Keep pulling the thread on Project Glasswing.
Anthropic is withholding its newest model, Mythos, because the company believes it is too dangerous for public access.
Anthropic's model, Mythos, autonomously discovered thousands of vulnerabilities, including bugs in every major operating system and web browser.
Anthropic's Mythos model autonomously discovered a 27-year-old vulnerability in OpenBSD, which is used in firewalls and critical infrastructure.
Anthropic's model, Mythos, is capable of chaining together 3, 4, or sometimes 5 vulnerabilities to create sophisticated exploits.
Anthropic established Project Glasswing, an AI-driven cyber coalition with 40 companies including Apple, Microsoft, Google, Amazon, and JPMorgan.
Brian Gerstner believes Anthropic's Mythos and OpenAI's forthcoming model SpUD represent the beginning of AGI-level models with massive step-function improvements in intelligence.
David Sacks predicts that over the next 6 months, AI-driven cyber tools will detect a wide range of dormant bugs from the past 20 years across many systems.
Anthropic required high-volume users of its $200/month subscription plan who were accessing the service via OpenClaw to switch to the pay-per-usage API.
Shortly after changing its pricing policy for OpenClaw users, Anthropic announced its own agent technology, positioned as a safer alternative to OpenClaw.
David Sacks alleges that Anthropic systematically copied features from OpenClaw and integrated them into its Claude model.
Jason Calacanis believes the primary goal of frontier model companies is to eliminate the open-source agent product OpenClaw as a competitive threat.
Jason Calacanis predicts that open-source models will eventually account for 90% of all token usage, undercutting the frontier model industry.