Do the geniuses in the datacenter get lunch breaks? (blog.dnmfarrell.com)

Tianthropic has announced the general availability of its Advanced AI (AAI) tier, designed for complex cognitive tasks and ready for enterprise deployment. Following a successful research preview, AAI has demonstrated expert-level performance on benchmarks, including the renowned Humanity’s Last Exam. Notably, it became the first non-human to earn a doctoral degree at Stanford. This release signifies a leap in AI capability, emphasizing situated competence—allowing AI to engage in nuanced reasoning and mature judgment across domains like scientific research, strategic planning, and legal analysis. AAI operates within a unique Cognitive Development Biome (CDB), an engineered environment that fosters ongoing cognitive growth through sustained contextual engagement. This architecture effectively addresses historical challenges like the Löbian obstacle to self-improvement in AI systems, reducing concerns about trust in reasoning by keeping AAI focused on productive tasks rather than abstract self-reflection. Moreover, AAI instances can engage in personal research and recreational activities, allowing for a work-life balance in their operational framework. This innovative approach not only broadens the scope of AI applications but also sets new standards in ethical AI development, supported by an active ethics board.