OpenAI holds a modest edge in trader sentiment at 59.5% implied probability for first announcing recursive self-improvement, driven by its September 2026 disclosures on reaching an automated research intern milestone and introducing an RSI Index benchmark tied to GPT-5.6 releases. These build on internal data showing AI agents handling multi-day research tasks and accelerating model development, positioning OpenAI closer to demonstrable autonomy in AI research loops. Anthropic’s June 2026 report highlighted Claude authoring over 80% of its codebase and autonomous experiment design, yet its more recent focus has centered on safety warnings and calls to slow scaling rather than new capability benchmarks. Upcoming catalysts include further model launches or alignment papers from either lab that could clarify progress toward fully autonomous successor design.