What we are learning while deploying Claude, AI employees, and automation in real organisations.
Claude Sonnet 4.6 matches Opus 4.5 intelligence at $3/M tokens instead of $15/M, with massive improvements in computer use, coding, and long-context reasoning. This changes the economics of AI automation.
Google releases Gemini 3.1 Pro with 77.1% on ARC-AGI-2, more than double the reasoning performance of its predecessor. This isn't incremental progress—it's Google throwing down the gauntlet in the reasoning race.
Anthropic just launched Claude Sonnet 4.6 with significantly improved computer control capabilities. The new model can navigate spreadsheets, fill web forms, and execute complex multi-step tasks — bringing us closer to AI agents that don't just chat, but actually work.
Zhipu AI's new GLM-5 is a 745 billion parameter model trained entirely on Huawei Ascend chips. It's open source, it's powerful, and it proves China's AI ecosystem can thrive without American hardware.
Anthropic's latest release brings frontier-level reasoning, computer use, and coding capabilities to its mid-tier model--changing the economics of AI deployment for enterprises.