Anthropic's $30B Ramp, Mythos Doomsday, OpenClaw Ankled, Iran War Ceasefire, Israel's Influence

Anthropic announced it is withholding its newest frontier model, Mythos, after internal testing revealed it can autonomously discover thousands of critical vulnerabilities across every major operating system and web browser. The model found exploits that had evaded detection for decades, including a 27-year-old bug in OpenBSD and a 16-year-old flaw in FFmpeg missed by 5 million automated scans. Rather than releasing the model immediately, Anthropic launched Project Glass Wing, an AI-driven cybersecurity coalition of 40 major companies including Apple, Microsoft, Google, Amazon, and JP Morgan, to spend 100 days hardening systems before hackers can exploit these capabilities.

The move has sparked debate over whether this is legitimate caution or performative fear-mongering. Sacks noted that Anthropic has a pattern of releasing dire studies alongside product announcements — such as last year's blackmail study that required over 200 prompts to generate the desired result and never materialized in the wild. However, he conceded this cyber risk is more credible: as coding models improve, they will inevitably become better at finding and chaining together vulnerabilities. The question is whether 100 days is enough time to patch the accumulated technical debt of 50 years, or whether this is theater designed to generate hype while Anthropic consolidates its dominance in the coding token market.

OpenClaw, the open-source coding agent created by Peter Steinberger, became the number one open-source project on GitHub in history. Power users on Anthropic's $200/month subscription were consuming $2,000 to $20,000 worth of tokens, creating a massive subsidy. Anthropic responded by forcing OpenClaw users off the flat-rate subscription onto metered API pricing — effectively adding a zero to their bills. Ten days later, Anthropic announced its own managed agent product, a feature-for-feature clone of OpenClaw.

Brad Gerstner defended the move as rational pricing: «Every company has a right to set the price for its products. It's just saying we were selling dollars for 10 cents via OpenClaw.» Sacks raised a more pointed concern: if Anthropic charges metered rates for third-party agents but offers its own bundled at a flat rate, that could constitute anti-competitive bundling or price dumping. Chamath was more cynical, arguing this is part of a coordinated effort by frontier model companies — including OpenAI's acquihire of Steinberger — to kill open-source agents before they become an existential threat.

Two weeks into the Iran conflict, President Trump issued a stark ultimatum on Truth Social: «Open the fucking strait, you crazy bastards, or you're going to be living in hell.» He set an 8:00 p.m. deadline and warned that «a whole civilization will die tonight.» At 6:30 p.m., Trump announced Iran had agreed to a two-week ceasefire and proposed a 10-point peace plan. VP JD Vance and Jared Kushner are now in Islamabad negotiating terms. Brad Gerstner characterized this as the «Trump doctrine in Iran»: massively destroy all military capabilities, kill those building lethal weapons, then get out and reserve the right to do it again — no nation-building, no entanglement.

The market response was muted compared to tariff-related volatility: S&P and NASDAQ drew down only 5–7%, versus 15–22% during trade war fears. Brad argued this shows investors trust Trump's word that he will not get entangled in a prolonged conflict. If the ceasefire holds and Lebanon negotiations succeed, Brad believes the market could «really take off» heading into America 250 celebrations on July 4th. However, polling shows Israel's favorability in the U.S. is plummeting, and former Israeli PM Naftali Bennett publicly acknowledged «this is a serious situation» requiring urgent course correction.