One of the best things about tools like Claude Code and Codex is that you can run historical analysis on almost anything.
You can pull the raw data, go back through the old record, and see how similar situations actually played out.
This is the exact AI strategy I used in 2025 to publicly predict the tariff bottom, and derisk before aggressively buying.
Here's why AI has a real advantage for moments of market crisis:
Most investors aren’t financial historians, and even the ones who are can’t keep 60 years of Fed statements and bond data in their heads. AI can go through decades of numbers and old documents at once and map out how comparable setups ended.
For this week, I ran this exercise on the recent bond sell-off to find anything interesting.
A TLDR of the current situation:
The 10-year Treasury just closed at 5.11%, its highest since 2007, and the Fed is hiking again only nine months after its last cut.
There are plenty of theories about what happens from here, so I started with a simpler question: when has the bond market looked exactly like this before?
I had Claude Code pull the data through APIs and MCP servers and score every trading day since the 1960s against today.
How closely the eight best historical matches line up with today, with February 2000 set to 100.
February 2000 lines up with today the most out of all other historical analogies.
The setup is very close:
The Fed cut three times in late 1998 after Long-Term Capital Management nearly collapsed
Then started hiking again in June 1999, the same kind of restart we just got after the 2024-25 cuts
Yields were climbing fast and oil was going up
Core inflation was running close to where it is today, and tech was leading the stock market.
Stocks and bonds were even moving together, which is what they’re doing now.
In 2000, yields peaked months before the Fed stopped hiking
The 10-year yield and the Fed funds rate from mid-1998 to mid-2001, with the main turning points marked.
The 10-year peaked in January 2000, a week after the Treasury said it would buy back up to $30 billion of its own bonds.
After that the Fed hiked three more times, the last one a 50bp hike in May, and the 10-year never got back to its January high.
By January 2001 the Fed was cutting, and a year after February 10 the 10-year was 162bp lower.
In other words - history says the best time to long bonds is before the Fed stops hiking.
Long bonds made 20% and the Nasdaq-100 lost 44%
What each major asset returned in the 12 months after February 10, 2000.
In the year after February 10, the 10-year Treasury returned about 20% and the 30-year returned a bit more.
The Nasdaq-100 lost 44%, and software and chip stocks each fell more than 40%.
Value stocks beat growth stocks by more than 60 points, and the dollar rose while gold fell.
Credit spreads widened by more than a point over the same year.
Long bonds and cheap stocks made money, while the tech names that had led the market lost almost half.
Real yields now pay more than stocks earn, like in 2000
The real 10-year Treasury yield against the S&P 500's cyclically adjusted earnings yield since 1980.
For equity investors, this is the chart I’d look at first.
A real 10-year Treasury yield now pays more than the S&P 500 earns on a cyclically adjusted basis, and the last time that was true was 2000.
The gap is a lot smaller this time, since in 2000 the real yield beat the earnings yield by about two points and today it’s about a third of a point.
It’s pressure on stock valuations, just much less of it than tech faced in 2000.
Where today is different vs 2000
How today compares with February 2000, with the matches and the differences marked.
The biggest difference is the government’s finances:
In 2000 the budget was in surplus and debt was 38% of GDP, while today the deficit is 5.6% of GDP and debt is close to 99%.
That surplus is the reason the Treasury was buying back bonds in 2000, and the shortage of long bonds helped push their yields down.
This year the Treasury upsized its long-bond buybacks on September 9, and the 10-year kept rising anyway.
The key difference in this cycle is we have an unlimited money printer .... this changes the calculus of what defines as "risk" if the Dollar looks very different 30 years from now.
How I’m looking at it
From my view, the market is asking two distinct questions:
Can I get a better yield on Treasuries?
Can I trust the US Dollar will be in a good shape by the time long-dated treasuries are due?
For question #1 (which I think most folks are focused on), we've already seen rates-sensitive equities sell off hard as those investors switch from clipping dividends to owning Treasuries.
Look at Utilities for example:
The answer to #2 ("Can I trust the US Dollar will be in a good shape") may be the same answer to why Bitcoin and tech equities have held up super strong during this episode:
If the government can't be trusted, hard assets will be lasting
If the government can outgrow the deficit issue, AI / Semis / Tech are the ONLY solution
I'm personally looking for opportunities on the long side of both crypto and tech / semis.
Flows also look like many of those who wanted to sell have already sold:
