Anthropic has commenced grayscale testing for its Claude Opus 5.2 within the Claude Code environment. This latest iteration bypasses version 5.1 and introduces substantial enhancements to its foundational capabilities. Key improvements include rapid response times, succinct output, elimination of sluggishness, and the capacity for autonomous task iteration and refinement. Users can verify their eligibility for the grayscale release by entering the prompt "Reset Bro Tibo." Furthermore, Anthropic is gearing up with three strategic plans to rival OpenAI's GPT-6 Astra. Claude Fable 5.2 is poised for launch as early as the end of this month or next. Internally, the company has already leveraged a model dubbed Model 2 to compose the majority of its code, with this model outperforming the current top model on the CoBench v2 leaderboard by a significant margin. Additionally, the RSI model has the potential to replace 85% of the research team's workload, surpassing Model 2 by 22 points and heralding a new era in AI autonomous iteration.