Good morning, and welcome back to Techzip AI Newsletter.

TL;DR: today’s issue tracks how frontier AI is becoming more capable while the systems around it become more deliberate: OpenAI is expanding frontier-agent monitoring into investigative agents; Mistral is opening the weights of its 1T-parameter Large 4 model; and DeepMind is examining how coordination, not just individual competence, could reshape the economics of intelligence. We also look at GPT-6 Luna, Anthropic’s evidence on agent influence over expert judgment, and the workflow shift from longer prompts to explicit goals, verification and selective delegation.