How We Got Here

As I documented over the years — in my many conversations with Silicon Valley executives, and in my film Forever Peace Now — the people running the technology sector are not concerned with the betterment of humanity. They think about individual financial gains, about "winning" at any cost, and about market dominance, regardless of the human consequences. I watched this up close. I was inside it. One conversation stays with me above all others. After leaving Amazon Web Services, I was in the job market and interviewing at a tech company whose name I will let the reader imagine. The Senior Vice President of Marketing told me, with what I can only describe as pride, that the U.S. Air Force was one of their most valuable reference customers — because the company's technology was helping them kill and target people more effectively. He said it the way a man talks about winning a sales award. Near the end of the interview, he gave me the floor. Ask anything, he said. I asked a simple question: as a company, where do you draw the line on customers and projects? Do you have a red line? He looked at me as though I had spoken in a language he didn't recognize. What do you mean, he said. We don't turn away business. If it's a registered business, we'd love to have them as a client. Silicon Valley has no red lines. They will lobby for any contract, pursue any customer, and sell their souls to any buyer willing to sign a check. This is not an opinion. It is a business model.

Who I Was When I Walked Into Those Rooms I want to stop here and tell you who I was when I first walked into those boardrooms. Not who I became. Who I was. I was twenty-two years old. I had come from a family of six living in a three-hundred square foot apartment in Berkeley, California. We were not poor in the way that word gets used as an abstraction — we were poor in the way that means six human beings sharing a space smaller than most American living rooms, making it work because that is what you do when you have no alternative and you are grateful to be alive and in a country that is not actively bombing you. I had graduated high school at sixteen, gone to community college, started a company before most of my peers had their first real job, and by twenty-two I was making a quarter million dollars a year in tech sales. I sat in boardrooms with men who ran companies worth billions of dollars. I raised capital. I negotiated contracts with nine figures attached to them. I wanted to be a good American capitalist. I say that without irony and without apology — because I think it is important to be honest about who you were before you understood what the system actually was. I came from Iran, from a war, from a country where the economy had been strangled by sanctions and ideology and the kind of revolutionary fervor that burns everything down and calls the ash freedom. I came to the United States and found that a person could build something, could earn something, could sit at a table with powerful people and be heard. I wanted that. I chased it. I was good at it. And then I had stock options. Founders in Silicon Valley receive common shares. This sounds like a good thing until you understand what it means in practice. You have the idea. You assemble the team. You work the hours that no employment contract could capture — the midnight calls, the weekend pivots, the months when the product is broken and the money is almost gone and you are the

only person in the room who cannot afford to lose faith in what you are building. You do all of that. And in return, you receive common shares — the lowest class of equity in the capital structure, the last in line when anything is distributed, the first to be diluted. The investor who writes a check in a Series A round gets preferred shares. Not common shares. Preferred. A different class entirely, with their own veto rights, their own liquidation preferences, their own protections that can be written into each subsequent funding round in whatever configuration their lawyers negotiate. The founder who had the idea and built the thing gets common shares. The investor who showed up with a checkbook gets preferred shares and a seat at the table that the founder cannot remove even when the investor's judgment is wrong. This is the foundational lie of Silicon Valley's meritocracy narrative. The system is not designed to reward the people who build things. It is designed to protect the people who fund things. Those are not the same people. And the gap between them — the gap between common and preferred, between the founder who cannot sell and the investor who can — is where most of the wealth that should flow to builders gets redirected to capital. I found this out the hard way. I had stock options in my company — what I call funny money, because at the time there was no secondary market for founder shares. I could not sell them to investors for cash. I could not convert them to anything liquid. I had paper wealth and a vision and the particular kind of poverty that comes from having something that looks like money but cannot be spent. What I wanted to do with those options — if they ever converted to something real — was donate a large portion to Doctors Without Borders. I mean that literally. I was twenty-two years old, I had come from a family where six people lived in three hundred square feet, and my first thought about what to do with hypothetical wealth was to give a significant share of it to the people providing medical care in war zones. I remembered what my mother had seen as a nurse during the Iran-Iraq war. I remembered the hospital corridors. I couldn't have articulated it as a political position at the time. It was just what felt right to a kid who understood, from personal experience, that the distance between having medical care and not having it was the distance between living and dying. The system did not accommodate that impulse. The system was not built for it. That is the moment I began to understand something that would take me another twenty years to fully articulate: the system was not broken. It was working exactly as designed. It was designed to concentrate capital in the hands of the people who already had it, to protect their preferred position in every subsequent round, to ensure that the funny money flowed upward — and that the founders and the workers and the communities that made the whole enterprise possible were positioned, legally and structurally and by deliberate design, to receive what was left after the preferred class had taken what they were owed. I was in boardrooms with these men. I watched them operate. I know how some of them acquired their wealth. I know how certain domain names were obtained at a time when the process governing them should have prevented it — when what was supposed to be an open, rules-based system for how the internet organized itself turned out to have rooms where different rules applied to the people who knew which doors to knock on. I know the mechanics of the deals I am about to describe not from reading about them but from having been present when the shape of those arrangements was discussed. I am not going to name every person I know from that world. I am too young to die, and I understand that the people who expose powerful men too specifically sometimes find that the legal system — the same legal system that gave Jeffrey

Epstein thirteen months for crimes that sent ordinary people away for decades — can be weaponized against the person doing the exposing rather than the person being exposed. But I will say this: I know who these men are. I know what they value and how they talk about people who are not in the room. I know what power looks like up close when you are twenty-two years old and hungry and sitting across the table from it. And what I can tell you is that the mythology — the visionary founders, the meritocracy, the disruption in service of humanity — is a story they tell in public. What they tell each other in private is something different. What they have built is something different. And the distance between the public story and the private one is the distance this entire book is trying to map. I got out. I chose something different. I am writing this book instead of cashing checks from the people I could be defending. That choice cost me money. It cost me relationships. It cost me the kind of professional comfort that comes from staying inside the network and keeping your head down and your opinions private. I would make it again tomorrow.

The PayPal Mafia I want to stop and define my terms — because I have been using two phrases throughout this book that deserve more than a passing reference, and because the people those phrases describe have spent considerable resources ensuring that the public understanding of who they are remains flattering. The first phrase is the PayPal Mafia. Silicon Valley uses this term proudly. It appears in business magazines as a celebration — a group of extraordinarily successful founders who met at a startup called PayPal in the late 1990s, built it into something valuable, sold it to eBay in 2002 for $1.5 billion, and then scattered across the landscape to build Tesla, LinkedIn, Palantir, YouTube, Yelp, and a dozen other companies that have shaped the modern world. The narrative is one of brilliance and ambition and the compounding returns of great networks. I never thought the word mafia was something anyone should be proud of. A mafia is not a network of ambitious people who know each other. A mafia is a criminal organization — a group of men who operate outside the law while presenting a legitimate face to the world, who use their collective power to bend rules that apply to everyone else, and who protect each other from the consequences of what they do. The fact that Silicon Valley adopted the term and made it a brand tells you everything you need to know about how these men see themselves in relation to the law and to the rest of us. The core of the PayPal Mafia as I use it in this book includes Peter Thiel, who co-founded PayPal and went on to build Palantir — the surveillance and targeting company whose tools I have documented operating in Gaza, in ICE deportation operations, and in the NHS health records of 67 million British citizens. It includes Elon Musk, who came into PayPal through the merger of his company X.com, and who now controls the world's largest social media platform, a private space company, a government efficiency operation, and the AI tool I documented producing antisemitic content and Holocaust denial. It includes Reid Hoffman, co-founder of LinkedIn, early PayPal executive, and a man whose platform I documented operating a covert surveillance system scanning 6,167 browser extensions without user knowledge or consent. It includes the network of investors, founders, and operators who move in and out of these companies through Founders Fund — Thiel's venture capital firm — and through the interlocking board memberships and investment relationships that connect these companies to each other in ways that are, on paper, separate transactions and, in practice, a coordinated system of mutual enrichment. And here is how that

system works — because it is important to understand that when I call them a mafia, I am not speaking metaphorically.

The Lawyers Who Were Never Yours

I want to talk about the people you hired to protect you. You walked into a funding negotiation and you hired a lawyer. That is what you are supposed to do. You paid them — sometimes five hundred dollars an hour, sometimes a thousand, sometimes more, depending on the firm and the partner and the prestige of the address on Sand Hill Road. You signed their engagement letter. You assumed, because you were paying them and because that is what engaging a lawyer means in every other context in American life, that they were representing your interests. Let me be direct about something that took me longer than it should have to fully understand: in most cases, they were not. The law firms that dominate Silicon Valley startup work — Wilson Sonsini Goodrich and Rosati, Cooley, Orrick, Fenwick and West and the others whose names you see on every term sheet and every funding announcement — do not primarily serve founders. They serve the ecosystem. And the ecosystem's center of gravity is not the founder. It is the venture capital firm. The VC is the repeat client. The VC is the reliable pipeline of deals, year after year, fund after fund, portfolio company after portfolio company. The founder is a transaction. The VC is a relationship. Think about what that means in practice when you sit down to negotiate your term sheet. The lawyer across the table from you — the one you are paying by the hour to fight for your interests — has known the partner at that VC firm for years. They have been to the same holiday parties. Their firms collaborate on dozens of deals simultaneously. The VC sends referrals. The law firm sends favorable terms. Not in writing. Not in any way that could be called corruption in a court of law. In the way that relationships work when money and proximity create alignment of interest over years. There is a phrase in that world that you will hear often. Market standard. When you push back on a term in a term sheet — the liquidation preference, the anti-dilution clause, the vesting acceleration, the board composition — you will be told it is market standard. Your lawyer will often be the one telling you this. What market standard means, in practice, is that this term has been used in enough deals, drafted by enough firms representing enough venture capital funds, that it has become the template. The template was not designed by founders. It was designed by the people negotiating across the table from founders. Market standard is the result of decades of the same law firms writing the same boilerplate on behalf of the same investors — and then telling you that the boilerplate is neutral. It is not neutral. It is the accumulated sediment of a thousand deals in which the lawyers were closer to the money than to the people who built the thing. Wilson Sonsini is the oldest and most prominent example of how deep this goes. The firm created an investment fund in 1978 — WS Investments — specifically so that its partners could take equity stakes in the client companies they were advising. They turned a seventy-two-thousand-dollar investment in Google into a twenty-eight-million-dollar return after the IPO. The same lawyers advising founders on how much to disclose in their IPO filings had a direct financial stake in maximizing the IPO price. A University of Illinois law professor put the conflict plainly: the more you disclose, the lower the IPO price. That creates a conflict of interest because if you have an economic interest in the highest IPO price, will you really advise them to disclose? The answer, which the professor left

tactfully unspoken, is: probably not. Not fully. Not in every case where disclosure might cost the fund money. The firm’s name partner, Larry Sonsini, sat on the board of Brocade Communications while Wilson Sonsini served as Brocade’s outside legal counsel. He reportedly encouraged the board to give the company’s CEO sole authority to grant stock options with minimal oversight — a “committee of one.” That committee of one became the center of an options backdating scandal that led to criminal prosecution. Many of the Silicon Valley companies investigated for options backdating at the time were Wilson Sonsini clients. When Brocade needed defense counsel in the resulting derivative lawsuit, a judge raised concerns that Wilson Sonsini was so tainted by conflicts of interest that the company had to drop the firm entirely. In a separate case, a company called Existence Genetics sued Wilson Sonsini for failing to disclose that while it was representing Existence Genetics, it was simultaneously representing a direct competitor — Navigenics. Two clients. One firm. Competing interests. No disclosure. These are the documented cases. The ones that became lawsuits, that produced court filings, that got written up in legal trade publications. The undocumented version of this — the version that never becomes a lawsuit because the founder doesn’t have the money or the evidence or the energy to fight a firm with two thousand attorneys — is orders of magnitude larger. And then there is the board seat. Some of these law firms, or their senior partners individually, will want something beyond hourly fees. They will want equity. An advisory role. A seat at the table that gives them ongoing access to company information and company decisions long after the initial deal is done. They will frame this as alignment — as skin in the game, as a signal that they believe in what you are building. What it actually is, is a mechanism for converting a service relationship into an ownership relationship, one that gives them preferred access to information that other shareholders do not have and creates a financial incentive that runs parallel to, and will sometimes conflict with, their legal obligation to advise you honestly. There is something that the smarter practitioners in this world will admit privately: when a VC firm recommends a law firm to a founder, the founder should ask a very simple question. Why does this particular VC prefer this particular firm? What is it that this firm offers this VC in return for that referral? The answer, in most cases, is not that the firm produces better outcomes for founders. The answer is that the firm produces smoother transactions for investors — deals that close faster, with less friction, with terms that are more favorable to the party that will be back next month with another deal to do. You are not that party. You are a transaction. You will not be back next month. And the lawyer you hired knows this. I am not telling you that these firms never produce good work. Some of their lawyers are genuinely skilled, genuinely committed to their clients, and genuinely uncomfortable with the conflicts their firm’s business model creates. I know some of them personally. The problem is not individual lawyers. The problem is structural. The business model of the dominant Silicon Valley law firm is built on serving both sides of the same table — the founders who pay the hourly fees and the venture capital firms who generate the deal flow — and calling it representation. You cannot serve two masters. Not when the interests of those masters diverge at the exact moment that matters most — which is the moment you are sitting across a term sheet from someone whose goal is to maximize their preferred return at your expense. The boilerplate they hand you is not neutral. The market standard they invoke is not a law of nature. It is a document written by people who were not thinking about you when they

wrote it. And the lawyer who tells you to sign it, who tells you it is standard, who tells you that pushing back will make you look difficult and harm your relationship with the investor — that lawyer is telling you something that is in their interest to tell you. Before you hire a lawyer in Silicon Valley, ask them one question: who referred you to me, and what do you owe them? If they cannot answer that question directly, you already have your answer.

The Oracle Problem — Or, How They Sold You a Map of a Country That Doesn’t Exist Before we get to what happens in the room with the investor, I want to talk about what happens before you ever walk in the door. Because the equity trap I just described — the preferred versus common, the liquidation preferences, the cap table that was never designed in your favor — that trap has a predecessor. And the predecessor is the number on the slide. You know the number I mean. The one that says your addressable market is four billion dollars. The one that says your segment will grow at thirty-two percent compound annual growth rate over the next five years. The one with the logo of a firm that sounds authoritative — Gartner, Forrester, IDC, take your pick — stamped in the corner like a government seal. The number that, when you were twenty-three years old and building your first company and trying to convince anyone that what you were building was worth funding, felt like permission. Like proof. Like someone credible had looked at the landscape and confirmed that the thing you believed in was real. Let me tell you what that number actually is. I am going to tell you this from an unusual position. I have sat in every chair at this table. I sold data center infrastructure in a direct sales capacity — I was the person presenting the contract and explaining why the terms were in your interest. I was also a customer of these same companies — I signed the agreements, accepted the credits, and learned what the fine print meant when the projections did not materialize. And I worked as a senior alliances manager inside Amazon Web Services, managing big data and analytics partnerships, which means I was present for the Proof of Concept credit conversations, the AWS All-In commitment negotiations, and the internal discussions about what we expected in return. I am not describing this system from the outside. I built parts of it. I operated inside it. And I am telling you exactly how it works. Those analyst firms are not neutral observers of the technology market. They are paid participants in it. The companies that appear favorably in their Magic Quadrants — the leaders, the visionaries, the firms positioned in the upper right corner of every chart — are, in most cases, also paying clients of the firms producing those charts. The research is sponsored. The coverage is purchased. The analysts who left large technology companies and landed at Gartner or Forrester or IDC did not stop serving the interests of large technology companies when they changed business cards. They institutionalized that service. They built a system in which the companies with the largest marketing budgets receive the most favorable positioning, and then sold access to those rankings as objective research. This is not a secret. It is a business model. It is disclosed, in the fine print, in language calibrated to be present without being legible. And it has been running long enough, and cited often enough, that the entire industry has forgotten to question it. The result is a generation of founders — many of them brilliant, most of them young, almost all of them operating with limited real industry experience — who built their financial models on numbers that were, at their origin, marketing artifacts dressed as market research. They took the analyst figure and built a five-year revenue projection.

They took that projection and walked into a meeting with Amazon Web Services, or Microsoft Azure, or Google Cloud, and said: here is our growth curve. Here is where we will be in eighteen months. Here is where we will be in three years. And the cloud providers looked at those numbers and said: great. We believe you. Let us structure a deal around that trajectory. Nobody in that room questioned the analyst numbers. Not the founder, who needed them to be true. Not the cloud provider’s sales team, who needed the committed revenue the numbers justified. The fiction was useful to everyone sitting at the table, so no one at the table called it a fiction. I want to be precise about what that deal looks like — because I have seen it from both sides of the table, and the structure is important. The large cloud providers will offer a startup meaningful discounts on infrastructure — sometimes fifteen percent, sometimes thirty percent — in exchange for multi-year commitments tied to projected consumption. The logic they present to the founder sounds generous. You are going to need this infrastructure anyway. Lock in now, get the discount, plan your budget with certainty. They will sweeten the entry point with what they call Proof of Concept credits — anywhere from ten thousand to a hundred thousand dollars in free usage — to get you building on their platform before the commercial terms kick in. That credit is not generosity. It is a migration subsidy. It is designed to get your engineers building on their infrastructure, your data living in their storage, your architecture dependent on their proprietary services — so that by the time the commercial contract begins, the cost of leaving is no longer a commercial question. It is an engineering question. A staffing question. A six-to-twelve month project that your operational budget cannot absorb and your investors will not fund. Inside AWS, we expected a ten-to-one return on every dollar of Proof of Concept credit issued. Ten dollars of future contracted revenue for every dollar extended. I want you to sit with that. I want you to think about what kind of institution demands a ten-to-one return on a short-term credit instrument extended to a company with no proven revenue. Not a bank. A bank charging those terms would be in violation of usury law in most jurisdictions. Not a reputable investor. A thirty percent annual return would make most investors ecstatic. The only class of lender that expects ten-to-one on a short-term extension of credit is the kind you do not want to owe money to. We dressed it in the language of partnership. We called it investment in your success. It was neither. And then the revenue projections — built on the analyst numbers, built on the market size figures, built on the Magic Quadrant that a paying client commissioned — do not materialize. Because they were never real. Because a company staffed by twenty-three-year-olds building their first product in a segment they learned about from sponsored research does not capture a significant percent of a four billion dollar market in three years. Now the founder is on the hook for the contract. Infrastructure costs are fixed and revenue is not. And now the cloud provider — which has already captured the architecture, the data, and the engineering team’s institutional knowledge of their platform — presents the AWS All-In program. I want to be clear about what AWS All-In actually is, because the name is doing a great deal of work. It is presented as a strategic partnership. A vote of confidence. An opportunity to consolidate your infrastructure and unlock deeper discounts in exchange for a long-term commitment to run everything — every workload, every service, every corner of your technical stack — on AWS. What it actually is, is a formalization of the trap you are already in. By the time AWS All-In is on the table, you have already migrated your infrastructure. Your engineers already

know AWS. Your data is already in S3. Your pipelines are already running on their proprietary services. The question is no longer whether you are dependent on AWS. The question is whether you will sign the document that acknowledges it and commits you to deepening that dependency in exchange for a discount on the rates you are already paying. The cost of leaving — the real cost, not the commercial cost — is measured in engineering hours. In months of rebuilding. In the operational budget line that your CFO will look at and say: we cannot afford this right now. Not this quarter. Not next quarter. And by the time the quarter after that arrives, the dependency is older, the engineers who built on AWS are more fluent in AWS than in anything else, and the cost of migration has not gone down. It has gone up. This is not an accident. This is the product. The infrastructure is the delivery mechanism. The lock-in is what they are selling. And the Proof of Concept credit — the ten thousand dollars, the hundred thousand dollars, the number that felt like a gift when you were a founder trying to keep the lights on — was the down payment on a decade of dependency. The analyst numbers manufactured the illusion of a market. The market projections justified the infrastructure commitments. The infrastructure commitments created the lock-in. The lock-in made AWS All-In feel like an option rather than a conclusion that had already been written. And the dependency is what the cloud providers were building toward from the first conversation. This is the trap that comes before the equity trap. And it is designed by people who understood exactly what they were building. If you are a founder reading this — if you are sitting somewhere right now with an analyst report in one tab and a Proof of Concept credit offer in another — I want you to understand something that nobody said to me when I was on your side of the table. The number in the corner of the slide is not evidence. It is the beginning of the trap. The deals are made in boardrooms, with a handshake and a wink. Company A — let us say a large cloud infrastructure company or an AI platform — decides to invest in Company X, a startup in an adjacent space. The investment is made on a given day. Thirty to ninety days later — at a carefully chosen interval designed to ensure that the two transactions appear unrelated in regulatory filings — Company X announces a series of commitments back to Company A: stock warrants, stock options, and a structured multi-year revenue agreement for Company X to purchase products or services from Company A. The investment and the revenue commitment were agreed upon simultaneously, in the same room, on the same day. But because they are executed on paper as separate transactions spaced weeks apart, they avoid the appearance of the quid pro quo arrangement they actually are. What this produces is the following effect. Company A's books show a capital investment in Company X as an asset, and simultaneously show Company X's revenue commitments as forward-contracted income — hyper-growth on paper, the kind of growth that drives share price, that drives analyst upgrades, that drives the valuation multiple that makes every insider's equity worth more. When Company X eventually fails to meet its revenue targets — as many of these startups do — Company A moves the warrants, the expected revenue, and any stock received to the loss column. Those losses are spread across three to five years of write-downs, timed to minimize their impact on any single earnings period, absorbed quietly into the accounting without the kind of market reaction that a single large write-down would produce. The stock price boost from the original deal has already been captured. The insiders who benefited from that boost have already been compensated. The write-down is paperwork. This is how the funny money of Silicon Valley

operates. This is how companies like OpenAI, Oracle, Amazon, and Nvidia participate in a circular ecosystem of investment and revenue that creates the appearance of explosive growth while the underlying economics are, in many cases, paper transactions moving from one column to another. It is not a coincidence that the same names appear repeatedly across these investment networks — investing in each other, contracting with each other, serving on each other's boards, and writing each other's reference letters when the SEC asks questions. I am not a securities lawyer. I am documenting what I have seen and what the public record reflects. What I will say plainly is this: the distance between what these men call aggressive accounting and what the some in law would call securities manipulation is, in many cases, measured in days on a calendar — the days between the handshake and the paperwork.

The Epstein Class The second phrase I use throughout this book is the Epstein Class. And I want to be precise about what I mean — because it is not synonymous with the PayPal Mafia, although there is significant overlap, and because the defining characteristic of the Epstein Class is not association with Jeffrey Epstein before his first conviction. It is association with him after it. Jeffrey Epstein was convicted of soliciting prostitution from a minor in 2008. He registered as a sex offender. He served thirteen months of an eighteen-month sentence — a sentence so light, secured through a non-prosecution agreement negotiated by then-U.S. Attorney Alexander Acosta, that it represented one of the most visible examples in recent American legal history of wealth purchasing its way out of consequences that would have destroyed any ordinary person. He was a convicted sex offender operating in the open. And after that conviction, certain men chose to continue associating with him. Not men who didn't know. Men who knew, and chose to continue anyway — because Epstein's network of money, access, and connections was worth more to them than the moral cost of the association, or because they believed, as men of that class often believe, that the rules applying to Epstein did not apply to them by extension. Bill Gates met with Epstein at least six times after his 2008 conviction, according to reporting by the New York Times and acknowledged by Gates himself. Gates has called this a mistake. I believe him that he now regrets it. I do not believe the meetings were accidental. A man of Gates' stature does not accidentally schedule six separate meetings with a convicted sex offender. Reid Hoffman — co-founder of LinkedIn later acquired by Microsoft, Bill Gates Company and a figure whose connections bridge both the PayPal network and the broader Silicon Valley philanthropic infrastructure — attended dinners with Epstein after his conviction and has publicly acknowledged doing so. Hoffman has said he was trying to secure funding for the MIT Media Lab through Epstein's connections. Joi Ito, the director of the MIT Media Lab, subsequently resigned after it emerged that he had accepted Epstein money and concealed its source. The philanthropic infrastructure of elite American universities was, for years after Epstein's conviction, being maintained in part by money that came through a convicted sex offender — and the men who facilitated those flows knew where the money came from. Peter Thiel's relationship with this network is documented through multiple channels, including video recordings in which Ehud Barak — the former Israeli Prime Minister who had one of the most extensively documented relationships with Epstein in the post-conviction period, including financial transfers and repeated visits to Epstein's properties — references Thiel's association and introduction into that network. Thiel is simultaneously a co-founder of PayPal alongside Musk,

the founder of Palantir, a primary funder of far-right political candidates and causes in the United States and Europe, and a man who has stated publicly that he does not believe freedom and democracy are compatible. That last statement is not an interpretation of his views. It is a direct quote from his 2009 essay "The Education of a Libertarian," published in the Cato Unbound journal, in which he wrote: "I no longer believe that freedom and democracy are compatible." A man who builds surveillance infrastructure, funds authoritarian political movements, and has openly stated his opposition to democratic governance is not a complicated figure. He is a man who has told you exactly who he is. The question is whether we have chosen to listen.

Why the Distinction Matters I separate these two categories — PayPal Mafia and Epstein Class — because they represent different species of the same genus. The PayPal Mafia is a financial and political network held together by shared capital interests, interlocking equity stakes, and the mutual protection of valuations built on the accounting architecture I described above. The Epstein Class is something darker: a social network held together by shared knowledge of shared crimes, and by the understanding — implicit, never spoken, understood by everyone in the room — that the protection flowing from that network runs in both directions. What connects them is power. The power to make the SEC look away. The power to make the DOJ cut a deal. The power to make the MIT Media Lab take the money and ask no questions. The power to ensure that the documents stay sealed, that the flight logs stay redacted, that the names stay private, that the record stays incomplete. That power has a cost. The cost is paid by the people who are not in the room when the handshake happens. It is paid by the startup founders whose companies were consumed by the circular investment machine and whose failure was someone else's write-down. It is paid by the workers in the warehouses and the content moderators in Nairobi and the engineers whose keystrokes are being captured to train the models that will replace them. It is paid by the women and girls who were in rooms with Jeffrey Epstein and could not leave. It is paid by every person whose suffering was, to the men in these networks, an externality — a cost to be absorbed by someone else, written off on someone else's books, buried in someone else's accounting period. I am documenting the cost. One chapter at a time.

We Have Been Here Before The blood-soaked hands of Silicon Valley's Big Tech executives are complicit in the enablement of the worst genocide the 21st century has witnessed — what is unfolding in Gaza. But before we examine the present, we must look at the past. Because we have been here before. And the names of the corporations involved are ones you know. In the 1930s, Adolf Hitler built his war machine and carried out the Holocaust with the direct assistance of major American corporations. U.S. bankers financed his regime through debt restructuring led by the houses of Morgan and Rockefeller. Standard Oil enabled the Nazis to refine synthetic fuel via the hydrogenation of coal — a technology critical to powering the German military. Henry Ford, a proud antisemite, provided the regime with Ford-Werke, which supplied transport vehicles for personnel and weapons. General Motors, through its Opel subsidiary, became Germany's largest automaker and built the trucking infrastructure alongside Ford for the German war effort. International Telephone and Telegraph, under CEO Sosthenes Behn, purchased a major stake in Focke-Wulf — the manufacturer of deadly fighter planes — and supplied the Nazi state with advanced telecommunications, radar components, and managed

telephone systems across occupied Europe. General Electric held cross-licensing agreements and patents with Siemens and AEG, and was a partner in forming IG Farben. Alcoa made agreements with IG Farben that restricted American production of magnesium — a key aircraft metal — handing Germany a significant technological advantage. IBM's German subsidiary, Dehomag, provided the Nazi state with punch-card tabulator systems. These were not simple counting tools. They were sophisticated data processors that enabled the bureaucracy of genocide — used to identify, categorize, and track populations for census, conscription, and the persecution of Jews and other targeted groups. Each of these corporations collected royalties on German military technology throughout the war. The Trading with the Enemy Act, passed in 1917, meant nothing when profit was available. These companies acted with impunity and enabled a genocidal war machine. Today, as genocide unfolds in Gaza, the names have changed. The pattern has not. Amazon and Google provide the foundational cloud and AI infrastructure through "Project Nimbus," a $1.2 billion contract that supplies the Israeli military with the data storage and processing power required for surveillance, intelligence, and real-time targeting. Palantir acts as the central brain, using its data fusion platforms to integrate intelligence from drones, satellites, and spies — creating an operational picture used for precise targeting that has killed tens of thousands of civilians, aid workers, and journalists. Oracle provides the critical database and cloud systems that manage vast surveillance data, forming the backbone of an intelligence apparatus used in a documented campaign of collective punishment. Microsoft powers the military's cloud infrastructure and AI tools, which process surveillance data and coordinate operations that have devastated civilian life across Gaza and Lebanon. The products built with this infrastructure have names that sound almost bureaucratic — clinical labels designed to keep the people who authorize them at a comfortable emotional distance from what they actually do. Gospel. Lavender. Where's Daddy. Blue Wolf. Red Wolf. Project Maven. The names change. The machinery does not.

The Gospel — The Mass Assassination Factory The Gospel, known in Hebrew as Habsora, was developed by Unit 8200 of the Israeli Intelligence Corps — the same elite military technology unit whose graduates flow directly into Silicon Valley and Israeli defense tech startups. The Gospel is an AI system designed to generate bombing targets automatically, at a rate no human intelligence operation could produce. Before the Gospel, Israeli military analysts could generate approximately 50 targets in Gaza per year. Once the Gospel was activated, it was producing 100 new targets per day. In the first weeks of the current genocide on Gaza, the Israeli Military claimed to have struck over 22,000 targets inside the Strip — at a daily rate more than double any previous operation. The Gospel does not merely assist human decision-making. According to a landmark investigation by the Israeli publication +972 Magazine, based on testimony from current and former Israeli intelligence officers, the system "generates targets almost automatically." One former intelligence officer described it plainly: a mass assassination factory. The system was used to justify strikes on private homes of junior Hamas members — and on homes where no known militant lived at all. The humans reviewing Gospel's recommendations were, by multiple accounts, rubber stamps. Some approved strikes in as little as twenty seconds. (Source: +972 Magazine and Local Call, "A Mass Assassination Factory: Inside Israel's Calculated Bombing of Gaza," https://www.972mag.com/mass-assassination-factory-israel-

calculated-bombing-gaza/; Democracy Now!, December 2023, https://www.democracynow.org/ 2023/12/1/israel_gaza_war_gospel_artificial_intelligence)

Lavender — The Kill List Generator Where the Gospel targeted locations, Lavender targeted people. It is an AI system that assigned every adult Palestinian in Gaza a numerical score from 1 to 100, rating the probability that each individual was a member of Hamas or Palestinian Islamic Jihad. At the outset of the current war, Lavender had flagged up to 37,000 Palestinians as potential assassination targets. The system's own architects knew it carried an error rate of approximately 10 percent — meaning that by their own internal calculation, roughly 3,700 people on the kill list were incorrectly identified. That error rate was considered acceptable. Officers who used the system testified to +972 Magazine that they treated Lavender's output "as if it were a human decision" — no independent verification required, no review of the underlying data. The AI had decided. The bomb fell. The system also flagged men based on criteria as broad as being in a WhatsApp group with a suspected operative, or sharing a name with someone on a list. Police officers, civil defense workers, and innocent family members were swept into the machinery. This is not targeted counterterrorism. It is algorithmic mass killing, validated after the fact by a human signature that had less than twenty seconds to think. (Source: +972 Magazine and Local Call, "Lavender: The AI Machine Directing Israel's Bombing Spree in Gaza," https:// http://www.972mag.com/lavender-ai-israeli-army-gaza/; West Point Lieber Institute, https:// lieber.westpoint.edu/gospel-lavender-law-armed-conflict/)

Where's Daddy — Killing Families to Kill One Person If the Gospel identifies what to bomb and Lavender identifies who to kill, Where's Daddy completes the system by identifying when. It was built specifically to track individuals on the Lavender kill list and alert military operators the moment those individuals entered their family homes at night — so that strikes could be timed for maximum lethality. Not on the street. Not in transit. At home. With their families. The name itself tells you everything about the moral universe in which this technology was developed. According to testimony from an Israeli intelligence official quoted in +972 Magazine, the Israeli Military "bombed them in homes without hesitation, as a first option" — meaning the family home was not a last resort but the preferred strike environment. Children were present. Spouses were present. Extended families were present. The system knew this. The operators knew this. The strike happened anyway. The only question the technology asked was: is the target home yet? (Source: +972 Magazine / Local Call, https://www.972mag.com/lavender-ai-israeli-army-gaza/; Wikipedia, AI-assisted targeting in the Gaza Strip, https://en.wikipedia.org/wiki/AI-assisted_targeting_in_the_Gaza_Strip)

Blue Wolf and Red Wolf — Biometric Occupation In the occupied West Bank, the architecture of AI-enabled control takes a different but equally chilling form. Blue Wolf is a smartphone application deployed to Israeli soldiers at military checkpoints. When a Palestinian approaches, the soldier photographs them. Within seconds, Blue Wolf matches the face against a vast biometric database — cataloguing name, family connections, address, and any "negative impressions" soldiers have noted in prior encounters. The system was gamified: military units competed weekly to see which battalion could photograph and register the most Palestinian faces, with prizes awarded to the winning unit. Amnesty International documented this explicitly in its 2023 report, Automated Apartheid. Children were photographed. Entries were

made in the middle of the night, at people's private homes, sometimes at gunpoint. Blue Wolf's companion system, Red Wolf, automates what Blue Wolf still left to humans. Deployed as fixed CCTV facial recognition infrastructure at checkpoints in Hebron since 2022, Red Wolf scans every face that passes without any soldier needing to lift a device. If a face is unrecognized, it is added to the database automatically. No consent. No notification. No recourse. A Palestinian resident of Hebron told Amnesty International: "They can tell you that your name is in the database, as simple as that, and then you're not allowed to pass through to your house." This is not security. This is the algorithmic management of a population — what Amnesty International has formally called automated apartheid. (Source: Amnesty International, "Automated Apartheid," May 2023, https://www.amnesty.org/en/latest/news/2023/05/israel-opt-israeli-authorities-are-using-facial-rec ognition-technology-to-entrench-apartheid/; Washington Post, November 2021)

Project Maven — And the Company That Built My Editor I do not exempt Anthropic — the company that built the AI I am using to write this book — from this accounting. That would be dishonest, and this book is not in the business of dishonesty. Project Maven began in 2017 as a Pentagon program to apply AI and machine learning to the processing of drone surveillance footage. The goal was to automate the identification of objects, people, and patterns of movement at a scale and speed no human analyst could match — and then translate those identifications into targeting recommendations. Google was the first major tech company to join Maven, in a contract initially worth approximately $9 million. What followed was one of Silicon Valley's most significant internal rebellions: thousands of Google employees signed petitions demanding the company withdraw, nearly a dozen resigned in protest, and the public pressure was enough that Google ultimately declined to renew the contract in 2018 and published a set of AI principles stating it would not build weapons systems. The principles meant something — for a moment. That moment did not last. Palantir stepped in where Google stepped out, and Project Maven has expanded steadily ever since. By May 2025, the Pentagon had raised the Maven Smart System contract ceiling to $1.3 billion through 2029. Maven is now a formal program of record across the entire United States Department of War. During the illegal war against Iran in 2026, Maven reportedly enabled the striking of over 1,000 targets in a single day — with future projections of 5,000 targets per day after the integration of large language models. The machine no longer assists the targeting process. It has become the targeting process. Now to the part I am obligated to say. In 2024, Palantir — the company that now operates Maven — partnered with Anthropic to integrate Claude into Palantir's defense software platforms. Anthropic stated publicly that Claude was being used to help the military process data and make decisions. In 2023 and 2024, Amazon invested in Anthropic. Amazon, as I have already documented in this chapter, is a core partner in Project Nimbus — the $1.2 billion cloud contract supplying the Israeli military's intelligence infrastructure. These are not separate worlds. They are the same world, connected by capital, contracts, and consequence. I use Claude as my editor. I have disclosed that from the first page. What I will not do is pretend that disclosure resolves the problem it names. The technology I am using to write this critique is manufactured by a company whose tools are embedded in the same military-industrial system I am critiquing. I hold that tension openly. I do not know how to resolve it except to keep writing, keep naming, and keep demanding that the

engineers and executives who build these systems answer for where their tools end up. That is what this book is. (Source: Wikipedia, Project Maven, https://en.wikipedia.org/wiki/ Project_Maven; Military.com, "Pentagon Expands Palantir's Role," March 2026, https:// http://www.military.com/feature/2026/03/22/pentagon-expands-palantirs-role-ai-contract.html; C4ISRNET, https://www.c4isrnet.com/it-networks/2018/07/27/targeting-the-future-of-the-dods-controversial-p roject-maven-initiative/)

The historical record shows that freezing assets or imposing penalties is not enough to stop genocide. The executives of Big Tech must be investigated and prosecuted for their roles. They must be held fully accountable for the crimes their technologies enable. They should not be free to retreat to bunkers in Hawaii or Patagonia, or sail their yachts openly in the Mediterranean, while their tools fuel slaughter. I think of Amal Khalil when I hear that argument. She was forty-three years old. She worked as a field journalist for Al-Akhbar, an Arabic-language newspaper based in Beirut. She regularly reported from high-risk areas of southern Lebanon, documenting what the weapons described in this chapter produce in real communities and real bodies. On April 23, 2026, she was covering the aftermath of an earlier airstrike when another strike hit near her location. She sought shelter in a nearby building. That building was struck too. When emergency teams finally reached her — after delays caused by ongoing shelling that prevented rescue access — she was dead. The Committee to Protect Journalists called for an urgent international investigation into what Lebanon's government officially described as the targeting of a journalist by the Israeli military. (Source: ParentsPlea.com/amal-khalil; https://cpj.org/ 2026/04/cpj-calls-for-urgent-international-investigation-into-israels-killing-of-leban ese-journalist-amal-khalil/) Her profile is on ParentsPlea.com. It is there because the same technologies I am describing in this chapter — the surveillance systems, the AI targeting tools, the cloud infrastructure — produced her death. And it is there because people like her, who documented what those technologies do, are being killed before they can keep documenting it. Someone has to keep the record when the people who were making it are gone.

--- Knowing the history of how we arrived here — IBM and Dehomag, Standard Oil and the Nazi war machine, and their modern equivalents in Amazon, Google, and Palantir — makes the evolution of my own response to this moment legible. Chapter Three is about that evolution: the specific moment I gave up on the belief that these companies could be reformed from within, and the decision to pursue the only form of accountability that does not require their cooperation.