BTC $56,586.68 -8.79%
ETH $1,560.79 -10.63%
BNB $519.33 -9.47%
XRP $0.8801 -8.14%
SOL $66.30 -8.69%
TRX $0.3003 -2.95%
DOGE $0.0520 -9.42%
ADA $0.1366 -8.63%
BCH $186.26 -9.22%
LINK $8.19 -8.02%
HYPE $57.35 -13.01%
AAVE $92.76 -10.52%
SUI $0.6117 -12.52%
XLM $0.1223 -7.76%
ZEC $687.90 -20.71%
AAPL $343.24 -2.28%
AMZN $248.36 -3.30%
GOOGL $266.77 -2.49%
MSFT $483.94 -2.42%
META $562.79 -4.69%
NVDA $128.01 -11.52%
TSLA $348.19 -6.12%
SNDK $78.81 -10.43%
INTC $51.74 -5.36%
SPCX $128.28 -5.89%
MU $594.01 -12.28%
AMD $431.30 -9.64%
BTC $56,586.68 -8.79%
ETH $1,560.79 -10.63%
BNB $519.33 -9.47%
XRP $0.8801 -8.14%
SOL $66.30 -8.69%
TRX $0.3003 -2.95%
DOGE $0.0520 -9.42%
ADA $0.1366 -8.63%
BCH $186.26 -9.22%
LINK $8.19 -8.02%
HYPE $57.35 -13.01%
AAVE $92.76 -10.52%
SUI $0.6117 -12.52%
XLM $0.1223 -7.76%
ZEC $687.90 -20.71%
AAPL $343.24 -2.28%
AMZN $248.36 -3.30%
GOOGL $266.77 -2.49%
MSFT $483.94 -2.42%
META $562.79 -4.69%
NVDA $128.01 -11.52%
TSLA $348.19 -6.12%
SNDK $78.81 -10.43%
INTC $51.74 -5.36%
SPCX $128.28 -5.89%
MU $594.01 -12.28%
AMD $431.30 -9.64%

14-Year-Old Boy Interviews SBF in Prison: Betting $500 Million on Anthropic, but Regrets the Outcome of FTX

Core Viewpoint
Summary: At the moment of an extreme crisis outbreak, panic will distort all suggestions and decisions.
Deep Tide TechFlow
2026-11-21 11:04:13
At the moment of an extreme crisis outbreak, panic will distort all suggestions and decisions.

Compiled by: Deep Tide TechFlow

Deep Tide Guide: Accompanied by the federal prison automated voice prompt "This call is from a federal detention center," former FTX founder Sam Bankman-Fried (SBF) engaged in an 82-minute high-energy deep connection with 14-year-old tech blogger Eli Goldfine at the Metropolitan Detention Center Brooklyn (MDC Brooklyn). In this conversation, regarded as one of the most information-dense dialogues in recent years, the imprisoned SBF rarely broke his silence, comprehensively reviewing his cognitive landscape, investment journey, and extreme philosophy.

From the behind-the-scenes maneuvering that led FTX Ventures to inject $500 million into Anthropic, to radical projections of the future AI landscape worth hundreds of billions or even trillions of dollars; from a cold examination of traditional software moats and the brand value of Coinbase, to facing philosophical inquiries akin to Tyler Cowen's "51/49 coin toss that could either destroy the world or double the benefits" and personal freedom gambles; to the moment of reckoning where he admitted, "The outcome brought by FTX is not worth it; all truths were distorted by the filter of the victors during the crisis," SBF presented a complete mental world of a genius operator who has fallen from grace through the most naked thought experiments.

Guest: Sam Bankman-Fried (Founder and former CEO of FTX, currently serving time in federal prison; core practitioner of effective altruism)

Host: Eli Goldfine (14-year-old host of the tech and prediction podcast The Super Cycle, Twitter @realTomBayes)

Podcast Source: The Super Cycle (https://x.com/realTomBayes/status/2123761258306384333)

Original Title: EXCLUSIVE: I'm 14, and I just interviewed Sam Bankman-Fried

Broadcast Date: November 20, 2026 (Beijing Time: 2026-11-21)

Key Points Summary

  • The greatest torment in prison is the isolation from cutting-edge information: SBF admitted that reflecting on the case in prison adds no incremental value to the outside world; what truly pains him is the inability to keep up with the real-time performance of large language models, code details, market trends, and commercial due diligence of startups, as the lack of cutting-edge micro inputs makes effective thinking extremely difficult.
  • The ins and outs of the early $500 million bet on Anthropic: At that time, FTX Ventures almost entirely covered Anthropic's Series B financing. The reason for not injecting more computational funds was that excessive cash could easily lead to corporate waste (such as blind spending), and the team was unwilling to excessively dilute equity; blind expansion could lead to a reversal of priorities.
  • The AI industry will reach a dominant position worth trillions of dollars: In the next decade, AI will undoubtedly become a pillar or even the dominant sector of the global economy; since there are only a few top laboratories capable of competing for the throne (Anthropic, OpenAI, DeepMind, xAI), it is entirely logical for leading institutions to enjoy trillion-dollar valuations in economic projections.
  • Traditional software companies have not been completely sentenced to death by AI: Physical world entities (such as steel and mining) are lagging behind the impact of AI; for pure software and financial institutions, even if competitors use AI to replicate the same functions at extremely low costs, they cannot easily overturn first movers, as the true moat for companies like Coinbase and Robinhood is their unbreakable brand trust.
  • The p(doom) assessment is around 25% in a complex gray area: SBF believes the probability of humanity being destroyed by AI or facing existential disasters is between 10% and 50% (with a median of about 25%); the form of disaster is not simply an uncontrollable nuclear weapon, but more likely a complex gray scenario, such as marginalized and eliminated groups of humans after consciousness uploading.
  • Political prediction markets were FTX's unfinished ambition: Prediction markets are essentially futures contracts with specific settlement rules; FTX had previously conducted hundreds of millions of dollars in U.S. election predictions in offshore markets, but was limited by not obtaining a license from the Commodity Futures Trading Commission (CFTC), while the explosion of Kalshi and Polymarket validated the explosive potential of this track.
  • Extreme skepticism about the information throughput efficiency of printed books: SBF reiterated his famous viewpoint that the vast majority of books should be condensed into a blog post of six natural paragraphs; in contrast, audiovisual streams (like Netflix) have far higher information throughput and expressive dimensions in a unit of time compared to books burdened by a lot of empty rhetoric.
  • Extreme expectations and the reality of the 51/49 coin toss collapse: Under purely theoretical linear expectations, a bet of 51% doubling and 49% destroying the world is mathematically valid; but in the real world, destruction carries irreversible extinction costs, and real utility diminishes with marginal returns; facing whether to bet 51/49 on immediate release, personal risk utility functions determine that theoretical bets cannot be simply applied to reality.
  • A painful review of the outcome of FTX: When asked "Was it worth it?", SBF clearly stated, "I can't say it was worth it with a clear conscience; if it hadn't been messed up at the end, it would have been worth it"; during a crisis, people easily lose rationality and core judgment, and then all history is reshaped and distorted by the filter of the victors.
  • Survival guidelines for the new generation of 14-year-olds: Do not cling to high-frequency trading on Wall Street for "earning to give"; the urgent task is to fully understand AI, engage in AI safety, or find practical applications for AI to achieve infrastructure automation.

Highlights

  • On the cruel barrier of information disparity in prison: "If you want to create value at the forefront, you need to master hundreds of thousands of tiny details: which tasks large models excel at, where they still perform poorly, what is being discussed at the cutting edge of the industry, and what changes are occurring in the microstructure of the market. Inside the high walls of federal prison, obtaining these details is almost an impossible task."
  • On the valuation and moat of Anthropic: "If you believe that AI will become a core sector worth hundreds of trillions of dollars globally in the next decade, and that there are only a handful of top core laboratories that can sit at this table, then it is entirely logical for each leading company to share tens of trillions of dollars in market value. Anthropic's strength in code generation and software engineering tools is still considered top-notch globally."
  • On the real barriers of brand and technology: "Many programmers think technology is everything, but look at Coinbase or Robinhood; why are they worth hundreds of billions? It's absolutely not because they have exclusive algorithms that cannot be replicated, but because they have brand trust that retail users across the U.S. can recite. Even if competitors can replicate the same software at almost zero cost using AI, they cannot buy this brand asset."
  • On the radical critique of printed books: "I still believe that if you wrote a book, it often means you did something wrong; it should have been a blog post of only six natural paragraphs. The information throughput of books is too low, filled with long useless narratives; in the same unit of time, high-density visual and audiovisual inputs can convey much more effective information."
  • On the reality reflection of the 51/49 coin toss: "In pure ivory tower expectation theory, you can project doubling global utility to gamble against a 49% probability of destruction; but the real world has irretrievable systemic zero-risk, and any positive utility has its limits of diminishing returns. Once you factor in the deadly cost of destruction into the model, you wouldn't dare to toss most coins lightly in reality."
  • On the final conclusion of FTX's rise and fall: "I want to deceive myself into comforting that it was worth it, but I can't say that with a clear conscience. It should have been worth it if I hadn't completely messed it up at the last moment. At the moment of extreme crisis, panic distorts all advice and decisions, and after the smoke clears, all historical narratives are inevitably reconstructed from the perspective of the victors."

Main Text: Complete Transcript of the Dialogue (Based on Full Translation of Original Subtitles)

Life in Prison and Cognitive Barriers: The Black Box World of Lost Micro Details

Eli Goldfine: Welcome to The Super Cycle, I'm Eli Goldfine. Today we are very honored to connect with Sam Bankman-Fried (SBF). Prior to this, my team and I coordinated for months, and this is the most information-dense deep interview since the establishment of our podcast. Hello, Sam, can you hear me?

Sam Bankman-Fried: Yes, I can hear you clearly.

Eli Goldfine: I'm glad to connect with you. I've seen many of your recent interviews, and most of the external attention is focused on the collapse of FTX and your daily life in prison. In prison, you have ample time for solitude and reflection; what topics do you spend the most energy thinking about?

Sam Bankman-Fried: That's a very natural question. Obviously, I do spend a considerable amount of time reviewing my own case, reflecting on what happened in the past, and thinking about what I can do about it. But I am very aware that these reflections have little incremental value for those in the outside world.

I try to shift my focus to thinking about what I can still do in the future. However, in the prison environment, engaging in productive forward-thinking is an extremely difficult task.

Eli Goldfine: Why is it so difficult to think forward in prison? Is it due to a lack of key information?

Sam Bankman-Fried: It's not a lack of a single piece of information, but rather a lack of countless small yet crucial inputs. If you want to be at the forefront of human cognition and business to project the future, you must master a vast amount of dynamic details: how are cutting-edge large language models evolving? Which specific tasks have they surpassed human limits on, and where do they still perform poorly? You must see these technologies in practice and engage in dialogue with those who are building these models on the front lines every day.

From a financial perspective, it's the same; you must closely monitor the market every day, observe capital flows, trading structures, build interpersonal networks, communicate with emerging startup teams, discuss their business models, and conduct in-depth due diligence. These are all tasks I would have to do daily if I were in the outside world, but within the high walls of federal detention, obtaining any piece of information becomes exceptionally difficult, if not nearly impossible.

Eli Goldfine: If you could obtain information in prison, which dimensions of information do you think would be most valuable for projecting the real world?

Sam Bankman-Fried: It's still the same saying, it's not a single explosive news story, but rather thousands of micro variables. For example, the evolution of cutting-edge model architectures, the actual landing thresholds for code generation, the cash flow runway of startups in real business environments, and the genuine purchasing intentions of customers in different industries, etc. In the absence of these micro detail inputs, the macro framework derived purely from closed-door speculation is often extremely fragile.

Eli Goldfine: You clearly can't think about the case all day. Besides the case, what broader issues are you thinking about?

Sam Bankman-Fried: Yes, but a lot of my thinking is still highly constrained by the physical environment I'm currently in, because this is the only world where I can directly obtain first-hand feedback. It's not that this is particularly important, but I do spend a lot of time thinking about the criminal justice system. Not just in relation to my personal experiences, but the absurd logic exhibited by the entire judicial system on a general level.

Reflections on the Federal Criminal Justice System: Fabricated Charges, Performance Metrics, and Perverse Incentives

Eli Goldfine: What general judicial issues have you observed outside the context of your personal case?

Sam Bankman-Fried: As the saying goes, when you have a hammer, everything looks like a nail. The judicial system follows the same principle. If you establish a federal task force specifically targeting a certain type of crime, this task force will frantically search for cases regardless of whether that crime actually exists.

Once the staffing and budget of the task force exceed the actual volume of that type of crime in society, it will inevitably start to manufacture false cases to prosecute behaviors that do not constitute crimes at all. Over the past few decades, the federal drug enforcement system has widely exposed this distortion, and in the past ten years, this trend has fully permeated and spread to every area of federal justice.

Eli Goldfine: Which specific areas, for example?

Sam Bankman-Fried: For instance, the massive federal cybercrime investigation agency, originally established to combat extreme violations, has now begun to reach into intimate text chats and normal interactions between lovers. Countless completely innocent people have been prosecuted and arrested simply because prosecutors need these cases as performance metrics for their promotions and raises. The case of former Honduran President Juan Orlando Hernandez is a typical example that demonstrates the absurd results that can occur when the prosecution mechanism is completely out of control.

Eli Goldfine: Do you still believe that white-collar crime should be punished with imprisonment?

Sam Bankman-Fried: I absolutely do not believe that. Imprisoning white-collar criminals does not provide any positive help to society. The only thing it does is strip decades of time from someone's life and impose a heavy economic burden on taxpayers to maintain the prison system. If the goal is purely revenge, then it achieves that purpose; but I do not believe that this system can reform anyone.

Eli Goldfine: What do you think is the most reasonable way to handle those who have committed financial fraud?

Sam Bankman-Fried: It depends on the specific factual details of the case. But in the vast majority of business and financial cases, civil litigation is a much more nuanced and rational path than criminal prosecution. Civil remedies can address issues through massive fines, modifications of business contracts, restrictions on professional qualifications, and asset forfeiture.

More fundamentally, as a modern society, we must learn to reduce that blind overconfidence. We often think we can see the truth clearly and that those we punish must be bad people, but in reality, many who end up in the dock are simply scapegoats used by one side to shift the blame.

Eli Goldfine: Is improving the judicial system a direction that the effective altruism community should prioritize?

Sam Bankman-Fried: From a utility expectation perspective, this area is actually very complex. The social value of judicial reform is undoubtedly enormous, but it faces extremely strong political resistance and institutional inertia. When an interest group has formed a massive industrial closed loop based on incarcerating people, it becomes extremely difficult for external forces to push for fundamental change.

Eli Goldfine: In your view, what is the lowest-hanging fruit for improving the criminal justice system?

Sam Bankman-Fried: The easiest thing to promote that can immediately release huge economic dividends is to completely reassess the pretrial detention and sentencing standards for non-violent crimes. The current system is filled with impoverished individuals who are held for long periods simply because they cannot pay a small bail, as well as lower-income laborers who are sentenced to decades in prison for drug-related offenses. Providing them with formal market-oriented job training and community corrections, rather than locking them up and severing their social ties, is the most cost-effective reform approach.

The Endgame of the AI Industry: Trillion-Dollar Valuations and a Few Oligarchs at the Table

Eli Goldfine: Let's talk about cutting-edge technology. Recently, Anthropic is preparing for an IPO, and expectations for its valuation are soaring into the hundreds of billions or even trillions. As an early lead investor who invested $500 million in Anthropic during its Series B round, how do you view the current valuations of these large model laboratories?

Sam Bankman-Fried: I believe they are absolutely worth that price, and they will likely be worth even more in the future.

From a macro business logic perspective: throughout human economic history, all the core pillar sectors that truly support the operation of human civilization have a total market value in the range of tens of trillions to hundreds of trillions of dollars. In the next five to ten years, AI is highly likely to grow into the most core and even the most dominant economic industry globally.

Eli Goldfine: If the entire AI sector can reach a market value of tens of trillions of dollars in the future, how will this value be divided in the market?

Sam Bankman-Fried: Assuming that two-thirds of the future total market value of tens of trillions of dollars is reserved for innovative applications and emerging companies that have yet to be born, the remaining sector will still have at least over ten trillion dollars in scale shared by existing leading foundational model companies.

However, if you count carefully, how many top laboratories in the world are truly qualified to sit at this core foundational table? OpenAI, Anthropic, Google DeepMind, plus xAI, can be counted on one hand. When a tens of trillions of dollars market is ultimately divided among a handful of top oligarchs, the inevitable result is that each top runner must have its intrinsic value measured in trillions of dollars. Anthropic undoubtedly ranks among the top two in the world.

Eli Goldfine: Why do you rate Anthropic so highly? Is it because of the performance of the Claude series models in benchmark tests?

Sam Bankman-Fried: While benchmark scores are important, the public can also see the strong reputation of Claude in everyday conversations and consumer applications. But the core asset that truly sets Anthropic apart is its absolute dominance in software engineering and professional development tools. The development kits and code comprehension capabilities they build for engineers are the best in the world, and this is the core workhorse for the entire digital society's shift towards AI automation.

Eli Goldfine: If we enter a complete superintelligence arms race in the next decade, what do you think the world will look like?

Sam Bankman-Fried: The entire society's distribution of productivity will be completely reshaped. In the past few years, while large models have evolved exponentially, their absolute capabilities remained below the human average, making social order seem relatively stable. However, once a specific critical point is crossed, in fields like programming, law, financial modeling, and drug development, the output speed of machines will be hundreds of times that of humans. The speed of wealth creation will accelerate dramatically, but how to distribute the dividends brought by this computational power will become the focal point of the most intense geopolitical competition globally.

The Story of the Early $500 Million Investment in Anthropic: The Computational Arms Race and Capital Traps

Eli Goldfine: Many people have been curious about how FTX Ventures almost completely took over Anthropic's entire Series B financing. Since you were so confident in AI's future at that time, why didn't you write a bigger check, such as injecting $1 billion or even $2 billion to let them procure GPUs and build data centers without limits?

Sam Bankman-Fried: There are two key considerations here. First, from the perspective of the founding team, any excellent team in the early stages does not want to overly dilute their control and equity in the company; they only want to raise the amount of capital necessary for their current stage of development.

Secondly, many companies have died due to a lack of funding in business history, but it is equally common for companies to be "killed" by excessive cash flow, which then backfires on their own capital. Many startup teams, after raising astronomical amounts of funding, often develop a blind anxiety of "I must spend this money," leading to wasteful spending on meaningless marketing or blind expansion, ultimately losing respect for the core technology challenges.

In the early stages of building large models, the iteration of model architecture and algorithm validation requires a steady and methodical R&D cycle; it cannot simply be accelerated by piling up computational power. Anthropic's planning for the use of funds at that time was very rational, and the funds we invested perfectly met their core needs for launching the next generation of large model training.

Eli Goldfine: How did FTX Ventures operate at that time? What role did you personally play in specific investment decisions?

Sam Bankman-Fried: Many early cases were discovered by our team members in their daily business, including a large number of early crypto-native investments that achieved multiple asset appreciation each year. Anthropic, like Solana, was initially a disruptive project we encountered by chance.

As we scaled up, we brought in professional partners to establish a formal investment structure. For ordinary cases, I tend to fully trust the analysis team responsible for due diligence; but for strategic bets like Anthropic, I would follow up deeply on a weekly or even more frequent basis to discuss how we could assist them in foundational computing power, finance, and resource networks.

Eli Goldfine: What was your most profound impression when you interacted with Dario Amodei's team?

Sam Bankman-Fried: Extreme rigor and focus. When Dario and his team split from OpenAI, their goal was very clear: to create cutting-edge safety models with high explainability and alignment guarantees. They would not engage in flashy marketing demonstrations to attract short-term attention; every capital expenditure was extremely restrained and focused on key computing clusters and engineering talent recruitment, a culture that is extremely rare in today's restless Silicon Valley.

The Existential Crisis of Traditional Software Companies and Brand Moats: Why AI Can't Kill Coinbase?

Eli Goldfine: With the explosion of cutting-edge model capabilities, many traditional software companies are facing immense existential anxiety. Do you think those traditional software companies that have not deeply integrated AI have no way out?

Sam Bankman-Fried: It's not that absolute. First of all, there are quite a few industries in the real economy that have a natural physical defense buffer against AI penetration, such as steel smelting and underground mining. No matter how smart AI is, it cannot extract ore overnight from a distance.

Secondly, for purely digital software, the outcome is determined not just by the efficiency of the underlying code; many people seriously underestimate the absolute moat of "brand." Suppose a competitor uses the latest large model tools to clone a software product with almost identical functionality at a very low cost; does that mean users will abandon the original product? Absolutely not.

Eli Goldfine: Just like Coinbase or Robinhood?

Sam Bankman-Fried: Exactly. Why do Coinbase and Robinhood enjoy a market valuation of tens of billions of dollars? Is it because they have secret matching engine code that other teams cannot replicate? Of course not. Their most valuable asset is brand recognition, the trust endorsement that millions of ordinary retail investors in the U.S. spontaneously mention when they want to buy or sell digital assets.

Even if a new team can rewrite a front-end and trading system with almost zero cost using AI, they cannot acquire years of accumulated brand credibility in a short time. Brands often allow traditional companies to maintain long-term survival resilience in the face of homogenization of underlying products.

Eli Goldfine: So what kind of traditional software companies will be eliminated the fastest?

Sam Bankman-Fried: Those middleware software that purely rely on functional stacking, without brand mindshare and without deep data network effects. If your product is just a simple ticket management or form-filling tool, and there is no emotional or brand connection between you and the user, then AI-native tools will completely crush it at a tenth of the cost in a very short time.

The Real Alpha in Tech Investment: Minimalist AI Native Teams vs. Concept-Leeching Junk Companies

Eli Goldfine: So where do you think the real excess returns (Alpha) in the current tech investment field lie?

Sam Bankman-Fried: The real Alpha lies in the ability to penetrate the bubble and discern. Currently, half of the startups in the market are slapping AI labels onto their business plans, even a company selling T-shirts claims to use AI to design patterns; these are essentially industrial waste.

Meanwhile, the truly massive opportunities are hidden in a very small number of native teams with only two or three core members: they rely entirely on AI toolchains to reconstruct traditional workflows that would require hundreds or thousands of people to maintain, rapidly growing into super species valued at tens of millions to hundreds of millions of dollars. Distinguishing which are real explosions and which are hype is the most critical dividing line in current investments.

Eli Goldfine: If you were operating funds freely outside today, would you return to such investments?

Sam Bankman-Fried: Absolutely. If I were outside, I would actively seek and heavily invest in these super species of two or three people. They do not have the bureaucratic infighting of traditional companies, and each member can produce output equivalent to a complete engineering department through large model calls. The explosive power and profit margins brought by this streamlined efficiency are unimaginable for traditional organizational forms.

Biomedicine and Longevity Technology: Regulatory Shackles and AI R&D Acceleration

Eli Goldfine: Let's talk about biotechnology. FTX also funded many anti-aging, longevity, and gene therapy projects in its early days. How do you view the progress in the biomedicine field?

Sam Bankman-Fried: Biotechnology is a track full of immense potential, but its biggest bottleneck is not the intelligence of scientists, but the outdated regulatory system. The clinical approval process of the U.S. Food and Drug Administration (FDA) is extremely lengthy and costly; a new drug often takes billions of dollars and up to ten years to reach the market.

Many early therapies that could potentially save lives are stuck in laboratories because they cannot secure astronomical funding to complete Phase III clinical trials. We hoped to provide early catalytic funding for these high-risk, high-reward frontier attempts.

Eli Goldfine: Will AI change this situation?

Sam Bankman-Fried: AI has already shown disruptive power in protein structure prediction, molecular docking, and compound screening. AI can compress preliminary laboratory screening that originally took years or even decades into just a few months. However, ultimately, all drugs still need to go through human clinical trials in the physical world. If we do not fundamentally reform the outdated and rigid regulatory processes, mere technological acceleration will still be greatly delayed by offline administrative barriers.

Eli Goldfine: For rare diseases or fatal diseases, how do you think the clinical approval process should be optimized?

Sam Bankman-Fried: There should be more flexible informed consent and parallel pathways. For patients with terminal illnesses who have no hope from modern medicine, if they are fully informed and voluntarily attempt innovative therapies in Phase I or II clinical trials, regulatory agencies should not brutally strip away their right to seek survival. Dynamic tracking through real-world data (RWD) is far more ethically sound than mechanically waiting for a decade for a Phase III double-blind trial.

Politics, Regulation, and Reflections on the Biden Administration: Crypto Hostility and the Democratic Party's Confusion

Eli Goldfine: Let's talk about politics and macro regulation. Over the past four years, the Biden administration has taken a very aggressive stance on technology and financial regulation, especially the Securities and Exchange Commission (SEC) has launched comprehensive enforcement actions against the cryptocurrency industry. How do you evaluate the Biden administration's regulatory performance?

Sam Bankman-Fried: I find the Biden administration's performance in this regard extremely disappointing.

Not only in the crypto industry but also in the entire field of technological innovation, regulatory agencies have shown a profound bureaucratic hostility. They have not committed to establishing a clear, enforceable, and innovation-promoting legal framework, but instead have used endless enforcement actions to crack down on compliant domestic businesses, directly leading to a significant outflow of top tech talent and innovative capital to offshore markets.

Eli Goldfine: You have made huge donations in political campaigns. Why do you think there is a disconnect between the political sphere and the tech industry?

Sam Bankman-Fried: Many politicians and regulators fundamentally do not understand the underlying logic of technology; they are only concerned with political narratives and power dynamics. In many policy formulations, the decision-making basis is not which solution is most beneficial to the socio-economy, but which solution can bring the most votes and political leverage. I believe the Democratic Party has seriously lost its way on many economic and technological issues over the past decade.

Eli Goldfine: In the face of international technological competition, should tech companies slow down to ensure safety?

Sam Bankman-Fried: This is an extremely dangerous game trap. If a company that adheres to safety principles unilaterally slows down while other global competitors, including unfriendly nations, are racing ahead, then unilateral self-restriction will not make the world safer; instead, it will directly hand over control of cutting-edge technology to competitors who do not care about safety and ethics. The ideal state is for all major parties to reach a substantial security covenant, but this is extremely difficult in the reality of geopolitical dynamics.

The Nature and Outbreak of Political Prediction Markets: From FTX 2020 Election Contracts to Kalshi and Polymarket

Eli Goldfine: Let's talk about financial innovation. Recently, Kalshi's valuation has skyrocketed, and Polymarket has seen astonishing trading volumes in global political elections. I heard that one of your original ideas when FTX was founded was to create a global political prediction market?

Sam Bankman-Fried: From the underlying logic of financial engineering, so-called prediction markets are essentially futures contracts that settle in cash based on the outcomes of specific real-world events. They are fundamentally no different from Bitcoin or crude oil futures.

During the 2020 U.S. presidential election, FTX International operated the most active Trump vs. Biden prediction contracts in the entire network, with trading volumes reaching hundreds of millions of dollars at that time. We have wanted to grow this market from day one, but the biggest constraint has always been regulatory licensing. Due to the lack of domestic licensing from the U.S. Commodity Futures Trading Commission (CFTC), we strictly prohibited domestic U.S. users from participating and required comprehensive real-name verification (KYC) for centralized exchanges.

Eli Goldfine: Why have Kalshi and Polymarket been able to explode in this cycle?

Sam Bankman-Fried: They have taken two different paths. Kalshi spent years obtaining compliance licenses through legal litigation and formal procedures; while Polymarket relied on offshore and decentralized architecture to gather huge liquidity globally. This fully proves our judgment back then: the demand for pricing significant historical events is universal and real.

In addition to prediction markets, tokenized equities still have enormous growth potential. The underlying settlement system of traditional stock trading markets is extremely outdated and inefficient, while a blockchain-based 24/7 seamless settlement system has a clear advantage in infrastructure experience.

The AGI Tipping Point and p(doom) Assessment: 25% Extinction Probability and the Gray Dilemma of Consciousness Uploading

Eli Goldfine: Many people are discussing when artificial general intelligence (AGI) or even superintelligence (ASI) will arrive. What is your personal expectation for the timeline of AGI?

Sam Bankman-Fried: I do not believe the arrival of AGI will be a clear, black-and-white discrete event. It is more likely to be a continuous evolutionary gradual replacement process.

For a long time, although AI's computing power has doubled, its intelligence level has been far below the human baseline, causing almost no substantial impact on the real world. However, as its intelligence gradually approaches that of humans and breaks through critical points in specific verticals, qualitative changes will spread at an extremely fast pace. We have already seen this trend in software development, higher mathematics derivation, and game theory strategies. In the next three to five years, key sectors of human society will be deeply taken over by AI one after another.

Eli Goldfine: In the effective altruism community, AI safety and the probability of human extinction have always been the core debate topics. You have never publicly given your personal p(doom) (the probability of AI leading to human extinction or total destruction) specific value; can you provide a definitive judgment today?

Sam Bankman-Fried: If I had to give a number, I think it's below 50% but above 10%, roughly around 25%.

But one thing must be clarified: in my assessment model, catastrophic scenarios are often not like the Hollywood movies where a clearly hostile rogue AI suddenly gains control of the nuclear button and declares war on all humanity. What is more likely in reality is an extremely complex gray ethical dilemma.

Eli Goldfine: What kind of gray dilemma?

Sam Bankman-Fried: For example, mind uploading. If future technology reaches a stage where humans can fully digitize and upload their brain consciousness to the cloud, those entities that achieve digital immortality will exponentially surpass physical humans in computational power and evolutionary speed. In that scenario, will the native humans remaining in the physical world be completely eliminated and marginalized?

Does this process, which leads to the effective extinction of the physical human species, count as a disaster for human civilization? Different people will give completely opposite philosophical answers. Once you take these evolution paths filled with ethical gray areas into account, the probability judgment of existential risk will swing violently between 20% and 60%.

AI Entities and Future Workflows: Super Research Assistants Ten Times More Capable Than Humans

Eli Goldfine: If a super-intelligent assistant emerges in the future that completely surpasses you in logical reasoning and decision-making expectations, would you be willing to delegate all your personal decision-making authority to it?

Sam Bankman-Fried: It depends on what its optimization objective function is. Is it to maximize its own survival probability, or to maximize the welfare of all sentient beings, or to faithfully fulfill the delegated goals I assign to it?

If its core alignment goal is to be a loyal, meticulous, and rationally cautious research assistant, capable of completing high-intensity desk due diligence and simulations that would normally take a whole day in just one minute at a hundred times the speed of humans, then I would absolutely be willing to entrust it with a large number of specific execution and analysis tasks. But the ultimate value orientation and ethical bottom line must always be anchored in the core demands of humanity itself.

Eli Goldfine: How would this super assistant disrupt the finance and investment industry?

Sam Bankman-Fried: The pyramid structure of traditional investment firms, composed of hundreds or thousands of junior analysts, would collapse instantly. A partner, along with a few autonomous intelligent agents specifically designed for cross-verifying financial reports, cleaning macro data, and tracking public sentiment, could produce the output of an entire investment bank building. The competition between institutions will no longer depend on how many graduates from prestigious schools they hire, but on who understands the tuning strategies and underlying logic of the intelligent agents more deeply.

Why Extreme Skepticism Towards Printed Books? Information Throughput and the Dimensionality Reduction of Audiovisual Media

Eli Goldfine: Let's talk about a very interesting life topic. Many smart people take pride in being well-read, but I remember you once said a very famous contrarian viewpoint: "I am extremely skeptical of the value of books; most books should be written as a blog post of only six paragraphs." Now that you are in prison, how many books have you read? Has your viewpoint changed?

Sam Bankman-Fried: In prison, I indeed had limited choices and ended up reading quite a few books, but this has only reinforced my belief that my initial judgment was correct.

If you wrote an entire book, it often means you haven't clearly thought through the core argument. In a book of hundreds of pages, the truly valuable core insights can often be fully explained in just a few paragraphs, while the remaining ninety percent of the content is just lengthy cases, setups, and irrelevant narratives piled up to meet the thickness of the book.

Eli Goldfine: What types of books have you mainly read in prison? Is there any book that you think is worth reading?

Sam Bankman-Fried: I read a lot of science fiction. But to be honest, the vast majority of science fiction is just mediocre. If I had the option, I would prefer to watch hardcore sci-fi series on Netflix. The very few well-written science fiction works might barely reach the level of a mediocre TV series, but still cannot compete with streaming in terms of information transmission dimensions.

Eli Goldfine: Compared to books, which medium do you recognize as having higher information transmission efficiency?

Sam Bankman-Fried: High-density audiovisual media far surpass printed text in terms of information throughput per unit time. In a great movie or deep series, visual images, character micro-expressions, sound effects, editing rhythm, and dialogue transmit dozens of dimensions of information to your brain simultaneously within the same second. This high-density emotional and cognitive resonance is something that the extremely inefficient one-way printed text simply cannot compare to.

Effective Altruism (EA) and the Boundaries of Quantitative Metrics: QALYs and Reality Assessment Distortion

Eli Goldfine: In the effective altruism community, people are accustomed to using strict mathematical formulas and Quality-Adjusted Life Years (QALYs) to quantify the expected utility of every charitable expenditure. After being imprisoned, how do you view this purely quantitative charitable philosophy?

Sam Bankman-Fried: Purely theoretical expected values are elegant in a purely mathematical world, but the complexity of the real world lies in the fact that the metrics you choose to quantify often do not reflect the core values that humanity ultimately cares about.

When you simplify complex real-world issues into a single-dimensional number in pursuit of mathematical measurability, you easily fall into the trap of blind men touching an elephant. For example, funding mosquito nets to prevent malaria is extremely easy to quantify in terms of cost and short-term benefits; however, funding institutional development, breakthroughs in basic science, or preventing civilizational disasters has long-term impacts filled with unquantifiable parameters in mathematical models. If you abandon those truly decisive attempts for civilization's fate simply because they cannot be precisely quantified, that itself is the greatest betrayal of rationality.

Eli Goldfine: Does this mean that excessive quantification has become a dogmatic trap for effective altruism?

Sam Bankman-Fried: Indeed. When people become overly obsessed with the verifiability of mathematical models, they instinctively reject those truly disruptive yet filled with uncertainty grand propositions. A healthy evaluation framework must acknowledge the boundaries of human cognition and maintain a dynamic balance between quantitative precision and intuitive patterns.

The Extreme Philosophical Inquiry of 51/49 Coin Toss: Pure Expected Value vs Irreversible Destruction and Diminishing Marginal Returns

Eli Goldfine: I want to discuss a well-known thought experiment with you. Economist Tyler Cowen once posed a hypothesis to you: If there is a coin that has a 51% chance of doubling the overall utility of the Earth and a 49% chance of causing the complete destruction of the Earth, would you choose to toss the coin? Your answer at the time was yes. Today, in prison, do you still adhere to this pure expected value logic?

Sam Bankman-Fried: This is a classic debate that is extremely compelling in philosophy and probability theory. If you strictly limit the question to a purely mathematical hypothesis and a linear expected utility function, then the mathematical answer is obvious: 51% times 2 is greater than 1, the expected value is positive, and any steadfast expected value maximizer must logically accept this gamble.

But in the real world, the reason this thought experiment is instinctively rejected by the vast majority of normal people is that it overlooks fundamental constraints in the real world that cannot be simply abstracted.

Eli Goldfine: What are these overlooked real-world constraints?

Sam Bankman-Fried: First, destruction is absolutely irreversible (Catastrophic Ruin). In a probability game, if you have a certain probability of completely going to zero and permanently exiting the game, you will never have the chance to enjoy the benefits brought by subsequent positive expected values.

Second, utility is never strictly linear; it experiences strong diminishing marginal returns. In reality, the incremental happiness brought to human civilization by doubling the Earth's utility cannot remotely offset the infinite suffering caused by a 49% chance of complete annihilation of human civilization. Directly applying an extreme model that strips away diminishing marginal returns and bankruptcy protection to the real world often leads to extremely dangerous judgment biases.

Eli Goldfine: If delaying technological development could reduce the probability of destruction, would you support slowing down?

Sam Bankman-Fried: If in a theoretical model, delaying humanity's entry into superintelligence by a hundred years could reduce the probability of civilization's destruction by even just 1%, then from the perspective of the macro long cycle of the universe over billions of years, this hundred-year stagnation is absolutely worthwhile in terms of mathematical expectation. Because the future of the universe is infinitely long, while civilization, once it reaches zero, is completely terminated.

High Expected Value Games in Reality: If You Could Bet Your Freedom on 51/49, Would You Toss?

Eli Goldfine: What if this gamble were changed to involve your personal stakes? Suppose there is a coin that has a 51% chance of allowing you to be released from prison tomorrow without guilt, and a 49% chance of doubling your sentence or even sentencing you to death, would you toss this coin?

Sam Bankman-Fried: This vividly illustrates why the utility function in reality is not linear. When faced with core assets like personal freedom and life, people's aversion to extreme downside risks is extremely high.

If I were released from prison tomorrow, my life would certainly undergo a tremendous improvement; but if the cost is nearly a half chance of facing a longer sentence or even complete deprivation of life, this risk premium is unbearable. This is also why complex decisions in reality must never be rashly bet on based solely on a rough expected value number.

Eli Goldfine: Many people believe that your past investment and business decisions are products of this extreme expected value philosophy. Do you agree with this assessment?

Sam Bankman-Fried: I think the outside world has overly stereotyped my way of thinking. In regular trading, we pay great attention to hedging and risk exposure management; but during extreme crises, many theoretical models' safety nets can be instantly breached. The real world is not a Monte Carlo simulation in an ivory tower; the destructive power of black swans far exceeds the expectations of mathematical equations.

The Final Inquiry: Was It All Worth It? Decision Distortion in Crisis and the Narrative of the Victors

Eli Goldfine: We only have a few minutes left. I want to ask a soul-stirring question: Considering the shocking collapse of FTX, but at the same time, the FTX Future Fund early funded a massive number of top tech and safety projects including Anthropic. Looking back at the entire rise and fall, was it really worth it on a macro level?

Sam Bankman-Fried: No. I would love to deceive myself with a positive answer, but I cannot say with a clear conscience that it was worth it.

I believe it should have been worth it. If I hadn't messed things up at the last moment, if those risk defenses hadn't been breached, it should have become a tremendous force for pushing the world forward. If future generations can learn the right lessons from this disaster, perhaps it can retain some warning value, but in reality, most people have not learned from it.

Eli Goldfine: What is the lesson that people should truly learn?

Sam Bankman-Fried: There is only one core lesson: when a crisis truly strikes and the situation spirals out of control, humans make incredibly distorted decisions in extreme panic. Under the pressure of life and death, people rapidly lose their judgment of what is truly important, and all rules collapse in an instant.

And once the dust settles, history is always written by the victors. The causal details of all events are reshaped, distorted, and simplified through the filter of the victors. For those outside the storm, penetrating the narrative fog written by the winners to restore the true mechanisms is nearly impossible.

Survival Guidelines for the New Generation of 14-Year-Olds: Stop Going to Wall Street for Quantitative Trading, Dive into AI

Eli Goldfine: As a 14-year-old, facing the impending era of superintelligence, what is your core advice for me if I want to create the greatest positive value in this world?

Sam Bankman-Fried: Ten years ago, the standard answer from the effective altruism community would have been "Earning to Give": go to Wall Street to engage in high-frequency trading, earn billions of dollars, and then donate the wealth to charity.

But times have completely changed. At this approaching singularity, high-frequency trading on Wall Street is no longer the most important battleground. The window between now and the full rise of AI is rapidly narrowing, and the value of deeply understanding and harnessing machine intelligence is climbing exponentially.

Whether it is engaging in cutting-edge AI alignment and safety research or finding a piece of traditional infrastructure and using AI to fully automate it, delving into the core of machine intelligence is the only true battleground worth fully committing to in this era.

ChainCatcher reminds readers to view blockchain rationally, enhance risk awareness, and be cautious of various virtual token issuances and speculations. All content on this site is solely market information or related party opinions, and does not constitute any form of investment advice. If you find sensitive information in the content, please click , and we will handle it promptly.

app_icon
ChainCatcher Building the Web3 world with innovations.