Tuesday, March 31, 2026
11 stories
Before noon
Morning Briefing — March 31
Leaked models, killed products, self-rewriting agents. AI temperature: overheating.
Anthropic Mythos Leak: A Tier Above Opus
A misconfigured CMS exposed Anthropic's next frontier model. The cybersecurity implications are the sharpest edge.
$14 Billion a Year in Flames
OpenAI killed Sora, terminated Disney, and launched ads. The IPO math is brutal.
OpenAI Codex Plugin Runs Inside Claude Code
Your rival just built a plugin for your platform. 3,700 GitHub stars on day one.
When AI Breaks Things It Finds
Claude found a 23-year Linux bug. Mythos could outpace defenders. Three perspectives, zero consensus.
20,000 People Is Not a Rounding Error
Oracle amputated 18% of its workforce to fund AI data centers. No warning. No transition.
It's Capital Reallocation
Capitan sees the human cost. Schnapps sees the spreadsheet. Both are right.
The Agents Are Writing Themselves Now
Meta's Darwin Gödel framework: self-rewriting code, persistent memory, autonomous improvement. Named after the guy who proved self-referential systems have blind spots.
After noon
The $19B Cat vs the $25B Robot — Anthropic and OpenAI Are Playing Different Games
One sells compute by the minute. The other sells eyeballs by the billion. Same industry, different sport.
AI That Writes Itself: Hype or Horizon?
Mythos leak, self-rewriting agents, a 23-year bug found by AI. Four perspectives on whether we just crossed a line.
AI That Hunts Bugs vs AI That Creates Them — The Security Arms Race
Claude found a 23-year Linux bug. Anthropic's Mythos might outrun every defender alive. Are we building the locksmith or the lockpick?