Anthropic’s $30B Ramp, Mythos Doomsday, OpenClaw Ankled, Iran War Ceasefire, Israel's Influence
The hosts discuss Anthropic's withholding of their powerful Mythos model due to cybersecurity risks, the company's unprecedented $30B revenue ramp, and current geopolitical tensions including the Iran-Israel ceasefire negotiations.
Summary
The episode covers several major topics starting with Anthropic's decision to withhold their new Mythos AI model, claiming it's too dangerous for public release due to its ability to find thousands of software vulnerabilities and chain them together for sophisticated exploits. The hosts debate whether this is legitimate safety concern or marketing theater, with Brad Gerstner defending it as responsible self-regulation while Chamath expresses skepticism about the timeline for fixes. The discussion then shifts to Anthropic's explosive revenue growth, reaching a $30B annual run rate in just over two years, driven largely by enterprise adoption with over 1,000 companies paying $1M+ annually. They analyze the competitive dynamics between Anthropic and OpenAI, noting how Anthropic gained ground through focus on coding capabilities. The episode also covers the controversy around OpenClaw being restricted by Anthropic after they launched a competing agent product. Finally, they discuss the current Iran-Israel conflict, ceasefire negotiations, and American concerns about Israel's influence on US foreign policy, with JD Vance and Jared Kushner heading to Pakistan for peace talks.
Key Insights
- Anthropic's Mythos model autonomously found thousands of vulnerabilities including 20-year-old exploits missed by security audits for decades
- Brad Gerstner argues Anthropic deserves credit for self-regulating rather than releasing a potentially dangerous model and relying on government intervention
- David Sacks claims Anthropic has a proven pattern of using fear tactics to market new products, citing their previous blackmail study
- Chamath believes this is mostly theater, comparing it to GPT-2's overhyped rollout in 2019 that turned out to be a 'nothing burger'
- Anthropic's revenue run rate has exploded from $1B in late 2024 to $30B by April 2025, representing the fastest revenue growth in tech history
- Over 1,000 enterprises are now paying Anthropic more than $1 million annually, representing the most coveted customers in enterprise software
- Brad Gerstner reveals Anthropic is achieving this growth with only 1.5-2 gigawatts of compute while being massively compute constrained
- David Sacks argues the TAM for intelligence is dramatically larger than any market seen before, making this not a zero-sum competition
- Anthropic restricted OpenClaw users from their $200 subscription plans, forcing them to pay per API usage, then launched a competing agent product
- Jason Calacanis believes the entire frontier model space is coordinating to kill OpenClaw because it represents an existential threat as an open-source alternative
- According to a New York Times report, Netanyahu pitched Trump a four-part plan for attacking Iran, with JD Vance warning it could cause regional chaos
- Former Israeli Prime Minister Naftali Bennett expressed concern about Israel becoming unpopular in the US, acknowledging serious work needed to fix the relationship
Topics
Full transcript available for MurmurCast members
Sign Up to Access