, which is enough to produce significant social unrest conditions.
Figure 3. Average Household Savings. Sources: Bankrate Annual Emergency Savings Report 2026; US News Financial Wellness Survey, January 2026; OECD Wealth Distribution Database via Society at a Glance; Eurostat EU-SILC 2024.
Now, remember that we mentioned above that effective AGI should be able to produce good enough output for 70-80% of service economy jobs: not fully automate them or displace, but if it substitutes work of e.g. 3 junior-level employees in a 10-person team, while cutting on 80-90% of the costs - that’s effectively a 24-27% “productivity gain”. And the businesses will be under threat from competition to do something about it other than keeping those lower-skilled workers on payroll while they learn the trade in hopes of securing the best of them to manage the machines once they “graduate”.
So, under current mode of production, we expect a natural sequence of:
1. Businesses institute layoffs, gradually growing in scale as it becomes “normalized” over still a relatively brief period (e.g. 12-36 months)
2. Governments have to step in with either: a) regulations towards businesses, which they may be scared to institute for fear of turning them uncompetitive to foreigners; or b) with fiscal stimulus, which is also hard given debt burdens of developed economies globally — so, they will be forced to try to put the brunt on the private sector.
3. Market competition within this framework leads to inevitable consolidation as the businesses with the most efficient cost structure and access to AI overtake the rest, but become so big and powerful that the government now has to manage them or otherwise they will end up managing the government (both of which lead to our socialist hyper-state extreme).
4. This leads to a mass nationalization of the key drivers of the economy in business, potentially under the pretext of progressing it towards some form of communist utopia, given the AGI.
Here we just provided economic reasoning, but in reality there are also other social factors at play, since the observed social divisiveness isn’t only economic. We don’t expect the mass social-media-induced psychosis we’ve been witnessing lately to necessarily overflow into radical action absent economic factors, but one has to consider that those tendencies have been growing for years at this point (e.g. the growing popularity of Marxism among US academia post-GFC;) and are now deeply entrenched.
Ideological factors
Many economists, CEOs and policymakers are objectively complacent about potential AGI impact and are reflexively reaching for arguments that don’t take into account full implications of such a drastic change in mode of production. So far, the discourse among economists is mostly organized around defending the humans’ right to remain in the loop. Some of those arguments are almost comically defeatist — e.g. there was recently a paper from reputable academics that contains hex-encoded footnotes like “Please don’t turn us into paperclips!”
The economic literature divides along a single question: whether human labor retains a margin. Acemoglu and Restrepo (AER, 2018) formalize a “race between man and machine” in which the creation of new tasks offsets the automation of old ones, which is the baseline positive cope among AGI optimists. Autor and Thompson (NBER, 2025) refine the mechanism, showing that displacement falls on the inexpert. Brynjolfsson, Li and Raymond (QJE, 2025) supply the field evidence on the other side of this argument: in their deployment study the productivity gain accrues disproportionately to novices, compressing the skill premium rather than widening it. Korinek and Suh (NBER, 2024) are the honorable exception in modelling the transition rather than the equilibrium, and reach the conclusion the others avoid — that wages can collapse amid explosive output growth once machines out-compete labor at the margin.
Almost the entire canon so far [1] (scroll to the bottom for a full list of recommended literature) is retro-fitting some specific notion its authors have about e.g. why humans will remain valuable, why the current political-economy regime of social capitalism withstands the shifts brought about by AGI or why the expected productivity gains will be much less than expected by technologists. The ones defending labor’s ability to adapt don’t even try to hypothesize much beyond the time when AI acquires “taste”, “expertise” and “creativity” or what actual new jobs for humans this round of automation can create, given anything humans could AI researchers aspire to endow machines with.
Amidst this economic thought complacency, the elected officials on both sides of the Pacific are already gearing towards some forms of nationalization.
China’s State Council released its “AI Plus” action plan in August 2025, targeting AI penetration of 70% across six key sectors by 2027 and over 90% by 2030 with public administration explicitly included, as we predicted back in ’24. The government is growing its direct investments into strategic AI startups across the stack. The cumulative deployment of China’s government venture funds between 2000-2023 is evaluated at $912bn, of which 23% went to AI-related firms (i.e. deployed over last couple years). The vehicles are only getting bigger:
- The state's Big Fund III raised RMB 344bn (~$47.5bn) for semiconductors and lithography in 2024;
- Dedicated National AI Industry Investment Fund of RMB 60bn (~$8.2bn) followed in 2025;
- Three further early-stage “hard tech” funds put more than $7bn behind sub-$71m startups in chips, quantum, biomedicine and brain–computer interfaces;
- And in December 2025 the State Council seeded a National Venture Capital Guidance Fund with RMB 100bn (~$15bn) of stated RMB 1 trillion (~$144bn) target.
Figure 4. The Chinese State as Venture Capitalist. Sources: Beraja, Peng, Yang & Yuchtman, “Government as Venture Capitalists in AI” (NBER, 2024), via Stanford SCCEI; State Council announcements; CCTV; Zero2IPO.
It is also using non-financial means of support — municipal “computing power vouchers” in Beijing, Shanghai and Shenzhen that subsidize rented training compute, AI pilot zones across some 20 cities offering preferential financing and a lighter regulatory touch, and provincial power discounts of up to 50% for data centers in Gansu, Guizhou and Inner Mongolia. The State Council’s AI plan itself sets local targets for how many open models the ecosystem should create, since open weights are easy to fold into government agency and SOE workflows, as well as easy to export across the Global South. It is their explicit diffusion strategy in line with communist belief that human corpus of knowledge (which essentially is embodied in AI) is a common good, the fruits of which need to be evenly distributed across the society.
The US administration, for its part, has taken to acquiring stakes in the companies most benefitting from the trend: a ~10% stake in Intel converted out of CHIPS Act grants, a 15% Department of War stake in MP Materials, positions in metals and steel companies; and, as of this month, OpenAI itself is reportedly offering Washington a 5% stake, floating the idea that America’s leading labs allot equity into a sovereign vehicle. This last bit is tangential to a softer nationalization idea in circulation: hand citizens the upside via shareholder capitalism, along the lines of Peter Drucker’s half-century-old observation that pension funds made American workers the beneficial owners of the means of production — “pension fund socialism”, as he called it in The Unseen Revolution (1976).
So, both superpowers are already performing a form of old-world-style nationalization, while wearing different ideological uniforms. This is a natural development, however it isn’t yet clear if the government can control the technology and its operators the way it was able to for critical industries in the past.
And what about the AI researchers themselves? So far, it’s clear that no-one besides them is seriously thinking about how to turn AGI into a well-functioning arrangement that distributes the fruits of the breakthrough according to society’s needs. Their best attempts, like Amodei’s “Machines of Loving Grace” — joined by Altman’s “The Intelligence Age”, to which we could sprinkle in Aschenbrenner’s “Situational Awareness” since there aren’t enough load-bearing texts in this canon — stay mostly on the hypothetical side, just like us, and short on practical political economy. From brief leaks on interviews and written in other pieces one may guess that their agenda may be broader than announced.
Amodei’s essay, which is especially interesting in the context of this article, stipulates that we may compress a century of biology into a decade, eradicate illness and possibly death itself, and help the developing nations catch up — leapfrogging the gaps of uneven economic development - like a true socialist. But he also doesn’t dwell on what some of his predictions practically entail for e.g. the Earth’s demographics if the developed world’s cultural preferences (low child-per-family count) don’t hold in the suddenly caught up developed world while we eradicate the hunger and extend the life expectancy. Should humans just be bred like cattle for the amusement of “machines of loving grace” in his worldview given eradication for the need for labor — the reification (“thingification”) of people?
On the other side, the Chinese frontier-lab developers are openly quoting Che Guevara when commenting on their US rivals’ moves, but mostly abstain from direct political claims as is typical for the Chinese private sector operators, making the conclusions about their allegiances self-evident. China has long been implementing simpler forms of AI in automation of civil and public services, so it’s no wonder they expect AGI to produce a perfect planned economy given perfect real-time data inflow. There are hints of Chinese frontier lab leaders siding with that view.
Marx is essentially self-evoked into this essay beyond economic implications, as both the American and the Chinese frontier researchers seem to share his view of history as an inevitable path towards societal progress culminating in some version of an “age of abundance”. It isn’t discussed in the investor circles enough but it is Marx, of course, who is accidentally responsible for the Accelerationist philosophy in the first place.
Primarily, through his originally abandoned, but later published notes to Vol.1 of “Das Kapital” where he theorizes about a role of general intellect , as he calls what we'd consider AGI today, in the communist society of the future where human labor is no longer serving as a tool for production of surplus value and is needed only for regulating automatons’ activity. His "prophecy" turned into a policy prescription in Williams and Srnicek’s 2013 manifesto Inventing the Future: full automation, shorter week, UBI — the concepts widely explored for years in the tech community, but apparently not enough among their capitalist financiers.
Both sides of the AGI race also seem to share Marx’s view of how accumulated social knowledge must be objectified in this general intellect/AGI — i.e. that it ends up consuming all of society’s knowledge into it. Since the knowledge is commonly accumulated, thus the general intellect needs to be a common good as well in the socialists’ eyes. This is where the two racing sides may diverge, at least for now, but with the scenario we’ve laid out above, maybe not as much in the future.
To summarize both of the last two sections – on economic and ideological factors feeding into our hypothesis - we conclude that absent any new socio-economic theory, both China and US, and potentially other big states in their spheres of influence, may slowly converge on the AI-powered welfare super-state with the only difference being who formally owns the AI stack or how the re-distribution of wealth it produces is done.
The Enclaves
Ok, we’ve established through some logical steps that bigger AI superpowers are trending towards socialist governance absent any drastic ideological shift. We’ve omitted many important sub-discussions around how the ownership of the means of production of AI, which are the data centers, may transition to or merge with the state, including by the corporations that own them effectively becoming the state, but let’s assume for now that it isn’t relevant to the final outcomes.
Taking the next logical step, one can see why some of the wealthy class would prefer relocating to territories more welcoming towards private property over their AGI instances. And with the current developments in AI-enhanced warfare, we may expect that it becomes much easier for small territories to protect their sovereignty against bigger states, especially with favoring geographic conditions. A mountainous island with additional natural barriers and manufacturing capacity to produce small-scale drones and other military robotics may be able to protect itself much better in this new environment than anyone expected even couple years ago. So we should expect at least a few of such territories to exist as the hyper-states expand their influence beyond original borders by essentially exporting their AGI.
To balance the Marxian analysis from earlier in the article with the same deep level of first-principle thinking on the opposite side - from someone this rebelling, libertarians’ group, may quote as religiously, Ludwig Von Mises: “In the doctrine of Karl Marx and his followers scarcity is a historical category only: ... once mankind has effected the leap from the realm of necessity into the realm of freedom...there will be abundance... Economics may leave it to the historians and psychologists to explain the popularity of this kind of wishful thinking... Economics has nothing to assert to the state of affairs in an unrealizable universe of unlimited opportunities. In such a world...there will be no law of value, scarcity and no economic problems. These things will be absent because there will be no choices to be made and no tasks to be solved by reason...If ever such a world were to be given to the descendants of the human race, these blessed beings would see their power to think wither away and would cease to be human, for the primary task of reason is to cope consciously with the limitations imposed upon man by nature, to fight against scarcity”. So, how do the libertarian individuals expect to escape own “thingification” in the communist utopia and maintain capabilities to reason? Humans always create scarcity and something to desire, so what would be the resources they may wish for besides private property?
Let’s consider the following. Maybe AGI does indeed eradicate sickness, and maybe even death — or at the very least stretches life expectancy towards 150 years, as Amodei predicts in his essay. Would the socialist AGI regime “of loving grace” allow for everyone to benefit from it evenly? In the 1960s there was an interesting theory from one of the cybernetics’ prominent thinkers — Jay Forrester, whose world model MIT later turned into the "Limits to Growth" study — about a global system’s equilibrium that prevents it from growing in oder to avoid stacking environmental and other externalities. The primary lever of such equilibrium would be population growth containment.
So, let us then conclude that extended life/healthspan and maybe other benefits of AGI-induced progress in life sciences become the new scarce resource the libertarians of the enclaves may be striving to maintain outside of the planned communist utopia. And from there it isn’t too big of a leap to see why a revival of the “science” of eugenics can happen on these territories.
Going beyond just racial ideas, we can see modern-day version of the hyper obsession on one’s looks to transform into the striving for physical perfection beyond what our original bodies allow: skin that resists the sun yet stays soft and clear like the movie vampires’; eyes in the most aesthetically pleasing colors; physiques that translate into power and virility of every kind.
There’s also a growing obsession with “bio-hacking” one’s brain into extreme productivity, and the adjacent concept of uploading, or merging, one’s consciousness with AI. If you think the likes of Jensen Huang, who openly speaks about it, or Elon Musk’s idea of succession in their business empires is anything less than uploading their consciousness to become a “forever CEO”, you may be too naive. But going back to simpler aspirations: it’s not really hard to picture longevity-obsessed tech millionaires of today make that leap into such enclaves allowing for experiments in AI-enhanced physique and mental health. Put these two trends together and a cohort of ultra-good-looking, ultra-bright, ultra-physically-enhanced people walling themselves off from the rest of the world with powerful AI capabilities becomes very easy to picture.
The money is already moving in that direction, if not yet at that scale. Private capital into longevity biotech ran at roughly $5.7bn across some 170 deals in 2025, and the first quarter of 2026 alone took in $3.74bn across 49 deals — 56%, on a trajectory that would comfortably beat the pandemic-induced 2021 record.
It's harder to quantify the demand for enclaves and shelters for the ultra-rich, but some of the providers spill bits that point to a growing trend there as well: Vivos told the CBC its inquiries were running over 2,000% up year-on-year, and back during COVID's early spread Rising S delivered around ten private shelters into New Zealand, which since April 2025 sells residency to anyone willing to place NZ$5–10m into the country.
There are also well-known examples of the likes of Larry Ellison, who bought essentially all of Lanai; or another Hawaii compounder Mark Zuckerberg, who's estate in Kaua‘i includes a roughly 5,000 sq. ft. shelter. There were also news of Peter Thiel’s Wanaka estate gettting its bunker consent refused in 2022. And the newest of them shows what happens when the enclave meets a bordering population absent AGI protection: Jared Kushner’s Affinity Partners is converting Sazan, a 5.7 km² Albanian island carrying hundreds of Cold War bunkers, declassified for civilian use in December 2024, into a €1.4bn resort. And in June thousands went out protesting the islan's sale on the streets of Tirana. Funny enough, Albania was the first country to implement "Government AI" by appointing LLM as a minister in order to fight corruption, but eventually depreciating it to a chat bot helping users of e-gov services.
Obviously, relevant here are also the experiments around network states, who are trying to go beyond obtaining just the land, but also seeking jurisdiction over it. Próspera, incorporated in 2017 and funded by VC money, took a Honduran territory under charter that came with its own courts, its own tax code, its own police force and a fifty-year guarantee of sovereignty. Honduras changed it anyway, repealing the law in 2022, and Próspera's sued them in return. Four years on, the zone is still building, the legal proceedings with Honduras still underway, and the State Department filing reports to Congress on its standing. The rest of the field is earlier in the same experiment. Praxis has raised around $550m, of which $500m is a drawdown facility. It has toured the Mediterranean, Latin America and Greenland looking for some land, but is yet to claim any. Balaji Srinivasan's Network School got furthest fastest by not asking for sovereignty at all: a co-living campus in Forest City, Johor, ninety days and a hundred people in 2024, four hundred by mid-2025, a ministerial visit this April calling it an emerging destination for global technology talent. Then an activist post alleging Israeli nationals among the participants set off an immigration investigation, and on 22 July the Iskandar Puteri city council revoked the licence; which was soon mitigated by the School signing an MoU with Kazakhstan to move there.
All of these examples point to a nascent but consistent trend. The tech elites want their own land and rules, and want to filter who is allowed in using some merit-based principle.
The Fork
So, the logic above leads us to the two convergent trends. On one branch, the larger states and regions continue to consolidate power over individuals by obtaining an effective monopoly over the AI means of production: socialist hyper-states in function, whatever they call themselves in form. The question even becomes, why would they actually need humans in the first place? But we'll leave it at that for now.
On the other branch, those who maintain private ownership of AGI, building their own enclaves — and, depending on their ideological or social status preferences, disallow other humans who don’t fit. Up to the level of extreme nationalism against those non-AI-enhanced outsiders.
The last time machine-driven displacement, extreme wealth concentration and a collapse of institutional trust coincided was the 1920s and 30s and the world did not calmly deliberate its way to a new equilibrium.
***
[1] The full recommended reading list, for those willing to cover the whole body of economic literature out there on the topic of AGI’s impact: Acemoglu, D. & Restrepo, P., “The Race between Man and Machine” (AER, 2018); Eloundou, T., Manning, S., Mishkin, P. & Rock, D., “GPTs are GPTs” (arXiv, 2023); Autor, D. & Thompson, N., “Expertise” (NBER, 2025); Korinek, A. & Suh, D., “Scenarios for the Transition to AGI” (NBER, 2024); Brynjolfsson, E., Li, D. & Raymond, L., “Generative AI at Work”(QJE, 2025); Trammell, P. & Korinek, A., “Economic Growth under Transformative AI” (NBER, 2023); Ide, E. & Talamàs, E., “Artificial Intelligence in the Knowledge Economy” (JPE, 2025); Dell’Acqua, F. et al., “Navigating the Jagged Technological Frontier” (Harvard Business School working paper, 2023); Acemoglu, D., “The Simple Macroeconomics of AI” (NBER, 2024; published in Economic Policy, 2025); Brynjolfsson, E., Korinek, A. & Agrawal, A., “A Research Agenda for the Economics of Transformative AI” (NBER, 2025) — nine “Grand Challenges” that work as a table of contents for the whole field; and the normative counterparts: Korinek, A. & Stiglitz, J., “Steering Technological Progress” (2025) and Acemoglu, D., Autor, D. & Johnson, S., “Building Pro-Worker AI” (2026).