Jason Calacanis hosts with regular co-hosts Chamath Palihapitiya and David Sacks, joined by guest Brad Gerstner, CEO of Altimeter Capital and investor in both Anthropic and OpenAI. David Freeberg is absent this week.

The conversation covers Anthropic's unprecedented revenue growth and their decision to withhold the Mythos AI model due to cybersecurity concerns, including the company's Project Glasswing initiative to patch vulnerabilities before public release.

The discussion explores the competitive dynamics between Anthropic and OpenAI, including Anthropic's controversial decision to cut off OpenClaw's access while simultaneously releasing their own competing agent technology.

The episode addresses ongoing geopolitical tensions, including Trump's Iran ceasefire negotiations and concerns about Israeli influence on U.S. foreign policy, while also celebrating technological advances like X's auto-translation feature.

Anthropic Withholds Dangerous AI Model Over Security Fears

Anthropic is withholding its Mythos model, claiming it autonomously found thousands of vulnerabilities including a 27-year-old OpenBSD bug and 16-year-old FFmpeg exploit missed by 5 million automated scans.

"The model that we're experimenting with is, by and large, as good as a professional human at identifying bugs" and "has the ability to chain together vulnerabilities" - Dario Amodei in promotional video.

Project Glasswing unites 40 major companies including Apple, Microsoft, Google, Amazon, and JPMorgan for 100 days of AI-driven vulnerability detection and patching.

Sacks notes Anthropic has a pattern of "using fear as a way to market their new products," citing their previous blackmail study that prompted the model 200 times to get desired results.

"If you actually think that Mythos is capable of doing what it says it can do, two things are true. One is a very sophisticated hacker can probably do those things right now with Opus" - Chamath.

Anthropic's Historic Revenue Ramp Reaches $30 Billion

Anthropic's revenue run rate exploded from $1 billion in late 2024 to $30 billion by April 2025, adding $11 billion in March alone - equivalent to Databricks plus Palantir combined.

The company now has over 1,000 enterprise customers paying $1 million+ annually with only 2,500 employees, compared to Google's 120,000 employees at similar revenue levels.

"They added $4 billion of revenue in January, $7 billion in February, $11 billion of annualized run rates in March" - Brad Gerstner, highlighting unprecedented growth velocity.

Enterprise adoption exploded after Claude Code launch on February 25th, with companies viewing this as "labor augmentation and labor replacement" rather than just IT budget items.

Gerstner predicts Anthropic could exit 2025 at $80-100 billion revenue run rate, calling it evidence that "the TAM for intelligence is radically different than anything we've seen before."

OpenClaw Controversy Sparks Anti-Competitive Concerns

Anthropic cut off OpenClaw's access to $200 subscriptions, forcing users to API pricing after they consumed 100x more tokens than average subscribers.

Ten days after cutting OpenClaw access, Anthropic announced their own competing agent technology, leading to accusations of anti-competitive behavior.

"OpenClaw became a phenomenon, the number one open source project in history on GitHub" with users burning through $2,000-20,000 worth of tokens on $200 subscriptions - Calacanis.

Sacks raises concerns about potential "price dumping or bundling" if Anthropic charges flat rates for their own agents while forcing third-party tools to use metered API pricing.

Multiple competitors are emerging including Perplexity Computer, Hermes agent, and Alibaba's Quinn-based solution, with Elon promising "Grock Computer" from "MacroHard."

Iran Ceasefire and Foreign Policy Influence Debate

Trump negotiated a two-week ceasefire after posting "open the f***ing strait, you crazy bastards, or you're going to be living in hell" with an 8 p.m. deadline.

VP J.D. Vance and Jared Kushner are in Islamabad this weekend working on peace negotiations, with Trump receiving a "10-point proposal from Iran" as a "workable basis."

The New York Times reported that Netanyahu gave Trump a four-part pitch on attacking Iran during a February 11th White House meeting, with J.D. Vance warning of regional chaos.

General Dan Kaine reportedly told Trump: "Sir, this is standard operating procedure for the Israelis. They oversell and their plans are not always well developed."

Former Israeli PM Naftali Bennett tweeted polling showing Israel becoming "very unpopular in the U.S.," expressing concern about the need to "fix everything."

X's Auto-Translation Breaks Down Global Barriers

X's auto-translate feature using Grok enables real-time cross-language conversations, automatically translating posts and replies between users who don't speak the same language.

"This auto-translate feature has done more for understanding across borders than anything I've ever seen" and represents "the most impressive tech feature I've seen released in years" - Calacanis.

The feature surfaces content from Japan, Israel, France and other regions, allowing Americans to engage with foreign perspectives previously inaccessible due to language barriers.

"Base Japanese makes like Fuentes and Alex Jones seem tame" with extremely direct cultural commentary that journalists aren't translating or covering - Calacanis.

X now operates with 70% fewer employees than when Elon acquired it while delivering a superior user experience, proving "these companies can do more with less."