10-Year Treasury Yield Falls Under 5%
The 10-year Treasury yield ($TNX) has fallen below 5% for the first time in a month. Today’s 14 basis point drop is the largest since August 2025.

The 10-year Treasury yield ($TNX) has fallen below 5% for the first time in a month. Today’s 14 basis point drop is the largest since August 2025.

While AI-related sectors are down sharply, four sectors are up more than 1%.

At the top of that list is Staples. The S&P 500 Staples Sector ETF ($XLP) is up 2.5%, its best day since July and before that April 2025 approaching its February all-time high.

We aim to play the long game with the shares, but here’s what the market focus will be near term.
“The sea was angry that day, my friends – like an old man trying to send back soup in a deli.”
– George Costanza
Today’s market is different than some of the past sharp drawdowns in the last 12-18 months.
* The drop is broad and deep — with other asset classes like gold and bitcoin joining the downside.
* There is no single catalyst like Leopold (Situational Awareness hedge fund liquidation) to blame on the market’s selloff.
* The weakness in financials over the last several months is now accelerating, with no respite whatsever.
* The decline in over stretched stocks is breathtaking — especially the over conceptualized AI stocks on the periphery (e.g. $GEV -$105 to $820). Look at the breakdown in $CAT’s chart (-$45 and down from $950 to $695 in a straight line).
* We are witnessing a classical (but violent) rotation. See the remarkable strength in staples and healthcare.
* A sign of flight to quality/safety seen in a pronounced rise in bond prices (and in lower bond yields). ($TLT +$1.20 today)
Position: None
– NYSE volume 3% above its one-month average;
– Nasdaq volume flat to its one-month average;
– VIX index: up 15.59% to 19.28
SOURCE: Interactive Brokers

S&P500 Sector ETFs
SOURCE: Interactive Brokers

SOURCE: TheStreet Pro


SOURCE TheFly


This has been one of my more controversial positions. Let’s dive into the report and see if I’ll be buying more.
For those who read my nightly market updates, I have noted the past couple of weeks that the equal-weighted S&P 500 ($RSP) has “been able to hold up better” versus both the cap-weighted indices and the Russell 2000.
RSP took another step forward Wednesday confirming its Tuesday close above the 50-day moving average and extending its gain since the October 1 intraday low to over 4%. In addition, the daily MACD is strongly positive while the RSI is now over 60.

The weekly chart is solid, well above a rising 50-week moving average with the RSI also near 60. The weekly MACD is the one fly in the ointment not yet confirming the intermediate trend.

I initially took a small position at the daily MACD crossover two weeks ago with a stop at the October low, and then added a slug Wednesday on the extension over the 50-DMA, taking it to medium sized. I will hold that for now, going back to small sized on a break back below the 50-DMA. I will consider adding more depending on how things develop.
This was initially drafted premarket but it is continuing to hold up better this morning down just a touch versus the -1.3% decline in the S&P 500.

At the time of publication Sethi was long RSP.
The AI complex is getting hit on earnings while small caps, biotech, retail and defensive pharma hold up. Here’s what I’m buying.
Well, per my More Tales from September (below on $META), it is amazing people could not see this coming.
It should have been obvious.
Quite the change from the numbers first reported and the big up move in the after hours, to what it is happening now after people had a chance to figure out what has happened with the balance sheet (and what might be off balance sheet) and cash flows.
Both gutted. Meta and Muse.
Cash flow negative.
This is what AI is, it is currently just a giant black hole.
It is money spent, with hope, but not purpose. It is digging holes, and filling them back up. And it is a money recycling machine on so many levels, including with all the venture-funded garbage startups that have no reason to exist, but effectively just funnel money back up the food chain in a giant money recycling operation and attempt at a self-fulfilling prophecy. Unreal that 70%-80% of frontier lab revenue might be coming from this stuff. Think about the other consumption-related ripple effects, and the ramifications.
No doubt the memos will be going out soon from the VCs to the portfolio companies, and it is going to say “start cutting your burn rate.” Eventually there is not enough money, and not a greater fool to step in line to carry the burden. Even the VCs will realize that. Character AI was just the beginning.
At any rate, what should have been obvious about Meta/Muse should be obvious about everything else too. But somehow, it isn’t?
Per below on all the garbage VC startups driving token spend, while corporate America with actual self financing businesses that can exist and spend of their own volition, is not because they realize AI is just an expense, here is some more data – Article, Chart, Torsten Slock Apollo:
Article: It is paywalled you can get a sense from the headline and the first few sentences you can see. The 1% of companies referenced are startups. The Torsten Slock piece that is 3rd on the list has some of the data as well: https://www.zerohedge.com/ai/1-companies-account-80-openai-and-anthropic-spending-and-it-just-stalled
Chart: In fact, adjusted for shifts in utilization rates, total factor productivity growth was zero in Q2. Said otherwise, there has been zero productivity boost from AI! In other words, paying to dig holes, and fill them back up, brought to you by the Venture Communities own circular reality.
Back to META, using the AOL gamebook of attracting eyeballs by offering a product for free…. I am not taking a “victory lap” but investors are finally realizing what I identified last month:
BY DOUG KASS · Sep 23, 2026, 7:35 AM EDT
“Muse personal AI agent is free to use with a generous weekly usage limit (such as up to 100 million tokens per week), though paid subscription tiers are available if you need higher usage capacity.”
– From AI
Does Anyone Do Their Homework Anymore?
Not that I am or have been trying to call the stocks, because that is a fool’s errand these days, especially in these types of mania-driven markets. But it is fun to occasionally comment on the markets relative to the underlying fundamentals and business issues.
Monday and Tuesday were a combination of a gamma squeeze aided and abetted by Facebook’s Muse announcement.
For those that don’t know, Muse is Facebook’s personal assistant app, a theoretical AI agent, that made it to the top of the app stores download list, helping to inspire the gamma squeeze rip of the last two days.
Why did it make it to the top of the list?Because it is both new and FREE! (I don’t recall a single person in the business media saying this!)
Hundreds of billions of investment, and this is what they have to show for it? A free, largely non-differentiated personal assistant app that Amazon (AMZN) has already cut off from shopping its site? And an app, that after the novelty factor wears off, will probably not be used all that much and likely never generate any economics, because it is free and of limited real utility and effectiveness. Guess what, people actually like shopping themselves, that is why they do it!
Heck, even Angry Birds costs money (and generates economics).
My guess, this thing goes the way of the Metaverse and Threads.
If any thing speaks volumes about AI, not sure if it is this, or it is OpenAI spending $22.5 million of compute to win a $1 million prize and needing to cheat along the way to even do that!
Coincidentally per prior on Muse, I just saw these.
The reference to Amazon is their theoretical no human convenience stores where AI was supposed to figure out everything you put in your basket. Turns out they had humans in India watching people on cameras. End of day, none of it ended up working, it lost too much money, and they shut all the stores anyway. Same for McDonald’s (MCD) automated AI ordering system, where “no onions” seemed problematic for the AI to get right. I guess I was too optimistic, hundreds of billions of investment, and Meta (META) still needs the humans too, because the AI still does not work.
Trillions of dollars of market cap added across the board due to this? And PS, Meta already paid for the junk they have in place to do this. They would have been better off buying offshore call centers for pennies on the dollar. Incidentally, that is one more thing AI was supposed to replace, but the opposite is happening, because it turns out Gen AI is multiples more expensive than cheap offshore labor.
Meta’s spending warnings and share-class fine print, Tesla’s capex cut, and a ‘not good enough’ beat from Lam Research are dragging Tech lower adding to a broader risk-off tape.