The 10-year Treasury yield closed at exactly 5% today. As we discussed yesterday, that's the level that stopped the Fed in October of 2023, when Powell looked at a 5% long bond and told the New York Economic Club that financial conditions had "tightened significantly."

The bond market had done the work. He stopped the tightening cycle, and within two months the 10-year was under 4%.

Kevin Warsh made the same observation himself in July, in his own press conference: "we haven't done much in 42 days. The markets have done quite a bit." He's talking about the rise in longer term bond yields. 

That said, the market says he hikes tomorrow. The bond market has given him reason not to.

Now let's talk about something bigger than the Fed meeting — the AI doom campaign.

Remember, it was just twelve days ago, OpenAI released its GPT-6 model.

Three days later, the man running the most valuable, and most important company in the world, said this about it …

AGI is "Artificial General Intelligence." GPT-6 itself defines it as "AI that can learn, reason, and solve problems across a broad range of tasks at roughly human level or better."

And this is the stage where the AI begins to autonomously improve itself, which creates a path to "superintelligence" — AI far beyond human capabilities.

Now, after this marinated a few days, one of the great hedge fund traders of our time, Paul Tudor Jones penned an op-ed in the Wall Street Journal sounding the alarm on AI safety. 

Was he talking his book? Very likely. 

By Saturday, the two leading frontier model CEOs were calling for a slow down on AI model development.

The media and politicians have since amplified the case.

But keep in mind, AGI has been declared by the man who sells the chips that AGI will need in unlimited supply. Meanwhile, the margin of performance between the frontier models, as measured by the third party Artificial Analysis Intelligence Index, is extremely tight.

Notice the cluster of the top models circled in the graphic below. This has GPT-6 leading, but at a level well below the performance reported from OpenAI's internal tests.

So, the American labs continue to lead: OpenAI, Anthropic, Google. But the Chinese labs, DeepSeek, Alibaba, Kimi, and others, have continued to compress the gap.

Until proof of AGI, the frontier is still converging.

And the consequences of who ends up on top are massive.

As we've discussed over the past eighteen months, whoever gets to human-level intelligence first sets the standard. They attract the talent. They decide what gets embedded in governments, in banks, in power grids, in weapons. It's the difference between AI that serves humanity and AI that serves the Chinese Communist Party.

You cannot slow down in that race.