Research Paper · AI and Governance Observatory (AGO) · Revised March 2026

How Microstates Will Win the AI Race

A strategic case for why microstates can use AI to deliver large-state capacity with small-state speed, raise the welfare of their citizens, and become global reference cases for the future of governance.

AGO Paper 01 Population cutoff: <1M Checked: 8 Mar 2026 Carlo Edoardo Ferraris AGO Director

1. Abstract

The core claim of this paper is simple: AI is the first general-purpose technology in a long time that can reduce some of the old advantages of administrative scale while preserving the advantages of small-state speed. Large countries will continue to dominate compute, frontier model research, and capital formation. That is not the race this paper is about. The race that matters for government is the race to implementation: who can reorganize public services, internal workflows, and state capacity around AI first, and do it credibly.

In the private sector, this shift is already visible. A compact team equipped with AI can draft faster, research faster, code faster, respond faster, and supervise more parallel work than its headcount once allowed. It still does not become a giant corporation. What changes is its effective operating capacity. A microstate sits in a similar position. It does not suddenly become a major power in raw size, but it can become much more capable than its administrative headcount and budget would traditionally imply.

For a government, that matters immediately. AI can already improve translation, public guidance, drafting, routing, case preparation, policy comparison, internal research, multilingual communication, and status lookup. These are not marginal conveniences. They are the everyday surfaces through which citizens experience the state and civil servants experience their own workload. If those surfaces improve sharply, a small country can feel dramatically more competent to its citizens, visitors, founders, and investors without first building a massive bureaucracy.

The argument is intentionally asymmetric. AI-assisted state capacity is the near-term certainty. A well-run microstate can use AI to raise service quality, compress response times, and widen institutional reach now. Autonomous sovereign decision-making is the optional upside. If, in the future, higher-autonomy governance proves technically robust, politically legitimate, and constitutionally tolerable, the country that built the AI operating layer first will be best positioned to evaluate it. If that future never arrives, the early adopter still keeps the state-capacity gains.

That is why this is not a science-fiction paper. It is a strategic state-capacity paper. It argues that microstates can use AI to deliver large-state quality with small-state speed, become reference cases for the next era of governance, and gain disproportionate influence relative to their size if they move early and execute well.

2. The Opening

Modern statecraft has a recurring pattern. Small states rarely lead broad technological revolutions. But they sometimes identify a narrow technical window early enough that the state itself becomes part of the story. The move is usually not invention. It is adoption at the level of law, administration, payments, identity, or public interface. When that happens, size stops being the main fact outsiders notice.

Estonia is the clearest recent example. In 2014 it opened its digital state to non-residents through e-Residency, becoming the first country to offer a government-issued digital identity of that kind beyond its own population.[15] Estonia did not invent the internet or digital signatures. It did something more specific for statecraft: it made a small country's administrative architecture visible, legible, and exportable. Since then, larger governments, multilaterals, and researchers have treated Estonia as a standing reference point in digital governance.

The Bahamas did something similar in monetary infrastructure. On 20 October 2020 the Central Bank took the Sand Dollar from pilot to national rollout, making the digital currency available to the general public nationwide.[16] IMF analysis later described it as the first state-backed digital currency in the world.[17] The underlying problem was local and practical: an archipelagic country with uneven physical banking access needed a better payments system. The result, however, reached far beyond the domestic problem. The Bahamas became a reference case in global discussions about central bank digital currency and digital public payments.

El Salvador offers a third case, sharper and more contested. In June 2021 its legislature adopted the Bitcoin Law, and by September 2021 it had become the first country to make Bitcoin legal tender.[17] In early 2025 legal reforms made private-sector acceptance voluntary and narrowed the public role in the project.[18] Whatever final judgment is made on that policy, the historical point remains. A small state used a monetary-technology window to force itself into the center of a global argument about money, sovereignty, and digital regulation.

These cases differ in quality and outcome. That is part of the pattern, not a weakness in it. Small states do not become globally consequential in new technical domains because they always choose well. They become consequential because, under certain conditions, they can move early enough to become the cases other states must study, copy, adapt, or reject. The window is usually narrow. It is often obvious only after the reference cases already exist.

Microstates are usually discussed as constrained actors. They are described as too small to shape standards, too dependent on external systems, too resource-limited to build deep institutional capacity, and too peripheral to matter beyond their immediate region. In many domains that has been true. Small states are often price takers, rule takers, and technology takers. What the cases above show is that this description is incomplete. At moments when the decisive move is not invention but state adoption, smallness can become an operational advantage.

AI creates a window of that kind because it changes the economics of competence. In an earlier era, if a government wanted more capability, it generally needed more people, more offices, more specialists, more committees, more process, and more time. Scale was expensive, but it was still the only way to extend reach. AI changes that logic by making more expertise, more drafting, more translation, more retrieval, and more first-pass analysis available through software. That does not eliminate human judgment. It changes how much work a small number of accountable humans can coordinate.

The same logic can be seen outside government. Smaller organizations can now extend their reach without taking on the full coordination costs of much larger ones. In statecraft, that leverage is especially meaningful because the bottleneck is rarely only intelligence in the abstract. The real bottleneck is throughput. Can a ministry answer citizens quickly? Can officials compare policy options without waiting weeks for a note? Can a licensing office route cases accurately? Can a small foreign ministry prepare briefs at a standard that looks serious in international negotiations? Can a government serve residents, businesses, and visitors in multiple languages without building a large multilingual staff? AI does not solve statecraft by magic. It does expand the amount of competent work a lean administration can supervise.

That matters more in a microstate than in a continental power because the administrative surface is more governable. There are fewer ministries, fewer legacy interfaces, fewer competing jurisdictions, and a shorter chain from political decision to operational rollout. A large country can possess superior technical talent and still fail to deploy at speed because every change touches too many systems, too many unions, too many agencies, too many legal layers, and too many political veto points. A microstate does not escape politics, but it can iterate faster.

There is also a more formal economic reason this matters. Economists know part of the problem as Baumol's cost disease: when wages rise across an economy but productivity in labor-intensive services does not rise at the same pace, the relative cost of those services keeps climbing.[13] Much of the administrative layer of government has historically behaved this way. A ministry can digitize forms, but large parts of the work still depend on people reading, drafting, routing, translating, checking, comparing, and explaining.

AI matters because it can raise productivity in exactly those cognitive service tasks. That does not mean the whole state becomes cheap or autonomous overnight. Courts, cabinet decisions, diplomacy, policing, and political judgment remain deeply human. But if Baumol pressure is weakened in the administrative layer, even partially, a microstate gets a disproportionate gain. It can expand throughput, responsiveness, and policy bandwidth without expanding payroll one-for-one. Recent field evidence on generative AI in customer-support work found meaningful productivity gains, especially for less-experienced workers, which is directionally relevant to many public-service workflows.[14]

Microstates do not need to dominate frontier AI research; they need to become the first states where AI makes government dramatically faster, clearer, and more capable.

This is also why the strongest version of the thesis is not "let the model rule." That is too crude, too early, and too easily dismissed. The immediate task is to build a national AI operating layer that improves the quality of services, internal analysis, coordination, and decision support across the state. Once that layer exists, the country gains two things at once: direct service improvement now, and the option to evaluate higher-autonomy functions later from a position of institutional maturity rather than from fantasy.

If one microstate gets there first, it will be read the way earlier small-state technology cases were read: as the first tractable proof that a different administrative model can function at sovereign scale. That is how a small country stops being treated as a peripheral user of other states' systems and starts being treated as a case that larger states must interpret.

3. Five Pillars of AI State Implementation

Once the historical logic is clear, the next question is practical. A serious national AI program still needs a structure that officials can act on. The most useful way to think about implementation is not as one giant promise about "AI government," but as five compounding pillars. Each pillar creates direct value on its own. Together they create something larger: a state that feels faster, more legible, more capable, and more consequential than its size would normally permit.

1. I

   AI Citizen Services

   One national front door for citizens, residents, and domestic businesses.

   The first pillar is the most visible and, in many cases, the most politically powerful. A citizen should not need to understand the internal map of ministries in order to get help from the state. An AI-enabled service layer can become a single public interface for permits, appointments, guidance, status checks, document preparation, and multilingual answers. It can route requests to the right office, explain requirements clearly, and stay available outside normal office hours.

   For a microstate, this is not just a digital-convenience project. It is a direct upgrade to how the state is experienced. In a small country, word-of-mouth spreads quickly. So does frustration. A government that becomes easy to navigate gains more than efficiency; it gains legitimacy. Citizens stop feeling that the state is a maze and start feeling that it is responsive. Parents, retirees, workers, students, and local firms encounter a government that is legible rather than fragmented. That internal legitimacy matters because trust in the state is built first at home.

2. II

   AI Judicial System

   Faster courts, cleaner precedent retrieval, and more consistent judicial support.

   Courts are one of the most powerful and least discussed leverage points in a small state. When judicial capacity is thin, delay spreads everywhere: contracts become slower to enforce, disputes last longer, citizens lose faith in fairness, and the whole state feels less reliable. AI can help here without replacing judges. It can retrieve relevant precedent, summarize filings, compare similar cases, surface inconsistencies, help structure opinions, and support scheduling or case triage.

   In a microstate, even modest judicial improvements can have system-wide effects. If hearings are better prepared, case materials are easier to search, and routine court administration moves faster, the country starts to feel more investable, more just, and more institutionally serious. The principle should remain strict: AI may assist judicial work, but authority stays with judges and lawful procedure. That is exactly why the pillar is compelling. It offers large upside without requiring a reckless claim about automated justice.

3. III

   AI Institutional Capacity

   Ministerial-grade analysis and administrative depth without large-state headcount.

   Microstates often operate with thin teams carrying broad mandates. The same official may need to draft a memo, review procurement language, summarize a technical brief, answer a parliamentary question, and prepare a meeting note in the same week. Small states also rarely suffer from a shortage of stakes; they suffer from a shortage of analytical bandwidth. AI can materially improve that environment by acting as a drafting assistant, retrieval layer, research aide, translation engine, policy scanner, and workflow accelerator.

   The best use of this pillar is not to replace ministerial judgment with model output. It is to raise the floor and compress the time required to become informed. A minister should be able to ask for a comparison of digital identity systems, a summary of how peer states structure AI procurement, or a briefing on tourism-platform regulation and receive a disciplined starting document in hours, not weeks. When that becomes normal, the state's quality of attention changes. The country becomes more capable of acting on opportunities because it can understand them faster and execute more of the follow-through internally.

4. IV

   AI Growth Interface

   Turn national AI capability into visible external-facing economic upside.

   Many microstates are externally facing economies. Tourism, financial services, international business, migration services, or specialized export sectors matter disproportionately. That means the state interface seen by outsiders has direct economic weight. A country with AI-enabled multilingual service layers can improve the visitor experience, the founder experience, the investor experience, and the compliance experience. It can become faster to onboard a business, easier to understand permit requirements, simpler to navigate residency or licensing pathways, and more attractive to the kinds of globally mobile people who amplify national reputation.

   This pillar is where administrative modernization becomes strategic signaling. A country that feels fast, legible, and digitally serious will be perceived as a more modern place to build, visit, register, or relocate. That does not mean AI alone creates growth. It means AI can improve the state layer through which growth sectors interact with the country. For tourism-heavy states, that can be visible quickly. For founder- or investor-oriented strategies, it can be even more consequential because the country begins to look like an unusually capable jurisdiction.

5. V

   AI Sovereign Operating Layer

   Build the supervised system first, then decide how much autonomy is desirable.

   The fifth pillar is the most strategic. Once a country has AI embedded in citizen interfaces, internal drafting, policy support, routing, and workflow supervision, it is no longer experimenting with isolated tools. It is building a national operating layer. At that stage, the state starts accumulating the prerequisites for more advanced autonomy: logs, feedback loops, override systems, human-in-the-loop review, clear task boundaries, and public familiarity with model-assisted administration.

   This is the right way to discuss higher-autonomy governance. Not as an ideological leap, and not as an excuse to automate sovereign judgment prematurely, but as an option that only becomes real after the operational layer is mature. If, over time, certain bounded domains prove suitable for greater autonomy, the early adopter will be best positioned to evaluate them responsibly. If they do not, the country still keeps everything it built: faster services, wider capacity, stronger coordination, and a reputation for leading rather than following.

The five pillars reinforce one another. Better citizen services generate cleaner feedback and stronger domestic legitimacy. Better judicial support improves trust, consistency, and the credibility of the rule of law. Better institutional capacity improves policy quality and execution speed. Better external interfaces improve how the country is seen by the world. And all of it contributes to a disciplined operating layer that makes future autonomy an option rather than a slogan. That is why AI matters so much more to a microstate than a narrow "chatbot for government" framing suggests.

4. First Movers Are Usually Recognized in Retrospect

Historical first movers rarely look inevitable while they are happening. Estonia's digital state was initially easy to read as administrative modernization in a small post-Soviet country. The Bahamas' Sand Dollar could be read as a payments solution for a dispersed archipelago. El Salvador's Bitcoin law could be read, depending on the observer, as either eccentric or reckless. Only later does the pattern become legible: when the decisive move is binding a new technology to law, procedure, and daily public use, size can matter less than decisiveness.

What later becomes obvious is that the first mover does several things at once. It creates the first administrative precedent. It forces larger institutions to respond. And it supplies the vocabulary through which later adopters and critics describe the model. Once that happens, chronology begins to outweigh scale. The record that gets studied is the one that exists.

This is why technological windows are often recognized only after they close. Larger states have deeper capital pools, larger bureaucracies, and stronger research bases. They also have more veto points, more legacy systems, and more institutional drag. When the relevant task is not inventing the technology but operationalizing it through public systems, a smaller jurisdiction can set the first serious precedent.

AI has the same structure. The most important question is not which state trains the best model. It is which state first makes model capability legible in everyday administration: services that work, courts that move faster, ministries that can analyze more, and public interfaces that feel unusually clear. If that happens in a small jurisdiction, larger states will not ignore it. They will have to decide whether it is a prototype, a warning, or both.

One careful way to frame the upside is as a possible transition in governance operating systems. Human societies have moved through different dominant forms of coordination: coercive rule, hereditary hierarchy, constitutional states, and democratic systems that broadened legitimacy and participation. If AI-assisted governance matures, the next step would not be "government by machine" in any crude sense. It would be a state whose administrative layer operates with much higher autonomy while remaining bounded by law, auditability, override, and the publicly expressed values of the people it serves. In the best case, that could make government less arbitrary, more consistent, and more legible than many current systems without severing democratic control.[10]

Scientific programs often advance this way: first prove the model on a tractable system, then scale. Connectomics reached a complete adult fruit-fly brain connectome before mouse-scale or human-scale whole-brain emulation became tractable, and Eon Systems explicitly frames a full digital emulation of a mouse brain as a step toward human-scale intelligence.[11][12] Governance may follow a similar curve. A microstate is not a toy country; it is a real sovereign system at a tractable scale. If a new operating model can be made to work there, larger states will study it as a prototype rather than dismiss it as a curiosity.

The first microstate that makes AI-native government administratively real will not be read as an exception for long. It will be read as a prototype.

None of this is automatic. A failed pilot, a reckless autonomy claim, or a sloppy rollout could easily produce the opposite effect. First-mover advantage only matters if the first move is disciplined enough to be taken seriously. That is why the near-term thesis remains rooted in AI-assisted state capacity. A country does not need to prove full autonomy to achieve outsized strategic gains. It only needs to prove that AI can make its government feel faster, clearer, and more capable in ways citizens and outsiders can actually see.

That is also the answer to the question of relevance. The point is not that a microstate suddenly becomes geopolitically large. The point is that it can become the case through which a larger governance transition is first understood. Once that happens, later debates do not begin elsewhere.

5. Supporting Comparison

The country comparison in this paper is not the thesis itself. It is supporting evidence that the addressable set is real. The screen is limited to sovereign states below 1,000,000 people and uses five public indicators: population, internet use, government effectiveness, political stability, and tourism receipts as a share of exports. The aim is not to produce an academic rank order. It is to show that there are already plausible microstates where the preconditions for serious AI adoption are visible.

Read the shortlist as a map of operating profiles. Some countries stand out as stronger reference cases because connectivity and state capacity are already high. Others stand out because their external-facing service economy makes the gains from better public interfaces easier to observe. In some cases, the first deployment would simply need to be narrower, simpler, and more tightly supervised.

Government Effectiveness and Political Stability are World Bank WGI estimates; higher is better. The "Operating Profile" column is a neutral shorthand used only to distinguish between cleaner reference cases, more service-led cases, and narrower pilot-scale cases.

The table should be read with restraint. It does not tell us which country will move first, because leadership is not a spreadsheet variable. But it does tell us that the opportunity set is not imaginary. There are already microstates with very high digital reach, credible administrative capacity, and economic structures that could make AI-based public interfaces materially valuable. That is enough to support the paper's central claim: the first real AI-native state need not be large.

It also suggests different deployment styles. Higher-capacity cases such as Andorra, Brunei Darussalam, Bhutan, and Seychelles look more plausible as disciplined reference cases. Countries with stronger tourism or external-service exposure can show quicker upside from visitor-facing and business-facing interfaces. Lower-capacity cases are not to be dismissed, but their starting point is narrower: a bounded, carefully supervised pilot layer rather than a sweeping state-wide claim.

6. Implementation Sequence

If the historical case explains why the window matters, the rollout logic explains how not to waste it. The first objective is not to make the most dramatic claim. It is to build a system that works, earns trust, and produces visible gains quickly enough that political momentum can compound. In practice, that means beginning with functions where AI can help immediately and where human supervision is straightforward.

A well-governed sequence matters because the credibility of the long-term thesis depends on the quality of the early layer. If a government skips directly to grand claims about autonomous rule, it invites skepticism and deservedly so. If it begins by making the state easier to use, easier to run, and easier to understand, it starts building a record of competence. That record is what turns a bold idea into a serious national program.

1. 01

   Build the state map first

   Inventory services, owners, rules, and data boundaries before automating anything.

   The first phase is not glamorous, but it is decisive. A government needs a clean map of its own service landscape: what services exist, which office owns them, what the legal rules are, what data they touch, where human approval is mandatory, and where current bottlenecks appear. Without that map, AI is layered onto confusion. With it, AI becomes a structured operating upgrade.

2. 02

   Launch the public interface layer

   Start where citizens and residents can feel the difference quickly.

   The first public wins should come from guidance, translation, request routing, appointment support, document preparation, and status lookup. These are high-volume, intelligible, and relatively observable functions. They generate usable metrics: response speed, escalation rates, completion rates, language coverage, and user satisfaction. They also create the simplest political proof that the program is not abstract. The state feels easier to use.

3. 03

   Instrument ministries and courts

   Give the state a real productivity layer, not just a public chatbot.

   Once the public interface is stable, the deeper gains come from internal capacity: drafting copilots, document retrieval, multilingual translation, summarization, regulatory comparison, meeting preparation, reporting support, precedent retrieval, and case-material structuring for judicial workflows. This is where microstates start to feel the large-state-capacity effect. A ministry or court that can generate stronger first drafts and research faster becomes more decisive without requiring a dramatic increase in headcount.

4. 04

   Introduce bounded decision support in consequential workflows

   Bridge the gap between assistance and autonomy before talking about self-governing systems.

   Before any serious autonomy discussion, the state should first use AI inside consequential but still supervised workflows: case triage, compliance review, policy-option comparison, anomaly flagging, draft administrative recommendations, and other high-value functions where humans remain the final authority. This is the real bridge stage. It tests whether the state can rely on AI in workflows that matter without pretending the model is sovereign.

5. 05

   Evaluate limited autonomy only after supervision is real

   Autonomy is a constitutional and operational stage-gate, not a branding exercise.

   Only after logs, overrides, testing, appeals, and role boundaries are mature should a government assess whether any higher-autonomy functions are worth attempting. Even then, the right candidates are bounded domains with clear objectives, explicit legal authority, and clear human recourse. A country should earn the right to ask the autonomy question by proving it can supervise the state-capacity layer responsibly first.

The sequencing principle is straightforward: begin where the upside is obvious and the accountability chain is short. This is how a microstate turns AI from an abstract promise into a national competence project. It is also how it creates the kind of credibility that later attracts better operators, better partners, better capital, and stronger political support for the next stage.

7. Safeguards, Sovereignty, and Optionality

The more ambitious the vision, the more important the safeguards. A serious government AI program must be explicit not only about what it intends to build, but also about what it will not delegate early. Rights-affecting decisions require special caution. Liberty, eligibility, enforcement, sanctions, visa refusal, and benefits denial should remain reviewable, contestable, and human-accountable unless and until a much higher standard of technical, legal, and political confidence exists.

This does not weaken the thesis. It strengthens it. A well-governed microstate can afford to be ambitious precisely because it can be disciplined. Smallness helps here too. Clear chains of accountability, fewer institutional layers, and tighter rollout boundaries make it easier to specify where AI is used, where it is not, and who remains responsible. The right model is not theatrical "human control" that exists only on paper. The right model is real human authority over outputs that matter, combined with transparent logging and credible redress.

• Human override for rights-affecting outputs and sensitive casework.
• Audit logs and incident reporting for all model-assisted workflows.
• Public disclosure of where AI is used, how it is supervised, and where it is excluded.
• Appeal and redress pathways that are usable by ordinary citizens.
• Evaluation across languages and user groups rather than one benchmark score.

With that said, optionality remains the deepest strategic reason to move early. Suppose that autonomous sovereign decision-making never proves desirable. In that world, the early adopter still wins. It has better service delivery, wider administrative leverage, richer institutional data, more digitally literate civil servants, and a stronger reputation for competence. Nothing about that investment is wasted.

Now suppose the opposite. Suppose that, over time, certain bounded classes of sovereign decision support or even partially autonomous execution do become trustworthy and legitimate under tight constitutional control. In that world, the country that moved early on AI-assisted governance has a profound advantage. It already has the operating layer, the logs, the supervisory habits, the public familiarity, and the institutional muscle memory required to evaluate the next step seriously. Late adopters do not. They are still arguing about whether to begin.

That is why the autonomy conversation should be framed as a call option built on real state-capacity gains, not as a leap of faith. The downside case is still strong. The upside case is potentially historic. A country that builds itself to use AI well is better off either way. That is the kind of asymmetric bet governments should pay attention to.

8. Conclusion

The default assumption has been that major powers would define the next era of governance, because they command the most capital, talent, and institutional capacity and can diffuse their models at scale. AI complicates that assumption. In the adoption race, scale is no longer the only relevant variable. Speed matters. Coherence matters. The ability to standardize services matters. The ability to align a small number of ministries around one operating model matters. On those dimensions, microstates may have an advantage that larger states genuinely struggle to match.

That is the deeper reason the idea is so powerful. For perhaps the first time in a very long time, smallness can be reframed from a structural limitation into a strategic asset. A microstate can move faster, iterate faster, standardize faster, and learn faster. AI amplifies those traits instead of punishing them. What used to look like insufficient scale can start to look like optimal surface area.

For officials, the practical message is clear: this is not a branding exercise. It is a path to better services, better institutions, higher productivity inside government, and a stronger national position in a world being rebuilt around AI. For builders and capital, the message is also clear: the first successful microstate deployment will matter because it will show that governance itself can be re-architected. That is a rarer and more consequential opportunity than almost any ordinary software rollout.

The decisive advantage in the AI era may belong not to the states that invent the most powerful models, but to the states that first turn them into a visibly superior way of governing.

The best-positioned microstate will therefore not be the one that makes the loudest claim. It will be the one that becomes visibly more capable, visibly more attractive, and visibly more modern through disciplined implementation. If one country does that first, its size will stop being the most interesting thing about it. It will become the case every other government has to explain.

9. References

1. World Bank Group. Small States Overview. Accessed March 2026.
2. United Nations Department of Economic and Social Affairs. UN E-Government Survey 2024. Published September 2024.
3. World Bank Data. Population, total (SP.POP.TOTL). Used with country-level API queries on 8 March 2026.
4. World Bank Data. Individuals using the Internet (% of population) (IT.NET.USER.ZS). Used with country-level API queries on 8 March 2026.
5. World Bank Group. Worldwide Governance Indicators. Government Effectiveness (GE.EST) and Political Stability (PV.EST) retrieved through the World Bank API on 8 March 2026.
6. World Bank Data. International tourism, receipts (% of total exports) (ST.INT.RCPT.XP.ZS). Used with country-level API queries on 8 March 2026. Latest pre-pandemic non-null observation shown in the table.
7. National Institute of Standards and Technology. AI Risk Management Framework. Accessed March 2026.
8. UNESCO. Recommendation on the Ethics of Artificial Intelligence. Adopted 23 November 2021.
9. OECD.AI. OECD AI Principles Overview. Updated May 2024.
10. Nature editorial team. Largest brain map ever reveals fruit fly's neurons in exquisite detail. Published 2 October 2024, summarizing the FlyWire adult fruit-fly connectome papers in Nature 634 (2024).
11. Dorkenwald, S. et al. Neuronal wiring diagram of an adult brain. Nature 634, 124-138 (2024).
12. Eon Systems. Careers. Accessed 8 March 2026. The company describes a goal of a complete digital emulation of a mouse brain as a step toward human-scale intelligence.
13. Baumol, W. J. Macroeconomics of Unbalanced Growth: The Anatomy of Urban Crisis. Reprinted in Baumol's Cost Disease, Edward Elgar Publishing. Original argument introduced in the 1960s and widely cited as the basis of Baumol's cost disease.
14. Brynjolfsson, E., Li, D., and Raymond, L. R. Generative AI at Work. The Quarterly Journal of Economics, published online 4 February 2025.
15. e-Residency of Estonia. Official programme site. Accessed 8 March 2026. The site states that Estonia was the first country to offer e-Residency, starting in 2014.
16. Sand Dollar. Nationwide Launch. Central Bank of The Bahamas public update, published 20 October 2020.
17. Appendino, M. et al. Crypto Assets and CBDCs in Latin America and the Caribbean: Opportunities and Risks. IMF Working Paper WP/23/37, February 2023. Used here for the Bahamas as the first state-backed digital currency and for the dates of El Salvador's Bitcoin legal-tender adoption in 2021.
18. International Monetary Fund. IMF Executive Board Approves New 40-month US$1.4 billion Extended Fund Facility Arrangement for El Salvador. Press Release No. 25/043, 26 February 2025. Used here for the 2025 legal reforms making private-sector acceptance of Bitcoin voluntary and narrowing the state's role.