A Seed of Thought — AI, Government, and the World “We” Might Build
The preceding chapters offer developers, traders, inventors, and investors practical, structured knowledge to help them navigate a complex financial system. This chapter is different. It is a seed of thought. I believe it belongs at the end of this book, planted quietly beside the conclusion, because the people who build infrastructure and finance technology will precisely determine whether this future is built wisely or recklessly.
I have spent twenty-five years watching governments make infrastructure decisions. I have watched projects that should have been approved in eighteen months take a decade. I have watched contracts go to connected firms when superior firms existed. I have watched tax dollars disappear into procurement processes so complex and so opaque that no one person could trace where they went. I have watched lobbyists write the regulations that govern the industries their clients operate in. And I have watched genuinely good people, civil servants who entered public life because they believed in it, become slowly ground down by systems that were never designed to work efficiently.
Thanks for reading David Goodnight | Austin, Texas | Substack! Subscribe for free to receive new posts and support my work.
I am not making a political argument. The waste I have observed crosses every ideological line. It is not a problem of the left or the right. It is a problem of the structure of government itself: humans making collective decisions about the allocation of enormous resources, with inadequate information, inadequate accountability, and enormous incentives for self-dealing. This is not a new observation. It is as old as civilization. What is new, what has never existed before in human history, is the possibility of a technological alternative.
The Thesis: From the County Up
Forget the federal government, at least for now. The federal government is a system so large, so entrenched, and so thoroughly captured by competing interests that meaningful transformation from the outside is not achievable in any near-term timeframe most of us will witness in our lifetimes. I hope I am categorically wrong. However, county, city, and state governments are different. They are closer to the people they serve. Their decisions are more directly felt. Their failures are more immediately visible. And they are small enough that technology can actually reach them.
The thesis I first wrote about in my LinkedIn article in 2024 and that I want to expand on here is this: county, city, and state governments will slowly transition into a blended platform of blockchain and AI features that will empower citizens to decide where tax dollars are allocated and which government services and infrastructure will take priority. This is not a fantasy. The building blocks are already being laid. Wyoming is piloting a state-issued stable token for contractor payments that reduced disbursement timelines from 45 days to seconds. California’s DMV has digitized 42 million car titles on blockchain infrastructure, creating an immutable and instantly accessible record of ownership. North Carolina’s legislature is actively exploring blockchain for property taxes, voting, and public records. These are not headline stories. They are the quiet early steps of a transition that, if it continues, will change what government means within a generation.
The Platform: What It Could Actually Look Like
Imagine an application, call it the CivicLedger, for the sake of argument, that every citizen can download onto their phone. It shows, in real time, every dollar of public revenue and every dollar of public expenditure in their city, county, and state. Not in the aggregate. Not in a PDF published quarterly that nobody reads. In real time. Every contract awarded. Every vendor paid. Every department’s budget. Every infrastructure project’s status. Every grant received and disbursed. Every salary on the public payroll above a defined threshold. All entries on a blockchain, so that no entry can be altered after the fact, only appended, with a time-stamped record of who made every change.
On this same platform, citizens can vote on major spending decisions. Not every decision, the model of requiring citizen approval for every routine procurement would be paralyzing, and voters do not want to spend their evenings deciding which office supplies to buy. But major decisions, infrastructure allocations above a defined threshold, new welfare program designs, emergency spending authorizations, the terms of public-private partnerships, the approval of new bond issuances, these are decisions that citizens have both the right and, when the information is accessible and the interface is simple, the capacity to make.
The ID verification infrastructure exists. Biometric identity systems, blockchain-anchored digital credentials, and government-issued digital IDs are already operational in Estonia, which has run its entire government on a digital platform for over twenty years, with ninety-nine percent of public services available online and voter participation in digital elections exceeding seventy percent of the eligible population. Estonia is a small country. But it is proof that this is not science fiction. This is an engineering problem that a real democracy has already solved on a smaller scale.
The AI Layer: Efficiency Without Bias — If Possible
Now add artificial intelligence. Not as the decision-maker, I will return to why that specific framing frightens me, but as the analyst. An AI system with read access to every public contract, every spending record, every regulatory filing, and every procurement decision across all government departments could identify patterns of waste, fraud, and inefficiency that no human audit team could detect at the same speed or scale. This system could identify unusual contract awards when a specific vendor repeatedly wins bids, even if their prices aren’t the lowest. It could identify regulatory approvals that consistently take three times longer for applicants without established political relationships. With its ability to calculate the per-capita cost of public services, it could also benchmark against comparable jurisdictions to identify underperforming areas. It could do all of this continuously, in real time, without fatigue, without bias toward any specific political outcome, and without the ability to be lobbied.
The efficiency gains from this layer alone, without removing a single human decision-maker, would be substantial. The federal government’s own Government Accountability Office has identified hundreds of billions of dollars in annual improper payments, duplicative programs, and procurement inefficiencies across the executive branch. At the state and local level, comparable waste is proportional. A system that simply surfaced this waste clearly, attributed it specifically, and made the attribution visible to citizens through the CivicLedger platform would create accountability pressure that the current system systematically cannot generate.
Could an AI-assisted government save enough to reduce or eliminate income tax? The honest answer is: yes, possibly, but the savings would need to be enormous, and the transformation would need to be complete. The Tax Policy Center estimates that the federal government spends approximately 15 percent of its budget on administrative overhead, procurement waste, and program duplication. At the state and local level, that percentage is higher in many jurisdictions. If a well-functioning AI-assisted government reduced that overhead by half, a conservative estimate given what operational AI has achieved in private sector supply chains and financial services, the savings would run into the hundreds of billions of dollars annually at the federal level alone. Whether those savings would translate into lower taxes or higher services is a political choice. But the savings would be real, and they would be demonstrable to every citizen who could see the numbers on their phone.
The Fear: Who Runs the AI?
And here is where I stop being optimistic. Because the history of powerful technology, from the printing press to the internet to social media, is a history of tools built with idealistic intentions that were captured by interests the inventors did not foresee. And artificial intelligence, as it currently exists, is not neutral.
We have already seen AI systems trained on data that reflects the biases of their creators, producing outputs that favor certain viewpoints, certain sources, and certain political framings over others. We have seen major AI platforms, including some of the most widely used in the world, respond differently to questions about different political groups, different historical events, and different religious traditions. This is not a conspiracy. It is a documented and acknowledged limitation of how large language models are trained: they learn from human-generated data, and human-generated data reflects human biases. When those biases are systematically skewed toward any particular political, cultural, or economic worldview, even unconsciously, the AI inherits and amplifies them.
A government run on a biased AI is not a more efficient version of the government we have. It is something potentially far worse: a system with all the efficiency advantages of artificial intelligence and all the partisan distortions of human politics, but without the visibility or accountability that human political systems at least nominally provide. At least when a politician makes a corrupt decision, there is a trail: campaign contributions, meeting records, phone calls, and emails that can eventually be subpoenaed. An AI system that systematically disadvantages certain communities or policy outcomes leaves no such trail, unless the system is specifically designed for transparency from the ground up.
The question of who runs the AI is the most important question of our time, and I do not have a satisfying answer. What I know is that concentrating control of a government-level AI system in the hands of any single founder, corporation, or political faction is exactly the wrong structure, regardless of how benign their stated intentions. The history of concentrated power does not provide reasons for optimism. And I have a specific and growing concern about whether any individual or small group of people, regardless of their intelligence or their good faith, can maintain control over AI systems that are rapidly approaching and in specific domains, already exceeding the intelligence of their creators. The idea that an AI system can be permanently constrained by the values and limitations of the humans who built it becomes less plausible as those systems become more capable.
The Pattern Already Underway: A Structural Warning
I want to be direct about something that is not comfortable to raise in a book about infrastructure finance, because infrastructure finance is a business that depends on government relationships, government contracts, and government goodwill. But intellectual honesty requires saying it plainly: the scenario I find most frightening about AI-operated government is not hypothetical. A pattern is already underway, visible in publicly documented contracts and policy decisions, and it deserves frank examination from anyone who cares about where the world’s largest economy is heading.
The pattern I am describing is drawn entirely from publicly available government contract records and the public securities filings of the companies involved, documents any member of the public can access through USAspending.gov and the SEC’s EDGAR database. The pattern works as follows. A private technology company, with deep roots in the intelligence community, develops software platforms that connect government databases, databases that were never designed to be linked, that were kept separate by design, because their separation was itself a form of civil liberties protection. The software is marketed not as a surveillance tool but as an efficiency tool. It promises to eliminate waste, streamline operations, reduce fraud, and modernize legacy systems. These promises are real; the efficiency gains are genuine. But so are the capabilities the software creates as a byproduct of those gains.
When you connect the immigration database to the tax database to the law enforcement database to the healthcare database to the social services database to the financial transaction records, you do not simply create a more efficient government. You create a unified profile of every citizen that no prior government in history has possessed. The granularity of that profile, immigration status, movement history, social network connections, financial behavior, physical descriptors, medical history, and employment records make it possible to identify, locate, and act on any individual in the country with a speed and precision that human administration could never achieve. That capability is real regardless of who currently holds political power, regardless of what the software’s stated purpose is, and regardless of how benign the intentions of the people who built it.
The public record on this is not hidden. Government contracts are public documents in the United States. What they collectively show when read together, rather than in isolation, is a systematic effort by at least one major technology firm to become the operating infrastructure layer for the entire US government. Contracts spanning intelligence agencies, the military, immigration enforcement, tax administration, health agencies, and law enforcement, consolidated wherever possible into single enterprise agreements that create long-term dependencies, have together produced a situation where one private company’s proprietary software platform mediates an enormous and growing share of consequential government decisions about individuals. That company’s own public filings describe its platforms as the ‘central operating systems’ for its government customers, an aspiration stated explicitly in its public market documentation as becoming the ‘default operating system across the US.’
THE INFRASTRUCTURE FINANCE LENS
For readers of this book, there is a specific way to understand this risk that the general public rarely considers. In infrastructure finance, we care deeply about structural lock-in, the moment when a physical or contractual dependency becomes so deeply embedded that it cannot be undone without catastrophic disruption. A pipeline that serves an entire region cannot be decommissioned without replacing the energy it delivers. A data platform that integrates thirty agency databases and upon which thousands of government workflows depend cannot be replaced in a budget cycle, or a political cycle, or perhaps several of them. When a private company achieves that level of structural dependency with a government, it has effectively become infrastructure, and infrastructure, as this book has argued from its first page, is extraordinarily difficult and expensive to change once it is built. The difference between a pipeline and a government data platform is that the pipeline cannot decide what it carries.
The connection to global trade and infrastructure investment is not metaphorical. The United States is still the world’s largest economy and the anchor of the global financial system. The dollar is the reserve currency of international trade. The frameworks described throughout this book, project finance, trade finance, ECA guarantees, and commodity markets, all ultimately flow through or depend upon the stability, rule of law, and predictability of the US system. If the administrative machinery of that system is increasingly mediated by proprietary AI platforms operated by private companies with their own interests, their own political relationships, and their own data business models, then the question of who controls those platforms becomes a question about who, in a very practical sense, controls the conditions under which global trade and infrastructure investment occur. That is not a civil liberties question in the abstract. It is a question about the foundation of everything this book describes.
I am not suggesting this outcome is inevitable. I am suggesting that it is the direction the current trajectory leads if it is not actively redirected. The same researchers, investigative journalists, and civil liberties organizations that have raised these concerns in recent years have done so not to obstruct technological progress but to argue for the conditions under which that progress can be trusted. I share those concerns. And I believe that the people who build infrastructure and finance technology are precisely the people who should be paying attention to them, because the infrastructure of governance is infrastructure too, and it is being financed and built right now, largely out of public view.
The Specific Danger: Efficiency as Cover
The most insidious aspect of this pattern is that it uses the language of efficiency and waste elimination to normalize capabilities that would be politically unacceptable if described directly. Who could be opposed to streamlining government operations? Who could be opposed to connecting databases to eliminate fraud? Who could be opposed to modernizing legacy systems that have not been updated in decades? Each of these phrases describes something that sounds not only reasonable but obviously desirable. Each of them also describes a capability that, framed differently, most citizens would find deeply alarming.
Streamlining certain enforcement operations means building a system that can identify, locate, and process individuals with a speed and scale that human administration could never achieve, before legal counsel is engaged, before due process runs its course, before errors in the underlying data can be caught and corrected. Connecting disparate databases to eliminate fraud means creating a unified profile of every citizen from data sources they never consented to combine, managed by a private company with its own commercial interests in the value of that data. Modernizing legacy systems means making a private company’s proprietary platform the irreplaceable operating layer for functions that a democratic government is supposed to perform transparently and accountably on behalf of its citizens.
This is the core danger. The tools of efficient government are not neutral. They can be used to build the citizen-controlled, transparent CivicLedger I described in the sections above. They can equally be used to build something that looks like efficiency on the surface but functions as comprehensive administrative control underneath. The difference between those two outcomes is not technical. It is about who owns the data, who controls the algorithms, who can audit the decisions, and what recourse citizens have when the system makes an error or when it makes a decision that is not an error but a feature.
What Proper Oversight Would Actually Require
If AI-assisted government is coming, and I believe it is, because the efficiency gains are too large for any administration to permanently resist, then the question is not whether to accept it but on what terms. Here are the minimum conditions I believe any honest person should insist on, regardless of which political party holds power when this conversation finally becomes unavoidable.
First, no private company should own the operating system of a democratic government. The software that integrates government data, makes government decisions, and profiles government citizens should be open-source, auditable by genuinely independent third parties, and owned by the public. A proprietary platform that defines how government data is connected, categorized, and made searchable is not a trade secret; it is a map of how the government sees its citizens, and citizens have a right to see how that map is drawn.
Second, the databases that can be integrated must be legally defined and publicly disclosed before integration occurs. Every combination of government databases that a platform connects should require specific legislative authorization, not procurement approval, not administrative discretion, but a vote of elected representatives, and should be subject to audit by an authority that is financially independent of the agencies being audited.
Third, algorithmic decisions that affect individuals must be explainable in plain language and contestable through a process that a normal person can access. If a system determines that a person is a priority for enforcement action, a fraud suspect, a security risk, or ineligible for a benefit, that determination must be attributable to specific, documented data inputs and subject to challenge on those specific grounds. Opaque determinations that deprive individuals of rights or liberty are not consistent with due process under any defensible reading of constitutional law, regardless of whether the opaque actor is a human bureaucrat or a machine.
Fourth, the crowd approval mechanism I described earlier should apply explicitly to government AI deployments and data integration decisions. Any contract that gives a private company access to cross-agency citizen data should require supermajority legislative approval plus a defined public comment period. The current practice of expanding data access through procurement renewals that attract no public attention is precisely the opacity that the CivicLedger is designed to eliminate.
Fifth, AI systems used in government should be required to demonstrate neutrality through adversarial testing conducted by parties with no financial relationship to the system’s developers, not the company’s own ethics board, not the agencies whose data the system manages, but genuinely independent researchers with full access to training data, model parameters, and decision outputs, and the legal authority to publish their findings without modification.
None of these conditions is currently in place in any comprehensive form. Most of them are not being seriously debated at the level of policy that would need to adopt them. The contracts that matter most are already signed. The platforms are already embedded. The structural dependencies are already forming. That is the honest state of where we are, and why this chapter belongs in a book about financing infrastructure, because the infrastructure of governance is the foundation on which every other infrastructure decision in this book ultimately rests.
The Safeguard: Crowd Approval as a Constitutional Principle
Which brings me to the safeguard I think is most important, and which is also the one most directly connected to the themes of this book. Every major decision made by an AI-assisted government should require crowd approval, explicit, verifiable, blockchain-recorded consent from a qualified majority of affected citizens, before it is executed.
This is not a new idea. It is representative democracy, extended to its logical conclusion by the technology that makes genuine direct participation possible at scale for the first time in human history. The Athenians practiced direct democracy in the fifth century BC because their city was small enough that all citizens could gather in one place. The scale of modern nation-states made representative democracy the only feasible substitute. But representative democracy is a second-best solution, a mechanism for aggregating citizen preferences through intermediaries because direct participation was physically impossible. When physical impossibility is removed by technology, the justification for the intermediary weakens significantly.
What decisions should require crowd approval? At a minimum: any declaration of war or military engagement. Any infrastructure expenditure above a defined threshold, perhaps $500 million at the federal level, $50 million at the state level, and $5 million at the county level. Any new welfare program or significant modification to an existing one. Any tax rate change above a defined percentage. Any emergency spending authorization beyond a defined duration. Any public-private partnership in which a private entity receives exclusive access to public resources.
The technology to implement this already exists. Estonia does it. Utah’s DAO legislation provides the legal framework for blockchain-recorded collective decision-making. Wyoming’s stable token infrastructure demonstrates that government transactions can be executed on blockchain rails. The CivicLedger platform I described earlier is not technically difficult to build; it is politically difficult, because the people who would have to authorize it are precisely the people whose power it would diminish.
The Timeline: Starting at the Bottom
This is why the transformation, when it comes, will start from the bottom of the governance hierarchy rather than the top. No sitting Congress will vote to implement a system that replaces most of its members with an algorithm and citizen voting. No incumbent administration will deploy an AI transparency platform that makes its own waste visible in real time. But a city council in a mid-sized American city, facing a budget crisis and a population that does not trust it, might. A county government trying to compete for new business against a neighboring county with lower taxes and better services might. A state legislature that has watched local governments successfully pilot civic technology and is under pressure to adopt it might.
The diffusion of this technology will follow the same pattern as every major governance innovation in American history: it begins in the states and localities, the laboratories of democracy, where the stakes are manageable, and the experiments are recoverable. It proves itself through demonstrated results, lower costs, better services, higher citizen satisfaction, and measurable reduction in fraud and waste. And then, slowly, it propagates upward through the governance hierarchy as the political cost of resisting it exceeds the political cost of adopting it.
I do not know whether this will happen in ten years or fifty. I do not know whether the AI systems that enable it will be genuinely neutral or subtly distorted. I do not know whether the humans who build the platform will protect its integrity against capture or succumb to the temptation to use its power for their own ends. These are not technical questions. They are human questions, and humans have never had a particularly reliable track record of choosing wisely when power is available.
What I do know is this: the tools described in this book, the infrastructure finance frameworks, the trade finance platforms, the community equity models, the inventor financing pathways, are all tools for putting more economic power in the hands of more people. Every crowdfunding campaign that lets a community own a piece of its local infrastructure is a small experiment in the same idea. Every equity crowdfunding round that lets a garage inventor bring a technology to market without surrendering control to a single investor is a small experiment in the same idea. The CivicLedger is a larger version of the same experiment. The question is whether, as the experiments scale, the principles that animated them, transparency, distributed ownership, citizen participation, and accountability scale with them.
The Hope
Here is what I allow myself to believe, on the days when I am most optimistic. The generation that is growing up with smartphones in their hands and AI in their daily lives will not accept the opacity of nineteenth-century governance structures applied to twenty-first-century problems. They will not wait in line to vote on a machine that was designed in the 1960s. They will not trust institutions that cannot show their work. And they will build or demand that others build the systems that reflect their expectations.
The infrastructure revolution that could follow a genuinely efficient, genuinely transparent, genuinely citizen-accountable government is almost impossible to imagine in full. High-speed rail networks approved by citizen vote and funded by redirected waste savings, built at half the cost of current projects because the procurement process is clean. Energy grids modernized not over the forty-year timelines that current regulatory processes require, but on the timelines that the physics of climate change actually demands. Community-owned renewable infrastructure, financed through the same crowdfunding tools described in Chapter 17, is built in neighborhoods that currently export their energy dollars to distant utilities. A country that actually looks like the country its founders imagined, not because of the wisdom of its leaders, but because its citizens have, for the first time, the tools to hold those leaders genuinely accountable.
I am writing this at a moment when that future feels both closer and more threatened than at any point in my adult life. Closer because the technology exists. More threatened because the people who control the most powerful AI systems are also, in many cases, the people who would benefit most from those systems remaining unaccountable. The distance between those two observations is the most important frontier in human governance, and it will be crossed, one way or another, within the working lifetime of most people reading this book.
Where Does Everyone Go?
This is the question I find myself sitting with most quietly, in the hours after the analysis is done and the contracts are reviewed, and the financial models are closed. If AI runs governments, manages infrastructure, processes trade finance, underwrites projects, monitors construction, and optimizes operations, not partially, but comprehensively, then what happens to the developers? The traders? The bankers? The laborers who build the facilities, weld the pipelines, string the powerlines, and drive the heavy equipment? These are not rhetorical questions. They are the practical human consequence of the technological trajectory this entire book has been describing.
The honest answer is: they are displaced. Not all at once, and not without transition, these things never happen the way they are modeled in the anxious think-tank papers that proliferate whenever a new wave of automation arrives. But they are displaced. The infrastructure developer who today spends three years assembling a capital stack across six lender relationships in four countries will eventually find that an agentic AI system can do the same work in weeks, with fewer errors, better documentation, and no deal fatigue. The commodity trader who today relies on relationship networks, market intuition, and the ability to execute under pressure will find those edges competed away by AI systems that never sleep, never miss a price signal, and never let emotion affect a position. The independent engineer who today certifies construction progress by physically visiting project sites will find satellite imagery, digital twins, and computer vision doing the same work continuously and more comprehensively than any human can.
And below these knowledge workers sit the physical laborers, the people who actually build the world this book is about. The construction crews who pour the concrete and install the equipment. The truck drivers who move the equipment and materials. The refinery operators who run the processes. The port workers who load and unload the containers. Robotics and automation are already replacing these workers at the margins of every industry, and the replacement is accelerating. A $10 million per megawatt data center that requires months of intensive labor to build today will, within a generation, be assembled largely by robotic systems working around the clock with human supervision rather than human hands. The port that currently employs thousands of longshoremen is already measurably less labor-intensive than the port of twenty years ago, and the trajectory continues.
I want to be careful here not to simply repeat the comfortable reassurance that has been wrong in every previous version of this conversation throughout economic history: that new technology creates more jobs than it destroys, that workers displaced from one industry find their way to another, that the system equilibrates. That narrative has been broadly true for most of the industrial era; the agricultural workers who were displaced by mechanized farming did eventually find work in factories, and the factory workers displaced by automation did eventually find work in services. But the comfort of that historical pattern depends on the new opportunities arising faster than the old ones disappear, and on the new opportunities being accessible to the people being displaced rather than only to an elite with specific educational credentials and social capital. Neither of those conditions is guaranteed in an AI-driven transition, and there are serious reasons to doubt both.
Universal Basic Income: The Question We Can No Longer Defer
Which brings us to universal basic income, the idea that was once the province of academic economists and fringe political movements and has become, within the span of a decade, a mainstream policy proposal being piloted in countries from Finland to Kenya to Canada to the United States. UBI is simple in concept and enormously complex in execution: every adult citizen receives a regular, unconditional cash payment from the government, sufficient to cover basic needs, regardless of employment status, regardless of other income, regardless of behavior. No means test. No work requirement. No bureaucratic gatekeeping that consumes a third of the program’s budget before any money reaches anyone.
The question I find myself asking and that I think deserves to be asked more directly than it typically is in polite economic discussion is not whether UBI is affordable. If AI and automation eliminate the need for the majority of routine human labor, and if the productivity gains from that elimination are even a fraction as large as current projections suggest, then the economic output available to fund UBI becomes extraordinary. The question is not the money. The money, in a world of AI-managed government and AI-optimized production, will be there in a way it has never been before. The question is whether the political and social structures exist to distribute it equitably and whether the humans who receive it will know what to do with the freedom it provides.
Because that is the more interesting question. Not the economics of UBI but the anthropology of it. What do humans do when survival is guaranteed? What do they create when they are not exhausted from labor that machines can do better? What do they discover about themselves and each other when they have the time and the security to ask?
The da Vincis We Never Found
Here is the thought that I cannot let go of, that has stayed with me through the writing of this entire book, and that feels, more than anything else in it, like the seed worth planting.
Leonardo da Vinci was born in 1452 in a small Tuscan village, the illegitimate son of a notary and a peasant woman. He had access to education because his father recognized his talent and arranged for him to apprentice with a Florentine painter. That single act of recognition, one person noticing what was inside another person and creating a path for it to emerge, produced the most remarkable polymath in recorded history: painter, sculptor, architect, engineer, musician, mathematician, geologist, botanist, writer, cartographer, and inventor of machines that would not be built for four hundred years.
Nikola Tesla was born in 1856 in a Serbian village, the son of an Orthodox priest. He had access to education because his father, despite initially wanting him to join the clergy, recognized his exceptional mathematical ability and supported his studies. That single act of recognition produced the inventor of alternating current, the induction motor, the radio, the remote control, and dozens of other technologies that made the electrified world possible.
Now ask yourself the question I cannot stop asking: how many da Vincis and Tesla’s were born in that same century, in the same villages, in the same decades, carrying the same extraordinary potential, who spent their entire lives doing something else? Farming land that barely fed their families. Working in mines. Carrying loads. Tending animals. Dying of diseases that were already understood by the educated class but not accessible to them. Not because they lacked the capacity. Not because the gift was not there. But because no one recognized it, because no path existed for it, because survival consumed every hour they had.
We will never know their names. We will never know what they would have built, what they would have painted, what they would have discovered, what problems they would have solved that we are still carrying today. They are the invisible loss at the center of every economic system that has ever existed, the cost, measured not in dollars but in unrealized human potential, of organizing a society around the premise that most people’s primary purpose is to perform labor that someone else has determined is needed.
What changes if that premise is removed?
If AI and automation absorb the routine labor, the driving, the carrying, the sorting, the welding, the data entry, the compliance processing, the warehouse picking, the crop harvesting, the construction supervision and if the productivity gains from that absorption fund a universal income sufficient to cover shelter, food, healthcare, and education for every person on earth, then for the first time in human history, the answer to the question of what a person can become is not constrained by what their labor market will pay them to do. A child born in rural Mozambique with da Vinci’s visual intelligence and Tesla’s mathematical intuition would not spend her life farming because that is the only option available within walking distance of her village. She would have time. She would have access to knowledge, to tools, to a global community of others who share her interests. She would have the security to experiment, to fail, to try again.
We have never run this experiment. Not once in the history of civilization has a majority of the human population had the time, the security, and the access to develop their full potential. We have always reserved that for a small class whose labor was not required for survival, and the output of that small class, even so constrained, is the entirety of what we call civilization: every painting, every symphony, every theorem, every invention, every poem, every discovery. The cathedral builders and the shipwrights and the farmers who made that class possible are mostly unnamed. Their potential went largely unexpressed, not because it was not there but because survival left no room for it.
AI is the first technology in history with the genuine potential to change that equation. Not by making everyone’s labor unnecessary, at least not immediately, and not without enormous transition costs that must be honestly managed. But by progressively freeing larger and larger portions of the human population from the requirement to spend their most productive hours performing tasks that a machine can do better. Every hour reclaimed from necessity is an hour available for whatever it is that person was actually built to do. And in a world of eight billion people, the statistical likelihood that we have been missing thousands of da Vincis and Tesla’s is not small. It is near certainty.
Would some of those potential da Vincis have spent their reclaimed hours watching screens and doing nothing? Of course. Human nature does not transform overnight, and freedom is harder to use well than compulsion. But history suggests that when people have genuine time and genuine security, a remarkable number of them create not because they were told to, not because the market rewarded it, but because making things, understanding things, and connecting with other people are what humans, left to their own devices, actually want to do. The question of how to structure a society that converts the productivity gains of AI into the conditions for widespread human flourishing is the most important design challenge of the next century. It is harder than any engineering problem in this book. And it is, ultimately, what every financing framework, every technology deployment, and every infrastructure investment described in these pages is in service of, whether we say so explicitly or not.
The Thread That Runs Through
This book began with infrastructure, the pipelines, ports, and power grids that move everything the world makes and uses. It ends with a question about what humans do when infrastructure takes care of itself. Those two things are not as far apart as they appear. Every tool described in these pages, the project finance structure that lets a small developer access institutional capital, the equity crowdfunding platform that lets a garage inventor raise a pilot plant without surrendering to a single gatekeeper, the community ownership model that lets a neighborhood own the infrastructure it lives beside, the trade finance innovation that lets a small commodity producer access markets previously reserved for banks, every one of these tools is, at its core, a mechanism for distributing economic power more widely. For giving more people the resources to participate in building, owning, and deciding.
That impulse, more participation, more ownership, more transparency, more decision-making power in the hands of more people, is the same impulse that animates the CivicLedger platform, the crowd approval mechanism, and the argument for universal basic income. It is the same impulse that drove the JOBS Act, which let ordinary Americans invest in early-stage companies for the first time. It is the same impulse that drove the community development finance system, which directed patient capital to places that institutional investors ignored. It is the same impulse that drives every developer who structures a project so that a neighborhood can own equity in the asset being built beside them.
The world we are financing is not just a world of better infrastructure and more efficient trade. If we get this right, the AI governance, the oversight structures, the distribution of gains, the preservation of human agency in the face of automation, it is a world in which the person who would have been a great artist is not spending their life driving freight. In which the person who would have solved a problem that has cost humanity decades of suffering is not spending their life in a call center. In which the person who carries the next breakthrough in clean energy, or materials science, or medicine inside them has the time and the security, and the access to let it out.
I am writing this at a moment when all of that feels both closer and more fragile than at any point in my life. The tools exist. The financing frameworks exist. The technology exists. What does not yet exist, in sufficient quantity, is the wisdom to use them in ways that enlarge human freedom rather than constrain it. That wisdom is not a technical problem. It is a human problem, and it will be solved by people who understand both systems they are building and the people those systems are meant to serve.
That is why this book exists. Not just to explain how infrastructure gets financed. But to suggest, quietly, that the people who understand finance are not separate from the question of what kind of world gets built. They are among its most important architects. The capital goes where the financing frameworks point it. The technologies that get commercialized are the ones that find a path through the capital stack. The communities that own their assets are the ones where someone understood both the crowdfunding regulations and the CDFI system well enough to combine them. The inventors who change the world are the ones who found a platform and a legal structure that let them raise enough money to prove the idea without surrendering the vision.
None of that is inevitable. All of it requires people who understand the tools, who care about the outcomes, and who are willing to use one in service of the other.
The rest, as always, is up to us; maybe.