Third-party dispatches, not graded receipts
Anthropic warns it is 'very close' to recursive self-improvement capability
Anthropic has stated it is nearing RSI (recursive self-improvement) capability, a milestone relevant to forecasts about superhuman AI systems and AGI timelines.
“Anthropic warns that it’s very close to RSI capability, aka self-building AI”
Bears on: superhuman-coder-2027 · metaculus-weakly-general-ai-2028 · hassabis-agi-5-to-10-years-2025