Top AI skeptics, including Dwarkesh Patel and the AI 2027 team, just drastically pulled their superintelligence timelines forward. The trigger?
Anthropic's amended SEC filing admitting it crossed its recursive self-improvement threshold—with Claude now leading 26% of internal model R&D (up from 1% in February). Meanwhile, internal models are hitting 38-hour autonomous task horizons, prompting figures like David Sacks and Congress to debate banning RSI entirely.
Are we at the tipping point?