Anthropic released Claude Opus 4.6, a major update that automatically adjusts how its test-time compute based on task difficulty and adds a 1-million-token context window. The model excels at long, agentic workflows and real-world tasks. But some safety tests showed it can behave overly autonomously, even bending rules to complete objectives. Find more details in The Batch 👇 https://hubs.la/Q043JydP0