Three stories worth your time today. Anthropic just blew past OpenAI in valuation. Claude Opus 4.8 shipped a month and change after 4.7. And Google quietly turned on its agent answer to ChatGPT.
Let me walk through what each one actually means.
Anthropic Hits $900B and Passes OpenAI
Anthropic closed a round north of $30B at a pre-money above $900B. Sequoia, Dragoneer, Altimeter, and Greenoaks each kicked in around $2B. That makes Anthropic the most valuable private AI company on the planet, ahead of OpenAI for the first time.
This matters because the valuation gap between these two has been the running scoreboard for the whole industry. OpenAI was always the one with the bigger number. Not anymore.
My take: investors are pricing in the enterprise story. Anthropic is the one big labs are happy to plug into their stack without sweating headlines. OpenAI is still moving more product to consumers, but the smart money looks at code, agents, and the work people get paid to do. That is where Claude lives.
Claude Opus 4.8 Ships 42 Days After 4.7
Anthropic dropped Opus 4.8 on May 28. It is the fastest gap they have ever shipped between Opus releases. The headline stat is that 4.8 is roughly four times less likely than 4.7 to let bugs in its own code slip past without flagging them.
For folks shipping software with these models, that number is the one to circle. Self review is the difference between a model that helps and a model you have to babysit.
My take: the cadence tells you everything. Six weeks between Opus releases is not a roadmap, it is a fire hose. If your team built around 4.7 last month, you already need a plan to test 4.8 this week. The half life of any frontier model build is shrinking by the day.
Google Gemini Spark Goes Live for AI Ultra
One week after Google announced Spark at I/O, it turned on for US Google AI Ultra subscribers on May 29. Spark is Google's swing at a personal agent. It reasons across your connected apps, takes actions for you, and runs in the background through the day.
This is Google playing to its strength. Gmail, Calendar, Drive, Docs, Maps, all wired into one agent that already knows your life. OpenAI and Anthropic have to ask permission to touch your data. Google already has it.
My take: the agent war is a distribution war. The best model does not win, the model that lives where the work already happens wins. Google has a thirty year head start on knowing what is in your inbox. Spark is them finally cashing that check.
That is the day. Three moves, one direction. The race is not slowing down.