Both labs' September 2026 disclosures on AI-assisted research have kept the market near even, with Anthropic at 52.5% implied probability. Anthropic highlighted Claude authoring over 80% of merged code by May and ongoing autonomous experiment proposals, while OpenAI detailed an "automated research intern" achieving 3.1 agent-workdays per human workday by mid-August, alongside chief scientist statements on sustaining progress toward recursive self-improvement. Neither has demonstrated or announced a closed-loop system where large language models independently drive full model advancement without human oversight, and both system cards and posts stress remaining gaps in capability and safety. Key upcoming catalysts include winter 2026 automation claims, new model evaluations, or regulatory pacing announcements that could shift trader consensus on which lab reaches verifiable first-mover status.