Recent progress in AI-assisted coding and research at OpenAI and Anthropic has fueled trader focus on recursive self-improvement, or RSI, where large language models autonomously accelerate their own development. Anthropic reports Claude authoring over 80% of merged code by May 2026, with engineers shipping eight times more output quarterly, while OpenAI achieved its automated research intern milestone in September with agents logging 3.1 workdays per human day. Neither lab has announced full autonomous RSI, however, with OpenAI stating fully closed-loop systems remain absent and both tying longer-term targets like automated researchers to 2027–2028. Competitive dynamics, safety concerns, and external calls for oversight on automation timelines continue to shape market-implied odds amid rapid but incremental gains.