GPT-5.4 lands, the agent revenue race heats up, and NVIDIA goes after enterprise agents

OpenAI's new flagship beats human experts on professional tasks. ARR numbers for OpenAI and Anthropic both leak. And NVIDIA launches its own enterprise agent platform.

GPT-5.4 is the standout this week, but the bigger story is the financial scale of the leading labs and what NVIDIA’s enterprise push means for the rest of the market.

OpenAI releases GPT-5.4 with a major leap in computer use

My take: Scores 75% on the OS World benchmark for operating computer interfaces, beating the human baseline. Beats or matches human experts 82% of the time on professional task benchmarks. The community consensus is it's the best model right now. If you've been holding off on agent automations because reliability was poor, the bar just moved. Worth a fresh look.

Anthropic at $19B ARR, OpenAI at $25B

My take: Within a day of Anthropic's leak, OpenAI's number leaked too. Both companies are growing at speeds that would be unprecedented in any previous tech wave. The competition between them is driving genuinely better products and faster price cuts. The losers in this race will be the smaller labs that can't keep up on capex.

Google launches a Workspace CLI for agents

My take: An agent-first command-line interface for Gmail, Drive, Calendar, etc. This is a quiet but important move. It means AI agents can interact with your Google Workspace far more efficiently than scraping the UI or hitting the slower API. If your business runs on Google Workspace, this drastically expands what's automatable.

NVIDIA launches Nemo Claw, its own enterprise agent platform

My take: Built on Open Claw, with enterprise privacy, security, and access controls. CEO Jensen Huang's quote: 'Every software company in the world needs to have an Open Claw strategy.' Translation: the agentic shift is now considered foundational by the largest enabler in the industry. Time to take it seriously if you haven't.

Amazon wins court order blocking Perplexity shopping agents

My take: Amazon argued the agents fraudulently accessed user accounts and breached terms of service. This will be a defining precedent for how agentic e-commerce can operate. Worth watching closely if you sell on Amazon or run any retail business. The rules of engagement for AI agents on commercial platforms are about to be written.