Anthropic's latest regulatory filings and benchmark studies indicate that recursive AI self-improvement may be arriving sooner than expected. At the same time, frontier models are matching top human political forecasters, severe infrastructure shortages persist, and public backlash against data centers is accelerating.