AI and Yields
We’ve been in semi-hibernation given our work on the Technology side of TEK2day migrating our T2D Pulse tool into Beta stage. If we do work with OpenAI Codex I’ll report it here. My view is that $200/month for a round the clock AI agent (Codex), that writes code while you sleep is a great value assuming the reasoning capabilities and memory of the underlying model is exceptional. OpenAI o3 is far from exceptional in these areas. I would go as far to say that memory is poor. Further, the models lack what I would define as common sense, often getting lost in the weeds. Why repair bad code when the code block can be entirely replaced?
Switching to the 10YR Treasury yield, I’ve long held that the 10YR should be well above 5%. Annual fiscal deficits of approximately $2 Trillion guarantee that the value of the Dollar will erode and that fiscal default risk will move higher in concert with the $36 trillion Public debt mountain (to say nothing of $70 Trillion in unfunded Social Security and Medicare liabilities). Therefore, yields should move higher to reflect these risks/truths. It is not going to get better over the long run, not while Social Security, Medicare and Medicaid are growing 6-9% per year and GDP is bouncing around 0-3%. The math simply does not work.
The Treasury market is too sanguine about the United States’ fiscal position and a Dollar that will continue to erode in value versus real assets. I could make the case that the 10YR ought to be at 10%.
