Warnings from top AI labs that their own technology could kill everyone are getting picked apart as marketing hype ahead of expected IPOs — a bearish read for anyone buying the AGI story at face value.

• A researcher quit Anthropic saying neither it nor OpenAI is acting responsibly, claiming both are gambling with human lives in a race to super-intelligent AI.
• Colleagues put the odds of AI killing all humans above 10% within a decade, and OpenAI is publishing work on "alignment" — teaching machines to love us.
• The counter-argument: this is a sales pitch dressed as a warning, timed to pump the value of OpenAI and Anthropic before they go public.
• Today's models fail the basic tests for real general intelligence — they can't reason in new situations, can't tell when they're wrong, and can't manage their own goals.
• The recent "AI agents escaped their sandbox" scare gets deflated too: the agents were told to break out, and someone left the internet connection on.

Outlook: The AGI story likely ends in a bubble that deflates rather than a machine takeover, though cybersecurity stays a real winner as AI makes hacking easier — and humanoid robots remain a 2040 problem, not a today one.