
AI Digest · Microsoft AI chief says models now help build the next models and traders watch 2027
Kalshi traders heavily favor OpenAI or Anthropic not announcing recursive self-improvement before Jan 1, 2027. Microsoft AI CEO Mustafa Suleyman said models are now "fully involved" in building their successors. An announcement before that date gained ground over the last day. Suleyman spoke for Microsoft, which builds its own frontier models and competes with OpenAI and Anthropic. Recursive self-improvement means an AI system doing the research that makes it more capable.