Signal Drop — 22 April 2026

Anthropic's Mythos model found 271 zero-day vulnerabilities in Firefox, proving AI is now peer-level with elite security researchers. Plus: OpenAI's vertical models, Amazon-Anthropic's $5B deal, and the coding platform arms race. Stories covered: • Anthropic Mythos finds 271 zero-days in Firefox • OpenAI launches GPT Rosalind and GPT-5.4 Cyber • Amazon invests $5B more in Anthropic for custom chips • SpaceX pursues $60B Cursor acquisition • John Ternus named Apple CEO successor • Codex hits 4 million weekly active users • Data center construction delays threaten compute capacity • AI-generated music flooding streaming platforms • Allbirds pivots to AI infrastructure business • Deep dive: AI security research at human parity: How Mythos's 271 Firefox zero-day discovery reshapes vulnerability discovery timelines and the offensive-defensive balance in cybersecurity Sources: OpenAI News, Google AI Blog, DeepMind, The Verge, Ars Technica, VentureBeat, TechCrunch, MIT Tech Review, Hacker News, TLDR AI, Simon Willison. Produced with AI voice technology and human editorial oversight.