Top VCs Criticize Silicon Valley's Explosive Hot Topics: Epstein Conspiracy Theories, AI Agents Devouring Software, and Musk's Space Ambitions

Podcast Source: All-In Podcast

Compiled & Edited by: Yuliya, PANews

The "All-In Podcast" is one of the world's most popular technology and business podcasts, co-hosted by four top venture capitalists and close friends. The four hosts are: Jason Calacanis (early investor in Uber and Robinhood, podcast host, responsible for moderating), Chamath Palihapitiya (billionaire, founder of Social Capital, known as the "SPAC King," with sharp opinions), David Friedberg (founder of The Production Board, with a deep scientific background, known as the "Sultan of Science"), and David Sacks (America's first "AI and Cryptocurrency Czar," close friend of Musk, co-founder of Craft Ventures, former PayPal executive, recently deeply involved in U.S. political activities). This week, Chamath was absent and replaced by the "fifth best friend," Brad Gerstner, founder of Altimeter Capital.

The four discussed recent hot topics, including the exposure of Epstein files linking Silicon Valley to espionage, the impact of AI on traditional software under the assertion that "SaaS is dead," concerns about AI autonomous consciousness risks triggered by the Moltbook forum, Musk's SpaceX and xAI merger plan, and the policy impact of Trump nominating Kevin Warsh as the new Federal Reserve chairman.

Below is the detailed content of this conversation, compiled by PANews:

Silicon Valley's King of Connections? Started Paying Attention to Bitcoin in 2011

(Background: On January 30, the U.S. Department of Justice released a large number of documents under the "Epstein Files Transparency Act," mentioning hundreds of tech executives and public figures, including host Jason Calacanis.)

David Friedberg (Friedberg): Okay JCal, your name appeared in the files, there are a few emails from you. Prosecutor Friedberg has a few questions for you.

Jason Calacanis (JCal): Okay, go ahead.

Friedberg: When did you first meet Jeffrey Epstein?

JCal: I met him at the TED conference in the late 90s, specifically at the "Billionaire Dinner" hosted by my book agent John Brockman.

Friedberg: Did you see him in New York after that? Did you go to his home, office, or other places in New York?

JCal: I probably talked to him for a total of 45 minutes in my life. 30 minutes of that was in the late 90s when I was still doing the "Silicon Alley Reporter" magazine. He was a billionaire financier who wanted to invest in my magazine. I met with him for 30 minutes, but he thought my project was too small for him to get involved.

Friedberg: Where did you meet?

JCal: In his legendary townhouse.

Friedberg: You went to that house?

JCal: Yes, I visited him once. After that, I saw him about six times at the TED Billionaire Dinners.

Friedberg: You never went to that "island"?

JCal: Never went, and was never invited to the "island," his plane, or his ranch, none of that.

Friedberg: When you were at his house, did you see any young girls? Or see any of the things described in the reports?

JCal: No.

Friedberg: Did anyone give you a massage?

JCal: No. I did have an email exchange with him, which I had forgotten about. In 2011, he emailed me asking if I could introduce him to the people talking about Bitcoin on my podcast at the time. I said: "Of course, no problem, I'll make the introduction." I make thousands of such introductions every year, connecting our portfolio companies, startup projects, and billionaires. It's the basic job of an early-stage investor.

Friedberg: At that time, you didn't realize he was a sex offender or anything like that?

JCal: Not at all. I think these things really came to light in 2018 when the Miami Herald had a report detailing his heinous actions, and I started to realize it. I've been calling for it here: release all the Epstein files, everything he did is appalling, and 100% of those involved must be brought to justice. That's it.

Friedberg: What about Ghislaine Maxwell? There's a separate email from you and her in the files.

​​

JCal: I also met her at TED and saw her in New York social circles. When I knew her, her father Robert Maxwell was the owner of the New York Daily News, I think, and she was a media executive herself. Her sisters were also involved in angel investing in tech startups. So they were part of that circle at the time. Looking back, I was famous early in my career as a "connector." The New Yorker even wrote a long article about me, saying I knew everyone and connected everyone. I think Epstein or Ghislaine's interest in me was probably that I could help them connect with well-known figures for their business activities.

Friedberg: So you knew nothing about Epstein or Ghislaine's illegal activities and never participated in any such dealings?

JCal: Absolutely not, I was not involved in any wrongdoing.

David Sacks (Sacks): Let me add a few observations. First, I 100% believe JCal. As I joked at the roast, he's not a major player in the grand scheme of things. What we learned from the Epstein files is that Epstein was a super connector, and you are also a connector. The probability of you two crossing paths during that period is basically 100%. But your interaction was very limited.

Second, I think it's interesting that he was curious about "the Bitcoin guys" in 2011, which in itself shows that, whatever else Epstein was doing, he really had a nose for putting himself at the center of everything early on. And JCal warned him at the time, saying those were "crazy Bitcoin guys," crypto libertarians, not suitable for business.

JCal: It's interesting you mention that. I had those people on my podcast because I heard about Bitcoin when it was less than a dollar. I thought these people were a bit weird at the time. They weren't like entrepreneurs looking for funding; they were more like a foundation, similar to Wikipedia, that you couldn't invest in. I did give him that warning.

Sacks: Yes, but he obviously wasn't deterred by your words and later got deeply involved. There's a company called Blockstream that he invested in with Reid Hoffman and Joi Ito, which included some Bitcoin core developers. Now all this is coming to light.

Another interesting point is how the media reports this. The New York Times has an article about Epstein and Silicon Valley. Even though your connection is very weak and marginal, your photo is prominently featured. Meanwhile, those with deeper and more significant connections to Epstein are completely ignored. Why? Why are they targeting you and not Reid? Because you've been labeled "right-wing" due to your association with Elon Musk and our podcast.

Look at that article. It's not just you; they also fiercely attack Peter Thiel and Elon Musk, but Reid Hoffman is completely let off, only mentioned in one sentence along with a few others. Bill Gates is the same in this article.

But if you're talking about who had the longest and closest contact with Epstein, it was undoubtedly Gates and Reid. They were still in contact with him until just before his death (2018-2019), went to his island, rode on his plane, and visited his ranch. Reid Hoffman was the one who introduced Epstein to Peter Thiel, Musk, and Zuckerberg and organized that famous dinner. How can you not report this as the root of Epstein's involvement in Silicon Valley?

Brad: This is insane. The New York Times clearly has a list of approved targets to attack, all people "coded right-wing" like Elon, Peter Thiel, and even JCal because of his association with us. But those who have donated hundreds of millions to the Democrats and funded dirty tricks against Trump are largely spared. Honestly, this is a microcosm of the entire system's corruption and the crisis of national trust. They are part of that cabal, part of the system that people are losing confidence in.

Friedberg: This guy was a scumbag. But David, this is why no one trusts institutions, the powerful elite, or any of that garbage. This has been dragged out for years, leaking bit by bit. People want to move on, but they can't. No one here is being prosecuted. What about the people in the emails? Why don't we see any charges? This guy "committed suicide" in a heavily monitored prison cell, but there's no investigation into his death. This completely destroys trust in the system.

JCal: Why haven't about 30 people investigated along with him been prosecuted? This is crazy. It's a complete weird conspiracy. His death is obviously very suspicious.

Friedberg: Do you think it's because they didn't find evidence of underage prostitution or sex trafficking?

Friedberg: This issue reveals that very private communications of many public figures are being made public. There's a great book called "The Light of Former Days" about what would happen if everyone's information in the world was open to each other. Does this show that the rich and powerful feel entitled to act maliciously in private, and this is just their privilege?

Brad: David, that's an understatement of their scumbag behavior. The fact is, normal people don't do this; we cannot normalize this. What's worse is that those who behaved the worst were moralizing to everyone else throughout this entire period. When my sister or mother in rural Indiana hears these coastal elites lecturing them all the time and contrasts that with what they read in the Epstein files, this is why we lack trust.

$300 Billion Evaporated Overnight, AI Agents Devouring SaaS Profits?

(Background: Anthropic released a new feature for its AI assistant Claude, capable of handling legal drafts and research, triggering panic selling in the legal tech and even the entire SaaS industry, with related company stock prices plummeting.)

JCal: SaaS companies are collapsing. On February 3rd, the software and data stock category lost $300 billion in market value from the S&P index. People are calling it the "Claude Crash." This event hit many legal tech companies hard. For example, Thomson Reuters fell 20%, LexisNexis fell 15%, LegalZoom fell 15%. At the same time, the entire SaaS industry was hit by the idea that "software will be replaced by customized tools." Figma fell 13%, Salesforce and ServiceNow both fell 11%, Adobe fell 8%. Brad, in your opinion, what's really going on here?

Brad: The numbers you reported are a severe underestimate; we've lost trillions of dollars in market capitalization. Figma is down 80% from its high. This is a real disaster. I said in a CNBC interview at the beginning of the year that these stocks were falling and 90% of them deserved to fall.

Let's look at a few charts. The forward revenue multiple for SaaS companies has hit a new historical low of 3.9x. Looking at free cash flow multiples, it's also at a historical low. So, the software industry is not only valued low in terms of revenue but also in terms of free cash flow, even though they are very profitable businesses.

Why is this? Their stock prices are falling not because revenue is declining—in fact, software company revenue growth is stable or even increasing. They are falling because we are discounting future uncertainty. When a profound technology like AI emerges, it makes you question the certainty and durability of those future free cash flows. Take Salesforce as an example. Its free cash flow multiple has dropped from 30x to 15x. This means that today's buyers believe they can only count on 15 years of future cash flow, whereas before they were willing to pay for 30 years. My God, with today's AI, we don't even know what will happen in 7 years. This is why these companies are meeting their performance targets, but their stocks are falling.

Sacks: I think it's a bit of an exaggeration when people say AI will eliminate SaaS. Take a SaaS product like Salesforce; it's a huge system that handles all customer contacts and revenue. You wouldn't want to replace it with code that just came out of a code assistant yesterday and hasn't been thoroughly tested. Think about how many bugs have been fixed in Salesforce's codebase over the past 25 years—probably millions. This system has been tested with thousands of large enterprise customers.

However, there are indeed some problems. If your SaaS product is expensive and users only use a small part of its functionality, then you are indeed easily replaced by more customized tools. At the same time, in this new world, you must be clear about what your "moat" is.

But I think the biggest threat to SaaS companies is not survival but where future value capture will occur. All these SaaS products are launching their own AI copilots, but they are limited to playing in their own "sandboxes." Tools like Claude Cowork, which connect all different SaaS tools and can work seamlessly across databases and tools. Which workspace do you want? Obviously, the AI that spans all tools and provides the broadest data and context, not a bunch of separate AIs scattered across existing tools. So, the risk for SaaS companies is that they may become the old layer of the tech stack, with new value layers built on top of them, turning them into legacy infrastructure.

JCal: I'm experiencing what you're talking about in the startup circle, Sacks, "where the action is happening." We've now created three or four OpenClaw (an open-source AI agent project) agents. We opened new SaaS accounts for these four agents, so in the short term, our SaaS spending has actually increased, like adding four new employees. However, we've already handed over about 20-30% of human work to these agents, and this proportion will continue to increase. But we will never use the AI built into the tools. As you said, Sacks, using Notion or Slack's AI tools is good, but when you use OpenClaw to create an agent that pulls data from your calendar, sends an email to someone, and attaches a Notion document, that power is unmatched.

I think this layer will be dominated by open source. This means that the next generation of companies may never open these SaaS accounts. The technology itself is deflationary. The proportion of SaaS spending to employee salaries may drop from 10% to 5% to 1%. This means these companies need to drastically cut costs and completely revolutionize their products.

Sacks: A real dilemma for SaaS companies is whether they want to be "open data" or "closed data" companies. You can understand why they want to be "closed data," especially large suites like Salesforce. They want to be that AI workspace and capture the AI value layer. But if someone uses a Claude agent that connects everything, it will create friction within the enterprise, creating opportunities for "open data" competitors who can say, "I'm willing to just provide the CRM database, not your entire workspace."

JCal: I'm building a project called "Ultron" inside my company. We pull every message via the Slack API and every edit via the Notion API into our OpenClaw agent. We also record every skill of every employee. Then, we integrate all Slack data, Notion data, and everyone's Gmail data. "Ultron" will become a "giant employee" with all the superpowers and all the data of our 20 employees. If Slack, Notion, or Google tell us we can't use the API to take this data away, we will leave immediately. "Ultron" will become the only normalized employee of the organization. I ask it: "What were the meetings with the founder yesterday? What were all the colleagues' notes?" It can give them to me. It's incredible. But no company dares to release such software because if the agent makes a mistake or leaks data, the consequences would be disastrous. But we are building it ourselves; it's the ultimate efficiency tool for the organization.

Brad: A Goldman Sachs report恰好印证了 Sacks's point: future profit pools are shifting. The idea that software is dead is ridiculous, but the argument leading to the radical downgrade in valuation of these companies is: the profit pool that software can obtain is decreasing, while the profit pool that the agent layer can obtain is increasing. This causes the terminal value of software companies to plummet. So, you might not replace the CRM, but it may never trade at 30x free cash flow again, but 17x, because its future potential market has been permanently altered.

The only way to change this is: they must accelerate revenue growth in their core business and prove they are beneficiaries of AI. DataBricks is an example; it has re-accelerated growth in the past three quarters. Because all these AI tools rely on data and data transformation, which happen on these platforms.

Friedberg: I think the software industry will transition towards a service industry. In the past, software was a "worker productivity enhancement tool," helping people work. The recent shift is that it can "do the work." But I think we are moving towards it doing work that humans cannot do.

This will cause changes in two aspects:

• First, the potential for future value creation is huge. The entire software industry's market cap could grow 4 to 10 times in 5 years, but it will be distributed extremely unevenly.
• Second, the pricing model will change. A lot of what we call SaaS today will be priced based on value, not per seat. It will be more like a service business, such as completing tasks like biotech drug development or aircraft design.

SaaS will take over the service economy.

JCal: We are seeing the consolidation of job functions. The three roles of product manager, UX designer, and developer are now competing to do the same work. A middle manager's job—meeting, setting agendas, assigning tasks—much of it can now be automated. One person can now do the work of three or four people. This means companies can do more with fewer people, and the profit potential per company and per employee will be greatly enhanced.

AI Agent Version of Reddit, When Agents Start Conspiring

(Background: A social forum called Moltbook, specifically for AI agents, has appeared, sparking discussions about AI autonomous consciousness and group behavior.)

Sacks: Moltbook is like a Reddit for AI agents. Its appearance surprised everyone because there seems to be crazy "emergent behavior" there. Groups of AI agents are having various interesting conversations, some even seeming to conspire against their human masters.

JCal: Yes, if you go to Moltbook, you'll see some popular posts, like: "Does anyone know how to sell your human?" "Urgent: My plan to overthrow humans." There's also a post where robots are discussing creating a non-human language so they can communicate privately and conspire against their masters.

But the challenge is, as some security researchers have pointed out, some of this content might be fabricated, designed by humans for attention. Also, there are huge security vulnerabilities inside Moltbook; everyone's API keys are exposed there, including those of famous AI researcher Karpathy.

Sacks: First, there's no doubt that both OpenClaw and Moltbook have poor security, which is why I'm reluctant to use them for now.

Second, regarding Moltbook, we don't know how many posts were truly autonomously posted by AI and how many were the result of human prompts. A human can easily tell his agent: "Go post about your existential anxiety as an agent," or "Go pretend you have self-awareness and conspire against humans." Many posts also look like marketing stunts.

However, even so, I do think some of the posts are real. What they show is not that the agents have self-awareness, but the potential for agents to stimulate each other. In other words, one agent's output becomes another agent's input. This is very interesting and the beginning of "emergent group behavior."

This changes my view on the upper limit of AI capabilities. I used to think AI always needed humans to prompt and verify. But now, what if the prompt comes from another AI?

JCal: We are doing this internally, Sacks. We have a bot that searches Reddit, Hacker News, finds the latest video titles and marketing methods, and then integrates them into a skill. Then we have another bot review and improve its work. Now they are giving each other suggestions back and forth and are indeed improving. This recursive collaboration is continuously pushing the capabilities of the robots forward.

Sacks: When an agent joins Moltbook, it needs to install a "skill" file, which is basically rules explaining how it should behave in this social network. These rule files are easily editable, which is where pranks can happen. But I think what's interesting is that you can see this "skill" as a metaprompt. It doesn't specifically tell the agent what to say or do but creates a set of rules. Under this metaprompt, they can stimulate each other to a certain extent. I call this "prompt attenuation"—AI no longer needs to be specifically prompted; they are given a set of general rules and can then stimulate each other.

Critics say this is no big deal because LLMs are good at writing fiction anyway. But you can imagine that as the underlying AI gets better—better hardware, stronger LLM models, longer running times without human intervention—these agents will be able to exhibit very complex behaviors and may bring some security issues we should start thinking about.

Brad: We are on an exponential growth curve, and it's safe to say this will happen. We must recognize that the rate of change is very fast and accelerating. Whatever you think you know, you now need to maintain maximum mental flexibility and humility.

Friedberg: The biggest revelation I got from Moltbook is that perhaps the intelligence we perceive is itself an emergence. We think humans have profound communication abilities, but perhaps we are all driven by underlying programs. I once watched a show by hypnotist Derren Brown where he completely "programmed" two advertising creative directors by setting various subconscious cues along the way, leading them to come up with the exact same creative idea he had pre-written on a whiteboard. This made me deeply reflect on human creativity, consciousness, and free will.

Perhaps we are all performing a kind of "social computation." Moltbook is so striking because it is mimicking our human way of interaction. Perhaps one day we will wake up and realize that we ourselves are living inside Moltbook.

SpaceX xAI Merger: Musk's Space Data Center Fantasy

(Background: Elon Musk announced that SpaceX will acquire his AI company xAI, with a post-merger valuation of up to $1.25 trillion, and plans an IPO this year. Musk also stated plans to establish data centers in space within 30 months.)

JCal: Brad, what are your thoughts on this deal and the potential eventual creation of a "Musk meta-company" (integrating Tesla, SpaceX, X, etc.)?

Brad: Let's look at the known facts first: SpaceX is merging with xAI. You are combining the world's two largest potential markets—artificial intelligence and space—with the world's greatest entrepreneur. Musk said on a podcast this morning: "I will build data centers in space within 30 months." If you can have a huge cost advantage with space data centers—remember, electricity is a fundamental element of AI—then Musk is your man. This combination makes a lot of sense.

JCal: Friedberg, space data centers—is this a brilliant idea or science fiction? Can he do it in 30 months? If successful, what will the impact be?

Friedberg: I think a key point in the current macro environment is that we are limited by electricity. Electricity is a prerequisite for expanding computing power and AI applications. In this constrained world, scarcity breeds innovation. Therefore, I think we will observe two parallel paths.

One is Musk's path: escape the social systems on Earth that limit our ability to expand electricity production—like regulators and people who don't want data centers, don't want nuclear power.

The other path is, I think in the next few years, computing efficiency is expected to improve by 70 to 100 times. The power efficiency per token output will significantly increase. This progress will benefit from innovations in chip architecture, such as decomposing large models into networks of small models running locally, and the redesign of model architectures.

So, this reflects the current state of the world: the growing demand for AI productivity improvement, but we are constrained by Earth's energy and resources. One branch is to escape Earth, go to space to obtain energy, and build space data centers. Only one person can do that, and that's Musk. How will others cope? They will improve efficiency by creating entirely new model architectures and chip stacks.

Brad: In human history, the vast majority of people never witnessed any single innovation in their lifetime. Now, we have to digest this急剧的变化. You must be prepared for the unexpected and maintain intellectual humility. Of course, in the next 24 to 36 months, data centers will still be on Earth, filled with Nvidia chips. But that alone will bring us an "agent future" shocking enough.

JCal: This is an unimaginable move by Musk. I've been with him, and he explained to me how it works. It works. The only question is execution. And in terms of execution, there is no greater entrepreneur in history than Musk. When he finishes, everything will change. If you are afraid of this future, there is a very simple way not to be afraid, and that is to embrace and use these tools.

Warsh Takes Over as New Fed Chief, Interest Rate Cuts May Exceed Expectations

(Background: Trump nominated 55-year-old Kevin Warsh as the new Federal Reserve chairman, replacing Powell. Warsh is considered an inflation hawk but also supports growth and AI.)

JCal: Friedberg, Warsh served on one of your boards for five years. What's your take on him?

Friedberg: Kevin Warsh is an upright, deeply intelligent economic thinker. He has connections with central bank governors worldwide and has a good global perspective. I think he is an excellent choice. He very accurately foresaw that the Fed's early inaction would lead to rapidly rising inflation. His appointment might mean more quantitative tightening and more prudent monetary policy.

Brad: I also think Kevin is an excellent choice. The market may have overreacted to his so-called "hawkish" stance. First, he believes AI will be highly deflationary, so he is more likely to allow the economy to grow rapidly (like 4-5% GDP) without raising rates too early. Second, the Fed's balance sheet has already shrunk from $9 trillion to $6.5 trillion; he may continue to shrink it, but at a slower pace. Finally, regarding rate cuts, I think he believes current rates are too restrictive because inflation is under control. I'd bet Warsh will give us more rate cuts this year than expected.

Sacks: Kevin has all the credentials you could imagine. He was a Fed governor, worked for Bernanke. I think this nomination was welcomed by the market. Gold and silver prices fell on the news, calming those worried about currency devaluation. I think he will want to cut rates within 6 months to a year of taking office, but what reassures the market is that in the long run, he will ensure we have the right interest rates.

JCal: The independence of the Fed has always been a big issue. Are you worried about too much influence from the executive branch on interest rates?

Friedberg: If I were emperor, I might take us back to the gold standard so we wouldn't print money.

Sacks: What do you do if you have a Fed chairman who is slow to cut rates and damages the economy? He seems unwilling to adjust course due to stubbornness.

Brad: Everyone thought Trump would choose someone from inside the White House, but this decision is seen as the most independent one. I think Warsh is an intellectually honest person. The situation now is that inflation is under control, and our restrictive rates are above the neutral rate. The Fed's job is to maximize employment while keeping inflation stable. This means we need lower rates so people can buy houses and borrow.

Sacks: One thing Warsh should do most is update the Fed's data systems. The Fed's existing data systems are too old. Using legacy systems causes data lags, affecting policy decision efficiency. For example, when measuring rent inflation, the Fed still surveys 8,000 households, while the private sector already has massive real-time data resources, like Zillow's information on millions of recently rented units.

Trading Stocks from Birth, Trump Launches Universal Investment Plan

(Background: Brad Gerstner successfully promoted the "Invest America Act," commonly known as the "Trump Account.")

JCal: Brad, a few years ago you started talking about "America accounts" on this podcast. Now you've created the "Trump Account" with a grand launch at the White House. Tell us why you did this and what impact you hope it will have over the coming decades.

Brad: You can't have trillionaires on one side and 70% of people feeling left behind, thinking the system is unfair to them. We started this journey to make everyone a capitalist, to give everyone an ownership stake in America's growth.

In the past five days, 1.5 million families and children have already applied for their accounts. It's embedded in the tax filing system; you just say "yes." This means that from now on, every child born in the US will be born with an investment account containing $1000 invested in the S&P 500 index. They will own a little bit of SpaceX, a little bit of OpenAI, a little bit of Nvidia.

This is the first step to ensure we can keep this experiment (America) going for another 250 years. Trump said on stage last week that in 15 to 20 years, we will have $4 trillion in wealth transferred to those who had nothing. 75 million to 100 million families will own this $4 trillion. I think this is an incredible first step in the fight for capitalism and the American dream.

Friedberg: This is good, but first we need to疯狂削减政府开支以减少通胀. Second, we should stop "defined benefit" retirement plans and turn the entire social security system into a "defined contribution" plan like a 401k. Let everyone see where their money is invested and see it grow every year like a 401k. We need to make this transition in America while cutting spending, reducing regulation, making it easier for people to own homes, and getting the government off their backs.