The Basics

What is the Won't Get Fooled Again Act?

It is a plan for the next financial crisis, written before it arrives. When a bank or a major AI company fails, the government does not bail it out and does not sell it to a bigger rival. It converts it. A failed bank of any size becomes a bank with no shareholders that lends to its community, owned by its depositors, by its state or city, or, for the largest, by the public. Failed AI labs become public research institutes, and failed data centers become a public computing utility. Shareholders and lenders take the losses, the industries that created the risk pay the costs, and the public and the banks' own depositors end up owning working infrastructure.

Why the name?

Because the public was told in 2008 that the bailouts were a one-time emergency, and then watched it happen again. The banks that were rescued came out larger. In 2023 three more banks failed, every large depositor was covered over a weekend with no premium ever having been paid for that coverage, and First Republic was sold to JPMorgan Chase, already the largest bank in the country. The Act exists so that the next crisis does not end the same way.

Does the plan assume a crash is coming?

We think one is likely, and the full plan lays out why: AI spending far ahead of AI revenue, companies investing in their own customers, debt hidden in shell companies, and loans secured by chips that lose value fast. But the Act does not depend on the prediction. It is insurance. If no crash comes, no institution is converted, and what remains is a safer deposit system, a clear record of who controls the nation's computing power, and an industry paying premiums for the risk it creates.

Does any of this apply to healthy companies?

The conversion powers do not. A company can be placed in receivership only if it has actually failed. A bankruptcy filing, a large payment left unpaid when its grace period runs out, or taking emergency federal support counts automatically. Short of that, regulators must find that it cannot pay its debts or owes more than it owns. A court can review either within 24 hours. What does apply to healthy companies is modest: the largest banks and technology firms pay premiums and levies, large clusters of AI chips must be registered, and the biggest firms cannot buy distressed computing assets.

How far along is this proposal?

It is a beta. The mechanisms are worked out, but many of the specific numbers, such as the size thresholds and the levy rates, are our best current judgment and are marked for further work. The legal design needs review by bankruptcy and constitutional lawyers before it could be introduced as legislation. We are publishing it now because a plan like this has to exist before the crisis, not be improvised during one.

Who Pays

How is this different from a bailout?

A bailout rescues the people who made the bets. It keeps shareholders in place, makes lenders whole, and usually leaves management in their jobs. The Act does the opposite on all three counts. Shareholders lose everything. Lenders get the current value of their collateral and lose the rest. Executives are removed and their pay is clawed back. What is rescued is the service, the jobs, and the customers, not the investors.

What stops the government from rescuing a firm anyway?

The Act itself. Any bank, and any systemically important technology company, that takes emergency federal support is treated as having failed: its shareholders are wiped out, its lenders are paid what their collateral is worth, and it is converted. The rule binds every federal agency and the Federal Reserve by name. Ordinary lending to sound banks against good collateral is not affected.

Does any of the cost fall on taxpayers?

The design is that they do not. The costs are paid by a tax on high-end chips and large training runs, by premiums from the largest technology companies, and by a surcharge on the largest banks. Because those funds start out small, they can borrow from the Treasury in a crisis and repay from future levies. This is how the FDIC is backstopped today. So public money is at risk for a period, and we say so plainly. If the levies were later repealed or set too low, taxpayers could be left with a loss. What the public buys, it buys at the market's price on the day of failure, and owns.

How much could it cost?

Nobody can know in advance. Our rough estimate is that in a serious crash the debts of failed AI-related companies could come to $150 billion to $285 billion. Under the Act the public does not pay those debts. It pays what the underlying assets are worth and receives assets of that value. If used chips lose most of their value in the crash, that might be $50 billion to $100 billion. If they hold their value better, it could be roughly double. The public cannot choose the lower figure: it pays what the market shows the assets to be worth. For scale, Congress authorized $700 billion for the 2008 bank rescue, later reduced to $475 billion. These figures are estimates and the full plan marks them as such. On the banking side, the FDIC's records give the scale: the failures of 2023 would have needed about $43 billion to $53 billion of capital, and those of 2008 through 2013 about $55 billion to $69 billion over six years. That capital stays inside working banks, and if a giant fails, its own loss-absorbing debt comes first.

Will the levies raise prices for customers?

Partly, yes. That is true of deposit insurance premiums too, and it is how insurance is supposed to work: the price of a risky activity should include the cost of the risk. Small-scale research and educational computing are exempt from the compute tax. And the alternative is not free. It is a bailout paid for by everyone, including people who never used the product.

The Lenders

Why do lenders receive anything when shareholders receive nothing?

Two reasons. The first is legal. A lender with a lien holds a property right, and the Constitution does not let the government take property without paying for it. The second is that the law asks for less than people assume. What a secured lender is owed is the value of the collateral on the day of failure, not the amount of the loan. A fund that lent $10 billion against chips now worth $3 billion gets $3 billion. The other $7 billion becomes an unsecured claim that will recover little or nothing. That is a real loss, and it is the point.

Will this make it more expensive to borrow for AI infrastructure?

Yes, somewhat. Loans for speculative build-outs will carry higher interest and stricter terms once lenders know that nobody will make them whole. That is intended. Cheap credit with an implied government guarantee is how bubbles are inflated. Companies with real revenue will still be able to borrow.

Could losses for lenders spread to the rest of the financial system?

This was the reason creditors were made whole in 2008, and the concern is a real one. The lenders are banks, credit funds, pension funds, and insurers, and their losses have to land somewhere. The Act's answer is to deal with each of those directly. A bank that fails because of these losses is itself converted. Pensioners are protected through a separate facility. Insurance policyholders are covered by the state guaranty funds that already exist. What the Act does not do is protect them by paying off the lenders, because that also rescues every other investor in the same funds.

How are retirees and pension funds protected?

This is one of the most important questions, and the plan addresses it directly. Pension funds rarely lend directly. They invest through private credit funds that also hold money from sovereign wealth funds, wealthy individuals, and insurers, so rescuing the loans rescues everyone. Instead, the Act creates a Pension Protection Facility, paid for by the industry levies, that makes long-term loans or grants to pension plans with documented losses. It protects retirees' benefits, not fund balances, and comes with conditions on future risk-taking. It is aimed mainly at state and local plans, which have no federal insurance. Congress did something similar for failing union pension plans in 2021.

Who gets paid ahead of the lenders?

A lender with collateral is entitled to that collateral's value, because a claim on collateral is a property right that the Constitution protects. It receives that value as the market sets it on the day of failure, which may be far less than it lent, and nothing more. The government does not cover lenders' losses. For everything else a failed firm owes, the Act sets the order. Employees come first, for everything they have earned. Then small suppliers, up to a cap. Customers, whose contracts and prepaid credits carry over to the new public entity. Then local governments and utilities. Banks, bondholders, and investment funds come after all of them for whatever their collateral did not cover. The order is written into the law in advance, so that lenders know it when they lend and nobody can accuse the government of picking favorites after the fact.

Public Ownership

Is public ownership like this unusual in the United States?

No. The plan calls for public ownership of institutions that have failed and that the country cannot do without, and the United States has done this many times: the Tennessee Valley Authority, the conversion of the bankrupt northeastern railroads into Conrail, the Bank of North Dakota, which has operated since 1919, and every weekend that the FDIC takes over a failed bank. Ownership by depositors is older still: mutual savings banks and credit unions have belonged to the people who bank with them for generations. The aim is a system in which the people who take the gains also take the losses.

Who would run a converted bank or data center?

It would not be run by a federal office. A converted bank keeps its staff and branches. A local or regional bank is governed by a board its depositors elect or, where a state or city takes ownership, by a board drawn from the places it serves, on the model of Germany's public savings banks, which kept lending through 2008 while private banks pulled back. A large national bank gets a board accountable to the public at large, with seats for its employees, its customers, and the regions where it operates. A converted technology firm keeps its engineers, gets an experienced outside chief executive from a roster vetted in advance, and can call on the Department of Energy's national laboratories, which already operate the largest computers in the country. Public enterprises can be run well or badly, as private ones can. The plan keeps the people who know the operation and changes who they answer to.

What protects the public institutions from politics?

This is a real risk, and the plan is designed around it. Public Benefit Banks are barred from speculation and limited to ordinary lending. Most converted banks belong to their depositors, who elect the board, so no official appoints it and no administration can remove it. Where boards are publicly appointed, at the largest banks and the technology institutions, they mix public appointees, employees, customers, and technical experts, so that no single interest controls them. The computing utility must offer non-discriminatory access at regulated prices and publish what it does. No design makes political influence impossible. Since the Supreme Court's 2026 ruling in Trump v. Slaughter, Congress may not be able to stop a President from replacing the directors of the federal institutions, so the plan does not rely on their tenure. It relies on what the law fixes: narrow mandates, a bar on privatizing without an Act of Congress, ordinary supervision, and public reporting. A hostile administration could still run these institutions badly. The plan is written for one that wants them to work.

Why does the plan apply only to companies that have failed?

Some people we respect would go further than this plan does. We limit conversion to failed firms for three reasons. It is on firm constitutional ground, because the owners of a failed firm have already lost their investment and nothing is being taken from them. It is fair to the public, which pays only what the assets are worth on the day of failure and owns what it pays for. And it is politically possible in the one moment when large changes can be made, which is a crisis. The Act does not rule out doing more. It makes sure that when the opportunity comes, the public is not once again handed the bill and nothing else.

Why is some of the hardware sold to private buyers?

There is no ceiling on what the public may take, and it gets the first pick of everything. But a crash means there are more chips than customers, and idle machines burn electricity and age quickly. What the public cannot put to use within 30 days must be offered to smaller private buyers. Cheap hardware in the hands of mid-size companies, universities, and startups builds competitors to the dominant firms, which serves the same goal as the public utility.

What are the Community Innovation Hubs?

They are buildings: community workspaces where residents who meet simple criteria get a free membership and can come to work, take classes, get advice, and meet other people who are starting things. Places like this already exist across the country. The efactory in Springfield, Missouri, is one example, and its clients have created roughly 3,000 jobs. The Act funds many more of them, and funds existing ones to grow, with at least one in every congressional district. They are paid for by a 1% fee on the largest AI and cloud companies, on the model of the fees that cable companies paid to fund public access television. They start when the Act passes and do not depend on a crash or on any public computing facility.

What happens to the workers?

They keep their jobs. Converted institutions are prohibited from mass layoffs, workers hold seats on the boards, and wages and benefits owed are paid first. Because much of the pay at AI companies is stock that a failure makes worthless, the agency pays retention bonuses from the first day so that the people who keep services running do not leave.

What does this mean for AI safety?

A public institute is free of the pressure to ship products before they are tested, and the plan directs it toward safety and interpretability research. That is an advantage, not a guarantee. The Act is a plan for resolving failures. It is not a safety regime for what AI systems do, and it says so. That is a separate and larger question that deserves its own law.

What happens to a failed lab's models and data?

They are frozen the moment the receiver is appointed and kept in public hands. The plan gives access to academic and nonprofit researchers under strict safety protocols. It does not publish model weights to the world. The freeze also means a failed firm's models and data cannot be deleted, sold to data brokers, or bought by a foreign government in a bankruptcy auction, which is what could happen today.

Monopoly and the Chip Rules

How does this keep the biggest technology companies from getting bigger?

In a crash, the natural buyers of failed companies' chips and data centers are the three or four firms that already dominate computing. The Act bars any company that controls more than 5% of the nation's AI computing capacity from buying those assets out of a bankruptcy or foreclosure. It closes the two obvious ways around that: such a company may not lease chips that someone else bought in a distressed sale, and may not buy a firm that has warned regulators it is close to failing. The rule is not a limit on size. A capped company can still buy as many new chips as it likes. The public gets the first option on the distressed assets, and everything else goes to smaller buyers. The banking equivalent already exists, a rule barring any bank from buying another if that would leave it with more than 10% of the nation's deposits, but it has an exception for buying failed banks, which is exactly when it matters. Ours has no exception.

How does this relate to antitrust?

The two work together, and nothing here stands in the way of antitrust enforcement. This Act does a narrower job. It stops a crisis from concentrating the industry further, and it builds a public competitor. Antitrust cases take years. A fire sale takes weeks.

What keeps the public utility from becoming too dominant?

It must offer access to everyone on equal terms at regulated prices, which is what separates a utility from a monopolist. The rule against hoarding forces idle capacity back onto the market. And it will be competing with some of the largest and best-funded companies in the world, which are left intact.

What is the chip registry, and why is it needed?

It is a record of who owns each large cluster of advanced AI chips, where it is, and who has lent against it. The model is the federal aircraft registry, which has recorded the owners of and the liens on every plane in the country for decades without slowing the sale of one. Without it the government cannot know who controls the nation's computing power, cannot enforce the 5% rule, and cannot tell where chips go when a failed company is liquidated. Today, once a chip is sold inside the United States, no law requires anyone to know where it is.

Who has to register or get a license?

Only the owners and operators of very large clusters: 1,000 or more advanced AI accelerators, defined by computing power rather than by maker, so that chips a company designs for itself count the same as chips it buys. Below that line nothing applies. Researchers and small companies need no permission to do their work, and gaming cards, workstations, university labs, and small company clusters are never touched. Above it, the license must be issued to anyone who meets published security standards, and the agency does not review what is run on the chips.

How do the chip rules stay fast and predictable?

The rules are designed to be simple. They are bright lines published in advance. Transfers go through automatically after 30 days unless the agency objects in writing. The agency is funded by the compute levy, not by hourly fees, so it gains nothing from a long review. Its mandate includes widening access to computing, not only preventing harm. Real discretion exists in one situation only, the sale of a failed firm's assets.

How does this fit with existing export controls?

The two do different jobs. Export controls already stop legal sales of advanced chips to China and other countries of concern, and the Act adds no new export rules. What export controls do not stop is smuggling, which already happens through shell companies, false paperwork, and shipment through third countries. A fire sale, with thousands of chips sold in small lots by liquidators who only want cash, would be the ideal market for it. Under the Act, chips from registered clusters are tracked by serial number through every resale. A paper trail will not stop a determined smuggler, but it will show whose chips went missing, which today nobody could say.

Deposits and Banks

Why insure every deposit?

The main reason is that a limit on deposit insurance does not work the way it is supposed to. Everyone knows the government will never let depositors at the largest banks lose money. So at the first sign of trouble, large depositors move their money from smaller banks to the giants. In the week after Silicon Valley Bank failed, about $120 billion left smaller banks, and tens of billions of dollars arrived at the largest ones. A limit therefore rescues the biggest banks by accident, whatever their condition, and drains their sounder competitors. Insuring every deposit at every bank ends that. It also means a converted public bank keeps its depositors, which it needs in order to function.

On risk-taking: a common concern is that full insurance removes depositors' reason to watch their bank. In practice depositors have not been good at policing banks. Silicon Valley Bank's customers included some of the most sophisticated investors in the country, and they noticed nothing until they all ran on the same day. Under the Act, discipline falls on shareholders, bondholders, and executives, who are paid to bear risk and lose everything when a bank fails. Banks pay risk-based premiums for the new coverage, and the largest pay the most.

What did the FDIC conclude when it studied this?

After the 2023 failures the FDIC studied three options. It found that unlimited coverage would deal most directly with bank runs, but it preferred a narrower option, covering business payment accounts, because of the cost to the insurance fund and the effect on risk-taking. We read the same evidence and come down differently, for the reason given above: anything short of full coverage keeps pushing money toward the largest banks. We agree with the FDIC that full coverage has to be paid for, and the plan has banks pay for it.

Which banks are converted?

Every failed bank that is still a functioning bank, whatever its size. Converting only the largest banks would leave small and mid-size failures to be sold in the usual way, usually to a bigger bank. A small town's bank matters as much to its customers as a giant does to the country, and the giants should not get better treatment. The FDIC can decline to convert a bank only by finding, in writing and in public, that it has no real base of customer deposits, no real lending business, or failed mainly because of fraud. In that case its accounts and sound loans go to a nearby converted bank, credit union, or community bank. They are never sold to one of the largest banks.

Who owns a converted bank?

It depends on the bank's size. A local or regional bank becomes the property of its depositors: its business moves into a newly chartered mutual bank, each depositor has one vote, and the depositors elect the board. Mutual savings banks and credit unions have worked this way in America for generations. A state or a city can take ownership instead, as North Dakota owns its bank, or its accounts and loans can go to a local credit union. Only the largest banks, those with more than $250 billion in assets, become federal public corporations, because no single community can speak for a bank with customers in every state. In every case there are no shareholders.

What happens to my account if my bank is converted?

Nothing you would notice. The FDIC already resolves failed banks over a weekend. On Monday your account, your card, and your loan work as they did on Friday, and the bank keeps its name. What changes over time is what the bank does with its money. And if it is a local or regional bank, you become one of its owners, with a vote for its board.

What would a Public Benefit Bank do differently?

It takes deposits and makes loans to local businesses, housing, and infrastructure. It is barred from proprietary trading, derivatives speculation, and lending to shadow banks. Its leadership answers to its depositors and the communities it serves, not to shareholders. The Bank of North Dakota has worked this way for more than a century, has never needed a rescue, earns a higher return on its assets than Goldman Sachs, and returns its profits to the state.

Markets, Innovation, and China

What does this mean for AI innovation?

Nothing in the Act limits what healthy companies may research or sell. Speculative borrowing will cost more, so some projects that depend on cheap credit may not be built. After a crash, the choice is not between this plan and a thriving private industry. It is between this plan and a fire sale in which the technology is broken up, shut down, or absorbed by a few giants. The Act keeps the researchers employed, the machines running, and access to computing power wider than it was before.

What does this mean for competition with China?

In a crash, the danger to the American position is that capacity goes dark, research teams scatter, and models or chips are sold to the highest bidder, who may be foreign. The Act prevents each of those. It keeps capacity running, retains the staff, freezes models and data against sale, puts every distressed sale through national security review, and tracks chips through resale. Keeping that infrastructure intact through a downturn is what protects the country's position.

Why isn't ordinary bankruptcy enough for these companies?

For most companies that is the right answer, and the Act leaves it alone. Bankruptcy has one job, which is to get creditors as much money as possible. For a few institutions that is the wrong goal. A liquidation trustee has no duty to keep a hospital's AI systems running, no reason to care whether model weights go to a foreign buyer or customer data to a broker, and every reason to sell the hardware to whichever giant bids most. Congress has long recognized this. It is why banks do not go through ordinary bankruptcy either.

How does the plan avoid picking winners and losers?

The rules are written before anyone knows who will fail. The thresholds for which firms count as systemically important are numbers set in advance. The order in which creditors are paid is fixed in the statute. The price the public pays is set by open bidding, with a published formula as the floor. What the plan rules out is deciding in the middle of a crisis which creditors should be saved.

Could passing the Act unsettle markets?

Passing it would reprice risk. Lenders who had been counting on a rescue would charge more or pull back, and some overextended companies would struggle sooner. We think gradual repricing is better than the alternative. Risk that is mispriced does not disappear; it builds until it fails all at once.

The Law and the Limits on Government

What is the constitutional basis for converting a failed company?

Congress's power to write special rules for failed institutions, which it has used for banks for 90 years. Converting a failed company is constitutional under four conditions, and the plan is built to meet them. The firm must actually have failed. Secured lenders must receive the value of their collateral. Every creditor must get at least what a liquidation would have paid. And courts must be able to review both the receivership and the price. In the 1970s Congress transferred the assets of the bankrupt northeastern railroads to a government-created company, and the Supreme Court upheld it. The plan also deliberately does not rely on the Defense Production Act. When President Truman seized the steel mills without clear authority from Congress, the Court reversed him. The power here comes from the statute itself.

Is the price the public pays fair to owners and lenders?

The owners of a failed company hold shares that the market has already made worthless. Nothing is taken from them. Lenders are paid what their collateral would actually fetch: the assets are offered to every eligible buyer, the public can take them only by matching the best bid, and it never pays less than a published formula price. Even that is not the last word: a creditor who believes a liquidation would have paid more can go to court, and the courts decide. The railroad precedent shows this is a real right. After years of litigation the government paid the Penn Central estate about $2.1 billion.

Who decides that a company is "systemically important"?

A council of existing regulators, applying numeric thresholds written in advance: a share of the nation's computing capacity, revenue and user counts, the number of hospitals, utilities, and agencies that depend on the firm, and how much it owes to the financial system. A company can contest its designation and must be answered in writing. We learned from the MetLife case, in which a court threw out a designation because regulators had not explained their reasoning.

Who decides that a company has failed?

For banks, nothing changes: the bank's regulator closes it and appoints the FDIC, as happens today. For a technology company, three events count as failure automatically, with no vote: it files for bankruptcy, it leaves a payment of $100 million or more unpaid when the grace period in its own contract runs out (30 days at most), or it takes emergency federal support. Regulators can also act earlier, when a firm is about to fail but has not yet crossed one of those lines. That is a judgment, so no single official can make it. It takes a written recommendation from two-thirds of the FDIC-Tech's board and two-thirds of the Federal Reserve Board, and a decision by the Treasury Secretary. Either way the company can go to court, which has 24 hours to rule. If it does not, the receivership goes ahead.

Does the chip registry give the government control over computing?

It records who owns very large clusters of export-controlled chips and who has lent against them. It does not record what anyone computes, and the agency has no authority to review it. Ownership of aircraft, ships, broadcast licenses, and nuclear facilities has been recorded this way for generations. We would rather have a narrow rule in place, with bright lines and hard limits on discretion, than have Congress write a sweeping one in the middle of a panic.

What keeps the agency's role from expanding over time?

That is a reasonable concern, and the plan includes limits for it. The conversion power applies only to failed firms, with a court looking over the agency's shoulder. The registry has a fixed size threshold. Transfers clear automatically on a deadline. The agency has no fee income to protect. Beyond that, the honest answer is that every law depends on Congress continuing to watch how it is used.

How is this different from the resolution authority in Dodd-Frank?

It borrows from it: the 24-hour court review, the rule that shareholders and creditors bear the losses, and the Treasury credit line repaid by the industry. There are two differences. Dodd-Frank covers only financial companies, and this extends to the technology companies the economy now depends on. And Dodd-Frank aims to wind a failed firm down and sell the pieces. This Act keeps the institution and converts it to public purpose. Dodd-Frank's authority has gone unused in part because regulators in a crisis reach for a rescue or a merger instead. This Act takes those options away.

If the Crash Comes First

What if the crash comes before this becomes law?

Then there is a second version of the plan, written for a crash already under way. The goals are the same. What changes is everything that depended on preparation: the funds are empty, there is no registry of who owns which chips, no firm has filed a plan for its own conversion, and the lenders made their loans expecting a rescue. The emergency version is a complete plan, not an outline, so that it is ready when it is needed.

How does the emergency version work?

It leans on a rule that is in both versions: any firm that needs public money to survive has failed. In 2008, the entire U.S. financial system failed. Our biggest banks were going to disappear without public support. Our government authorized $700 billion in direct bailouts, plus a whole range of other support, with almost no conditions. Under the Act, help comes only on its terms: shareholders wiped out, lenders paid what their collateral is worth, executives' pay returned, and the institution converted. That rule applies to every federal agency, including the Federal Reserve. Ordinary lending to sound banks against good collateral continues as usual.

What changes?

The whole cost is borrowed from the Treasury up front and repaid by the industry afterward. The FDIC acts as receiver for technology companies from the first day, because a new agency cannot be built during a panic. The dominant firms barred from buying distressed chips and data centers are identified in the bill, because there is no registry yet to measure market shares. The rules take effect on the day the bill is introduced, and sales to the dominant firms since the crisis began can be unwound. Failed companies already in bankruptcy court are moved into the new process, as Congress did with the bankrupt railroads in the 1970s. And every bank deposit is guaranteed at once by Act of Congress, with the cost billed to the banks afterward.

Why does waiting cost more?

Some data centers will already have gone dark and their staff will have left, and a dark site is much harder to restart than a live one is to keep running. Some of the best hardware will already have been sold. Lenders who never expected losses will have lent more, against worse collateral. Their lawsuits will be stronger, because the rules changed after they lent. And the public has to advance all of the money before the industry has paid in anything. None of that makes the emergency version unworkable. It is the reason to pass the Act in advance.

Why is the emergency version a full-length plan and not a short emergency bill?

Because of what happened in 2008. The first draft of the bank rescue was three pages long. It asked for $700 billion and said the Treasury Secretary's decisions could not be reviewed by any court. Congress read it as a blank check and voted it down, and the bill that passed days later was hundreds of pages written in a hurry, much of it by the industry. In a crisis, whoever has a finished text ready has the advantage. The parts of the emergency bill that must work immediately take effect at once, and the parts that take time to build, such as the chip registry, phase in over the following year.

Politics

What would it take to pass this?

It would be difficult in ordinary times. The largest banks, technology companies, and lenders have strong reasons to oppose it. It becomes possible in the window that opens when a crisis hits and the public is asked, once again, to pay for a rescue. That window is short. In 2008 the only plan on the table was the bailout, so the bailout is what passed. The purpose of writing this now is to have a worked-out alternative ready when the moment comes.

Who is this for?

Anyone who wants the next crisis handled differently from the last two. That includes community bankers squeezed by consolidation, smaller technology companies that cannot match the giants' computing budgets, researchers priced out of their own field, conservatives who believe that failure should have consequences, and working people who have already paid for two rescues.

What if the crash never comes?

Then no bank and no company is ever converted, and the country will have paid a modest insurance premium. It will still have a deposit system that no longer favors the largest banks, a record of who controls its computing power, Community Innovation Hubs in every congressional district, and a standing rule that the next failure, whenever it comes, will not be paid for by the public.