The Housing Freeze Outlasts the Bust · Morning Gist
Saturday, October 175 stories · 14 min
The Housing Freeze Outlasts the Bust
Housing is still stuck years after rates jumped, plus a formula for sizing life insurance, a new 2027 tax credit, and AI agents reading bank earnings.
Picked from 54 new articles across 16 sources · arrow keys for other days
Today in 30 seconds
30s left
Adjusted for population, existing home sales have stayed below the 2008 bust low for three years. Carlson blames how fast rates rose. story 2 ↓
The Case-Shiller national index rose 1.9% in the year to July, barely beating inflation, while list prices keep slipping. story 3 ↓
Adam Grossman sizes term life as mortgage plus tuition plus living costs, minus savings, rounded up. He would skip whole life. story 1 ↓
From 2027, gifts to approved K-12 scholarship charities earn a federal tax credit of up to $1,700 per person. Thirty states have opted in. story 5 ↓
Ex-analyst Rubinstein had four AI agents digest the six biggest US banks' 250+ pages and eight hours of earnings calls. story 4 ↓
Character.AI is filing for bankruptcy after Google hired its founders and licensed its technology instead of buying the company.
1/5HumbleDollar· Adam M. Grossman· Oct 17
An adviser's step-by-step formula for how much term life insurance to buy
Adviser Adam Grossman points to lawsuits over how whole life policies are sold and recommends plain term insurance instead. To size it: mortgage balance, plus tuition, plus a fund for living costs, minus savings, rounded up.
Why it matters · Life insurance is one of the few big money decisions where a salesperson's pitch often decides the outcome. A simple formula lets you check that pitch yourself.
Know this first
Context from the editor, not the article · 30 sec
1 · Term versus permanent insurance
Term life insurance pays a set sum if you die within a fixed period, such as 20 years, and then ends. Permanent insurance, meaning whole life or universal life, lasts for life and adds a savings account called cash value. It costs much more.
2 · Policy illustrations
An illustration is a chart a salesperson shows a buyer. It projects how a permanent policy's cash value might grow based on assumed returns. The growth isn't guaranteed, and real results can fall short.
3 · The 4% withdrawal rule
This is a common retirement rule of thumb. If you take out about 4% of a portfolio's value each year, the portfolio should last for decades. It turns a yearly spending need into a target portfolio size.
2/5A Wealth of Common Sense· Ben Carlson· Oct 16
Rates aren't too high by history, but the jump from zero hurt housing
Ben Carlson argues there is no "normal" interest rate. Today's 7.5% mortgages are ordinary by historical standards. The trouble is that home prices were set when mortgages cost 3%, and the sudden move to higher rates has frozen home sales.
Why it matters · If you're buying a home, refinancing or holding bonds, the speed of a rate change can matter as much as the level, and housing is a big enough slice of the economy to slow everything else.
Know this first
Context from the editor, not the article · 28 sec
1 · Interest as the price of time
An interest rate sets how much a dollar promised in the future is worth today. Higher rates make future income worth less now. That affects stock and bond prices, how much it costs to borrow, and how much savers earn.
2 · Zero interest rate policy
ZIRP means the Federal Reserve holding its short-term policy rate near 0%. It did this roughly from 2008 to 2015 and again in 2020–2022, which pushed 30-year mortgage rates down to about 3%.
3 · The mortgage lock-in effect
Most U.S. mortgages carry a fixed rate for 30 years. An owner paying 3% who sells and buys again must borrow at today's rate. Many decide to stay put, so fewer homes come up for sale.
3/5Calculated Risk· CalculatedRisk by Bill McBride· Oct 16
US home prices are flat to slightly up as list prices keep slipping
Bill McBride of Calculated Risk reports that median US asking prices have fallen year over year for 11 months in a row. The Case-Shiller index, which tracks actual sale prices, rose only 1.9% in the year to July.
Why it matters · If you're buying, selling or holding a home as wealth, this shows prices barely beating inflation four years after rates jumped. Nobody is clearly winning yet: no crash, no boom.
Know this first
Context from the editor, not the article · 27 sec
1 · Why 2022 is the turning point
In 2022 the Federal Reserve raised interest rates fast to fight inflation, and 30-year mortgage rates roughly doubled from around 3%. Monthly payments jumped, demand cooled, and the pandemic-era price boom stalled.
2 · List price versus sale price
A list price is what sellers ask. A sale price is what buyers actually pay. Asking-price data comes out sooner, but it can drift away from what homes really sell for.
3 · How Case-Shiller works
The S&P CoreLogic Case-Shiller index uses repeat sales. It compares the prices of the same homes each time they sell, so changes in which homes are on the market don't skew it. It is published with a lag of about two months.
4/5Net Interest· Marc Rubinstein· Oct 16
A former bank analyst uses four AI agents to read bank earnings
Marc Rubinstein, who writes Net Interest, built four AI agents to digest third-quarter US bank earnings. They summarise filings, track themes, predict analysts' questions and update his Excel model. His read: results were solid, especially lending income.
Why it matters · Earnings days are when stock prices move most. Seeing how a professional now uses AI to work through the flood of disclosure shows where an analyst's edge is moving.
Know this first
Context from the editor, not the article · 28 sec
1 · What earnings season is
Each quarter, US public companies publish their results and hold a conference call with analysts. Banks go first. Investors compare the numbers with what analysts expected, and the size of the surprise drives the share-price reaction.
2 · Sell-side vs buy-side
Sell-side analysts work at brokers and publish research for clients. Buy-side investors manage money and trade on their own analysis. Rubinstein has done both jobs and now writes the Net Interest newsletter about financial companies.
3 · Net interest income
Net interest income is what a bank earns on its loans and securities minus what it pays depositors and lenders. It is the biggest revenue line for most large banks, and interest rates move it.
5/5HumbleDollar· Bogdan Sheremeta· Oct 16
New federal tax credit for K-12 scholarship donations starts in 2027
Starting in 2027, the Education Freedom Tax Credit cuts your federal income tax by up to $1,700 ($3,400 joint, the article says) when you donate cash to approved K-12 scholarship nonprofits.
Why it matters · A credit reduces your tax bill dollar for dollar, so this gift costs a lot less than an ordinary deductible donation. It's worth planning for before the 2027 tax year.
Know this first
Context from the editor, not the article · 23 sec
1 · Credit versus deduction
A deduction lowers the income you're taxed on, so it saves you your tax rate times the gift. A credit comes straight off the tax you owe. $1,000 of credit saves $1,000.
2 · Where the credit comes from
The credit was created by the 2025 federal tax law, often called the One Big Beautiful Bill Act. It pays for K-12 scholarships through private charities, with no direct government spending.
3 · School choice tax credits
Many states already offer tax credits for gifts to private-school scholarship funds. This is the first federal version, and it works through state-approved scholarship nonprofits.
Pacific Life Insurance Company recently agreed to pay $58 million to settle a 1, a lawsuit brought by a group of policyholders together. They argued that its marketing was misleading. The company admitted no wrongdoing. In another case, race car driver Kyle Busch sued an agent who sold him coverage costing millions by calling it a "tax-free retirement plan."
Writing on HumbleDollar, adviser Adam M. Grossman says such suits are common for permanent life insurance. He uses them to argue for plain term coverage, and then shows how to size it.
The problem with permanent policies
Whole life and universal life policies combine insurance with an investment account called "2." Salespeople show projections of how that account might grow, called "3." When real growth falls short of those projections, buyers are disappointed and some sue. Term insurance pays only if you die during a set period and has no investment piece. Grossman recommends it for that reason.
Who can skip it
Grossman says single people with no dependents and people who are financially independent usually don't need life insurance. He dismisses the "buy young while it's cheap" pitch: a better price on something you don't need isn't a deal. Business owners are a narrow exception. Insurance can fund a 4, a contract that lets partners buy out a partner who dies.
The formula
Start with your mortgage balance.
Add children's future tuition. Use the midpoint between your highest and lowest guesses.
Add an endowment for living costs, sized like a retirement portfolio. If your family needs $100,000 a year, a 4% to 5% 5 means $2 million to $2.5 million. Subtract the savings you already have.
Round up to allow for inflation and changing needs.
For example (illustrative numbers): a $400,000 mortgage, plus $400,000 of tuition, plus a $2.5 million endowment minus $700,000 of savings comes to $2.6 million. Rounded up, that is about $3 million.
Term length and couples
Tuition is needed for a set number of years, but living costs go on indefinitely. One option is two policies: one ending when your youngest child turns about 22, and one ending around the age you expect financial independence, such as 60. The other is a single larger policy. Many term policies let you reduce coverage later, which Grossman says can make one policy more flexible.
He suggests insuring each spouse in proportion to the income he or she earns, since a surviving spouse often works less. He would ignore employer coverage, because jobs change. He adds that 6 can matter just as much.
The article doesn't cover cases where permanent insurance has stronger defenders, such as large estates. It also gives no data on how often illustrations overstate real results.
Earlier this year, Pacific Life Insurance Company agreed to pay $58 million to settle a 1, a lawsuit brought by a group of policyholders together. They argued that its marketing materials were misleading. Pacific Life admitted no wrongdoing. Writing on HumbleDollar, adviser Adam M. Grossman says the case isn't unusual. Companies that sell permanent life insurance get sued often over how they promote it. In another recent case, race car driver Kyle Busch sued his insurance agent. Busch said the agent sold him coverage costing millions by calling it a "tax-free retirement plan."
Grossman runs Mayport, a wealth management firm that charges a 2. He uses these cases to argue for a simple position: buy term insurance, not permanent insurance. He then explains how to decide how much term coverage to buy and for how long.
Why permanent insurance causes trouble
Grossman's complaint is about structure. Whole life and universal life policies combine two different things. One is the insurance, which pays out when you die. That part is simple. The other is an investment account called "3" that sits next to the insurance.
Nobody knows how an investment will perform, so salespeople show buyers "4." These are projections of how the cash value might grow over time. If the real growth falls short, buyers are disappointed, and the lawsuits mostly come from that gap. Term insurance has no investment piece, so this problem can't arise. Grossman recommends term for that reason.
Who doesn't need it
Grossman says two groups usually don't need life insurance:
Single people with no dependents. Agents sometimes argue that you should buy young because rates are lower. Grossman says to be wary of that pitch: a better price on something you don't need is not a better deal.
People who are financially independent. This means retirees living on their savings, or anyone with enough assets to retire.
The narrow exception is business owners. Insurance can fund a 5, a contract that lets surviving partners buy out the share of a partner who dies.
The coverage formula
Grossman builds the coverage amount in four steps.
Mortgage balance. Survivors don't have to pay off a mortgage. In his experience, though, life is much easier for them without one.
Children's future tuition. This is hard to estimate. College costs vary widely, tuition has historically risen faster than inflation, and one child might win scholarships while another goes to graduate school. He suggests taking the midpoint between your highest and lowest guesses.
An "endowment" for living expenses. Size it the way you would size a retirement portfolio. If your family needs $100,000 a year, a 4% to 5% means a portfolio of $2 million to $2.5 million. Subtract the savings you already have. With $1 million saved, you need $1 million less coverage.
Worked example (illustrative numbers, not from the article): Suppose a family has a $400,000 mortgage and two children. Tuition guesses range from $200,000 to $600,000, so the midpoint is $400,000. The family needs $100,000 a year, which at a 4% withdrawal rate means $2.5 million. They have $700,000 saved, so the endowment piece is $1.8 million. The total is $400,000 + $400,000 + $1.8 million = $2.6 million. Rounded up, that is about $3 million of coverage.
How long the term should run
The formula has a timing problem. Tuition is needed only for a set number of years, but the living-expenses fund has to support a surviving spouse indefinitely. Grossman gives two ways to handle this:
Two policies. One covers tuition and ends when your youngest child is about 22. The other covers living expenses and ends around the age you expect to be financially independent, say 60.
One larger policy. Grossman says this can, counterintuitively, be more flexible. Many term policies let the owner reduce the coverage on request. You can cut it as your savings grow or your needs become clearer, without having had to predict the split at the start.
The article doesn't compare the costs of the two approaches. A buyer would need to get quotes for both.
Couples and workplace coverage
Grossman suggests insuring each spouse in proportion to the income he or she earns. This holds even if the family could live on one income, because a surviving spouse often can't work as much, especially with children at home. He recommends both spouses keep coverage until they are clearly financially independent.
He would not count group life insurance from an employer. It is a nice benefit, but you may change jobs.
His last point is that 7, which replaces income if you can't work, can matter as much as life insurance.
What to make of it
The advice is mainstream fee-only planning. Grossman is paid a flat fee, not commissions, so he has no stake in which policy you buy. The article doesn't address the cases where permanent insurance advocates make their strongest arguments, such as estate planning for very large estates or a dependent who will need lifelong care. It also offers no data on how often illustrations overstate real results. The $58 million settlement shows that disputes happen, not how common the shortfalls are.
The formula also leaves out some costs, such as funeral expenses and childcare for a surviving parent. Grossman's "round up" step is the only cushion for these.
Remember
Most permanent life insurance lawsuits come from the investment piece, where projected cash-value growth wasn't met.
Using a 4–5% withdrawal rate turns a yearly living-cost need into a lump-sum coverage target.
A single term policy that lets you reduce coverage can be more flexible than two separate policies.
10-year Treasury yields since 1940: suppressed in WWII, peaking in the early 1980s, bottoming in the pandemic.1 / 4
Ben Carlson, a financial adviser who writes the blog A Wealth of Common Sense, argues that there's no "right" or "normal" level of interest rates. In his view, what has hurt the economy since the pandemic is how fast rates went from one extreme to the other. His main evidence is housing. Adjusted for population, 1 have been below the low point of the 2008-era housing bust for three years.
Rates set the price of time
Carlson draws on Edward Chancellor's book The Price of Time. Its point is that interest sets how much future money is worth today, and so drives saving, borrowing and risk-taking. Chancellor blames the 2 after the 2008 crisis, when the Federal Reserve held short-term rates near 0%, for inflating asset prices and propping up weak "zombie" firms.
ZIRP's critics got the crisis wrong
Carlson points out that the hyperinflation, dollar crash and speculative blowup many predicted in the 2010s never happened. He also argues that rates don't drive markets as much as people think:
The 3 has done better in the 2020s, with 4–5% rates, than it did in the 2010s.
Stocks boomed in the 1980s even though the 4 averaged over 10%.
Housing stayed quiet through the cheap-money 2010s and only took off during the pandemic.
A chart of the 10-year yield since 1940 shows why he says there is no normal. Rates were held down during WWII, climbed with inflation until the early 1980s, fell for four decades, hit bottom in the pandemic, and have risen sharply since.
The real cost: no margin for error
A 7.5% mortgage isn't unusual by historical standards. Rates never went below 9% in the 1980s and averaged over 8% in the 1990s. But today's home prices, Carlson writes, "only work with 3% rates." Owners holding 3% loans won't sell, because moving would mean borrowing at today's rates (the 5). Buyers can't afford what is for sale. As an illustration, a $400,000 loan costs about $1,690 a month at 3% and about $2,800 at 7.5%.
The result is a frozen market. In 2000, more than 5 million existing homes sold among 105 million households. Today it's just under 4 million among 135 million households. That's roughly 3 sales per 100 households compared with nearly 5 then (my arithmetic, not Carlson's). The housing research firm John Burns estimates housing drives about 20% of economic activity.
What he isn't saying
Carlson doesn't argue that ZIRP was better. His charge is that policymakers didn't think about what a fast swing between extremes does to particular households. He offers no alternative path, and he doesn't say how the gap between home prices and mortgage rates will close.
Ben Carlson, a financial adviser who writes the blog A Wealth of Common Sense, has a two-part answer to the question of what interest rates the economy should have. First, there is no "normal" level. Second, the real damage of recent years came from how fast rates moved from one extreme to the other, not from where they ended up. His clearest evidence is the housing market. Once you adjust for population growth, the number of homes changing hands has been lower for three years than it was at the bottom of the 2008-era housing bust.
Why rates matter so much
Carlson starts with Edward Chancellor's book The Price of Time. Chancellor describes interest as the "1", the rate that turns a stream of future income into a value today. "Without interest, there can be no capital," Chancellor writes. "Without capital, no capitalism."
A quick illustration (not from the article): $1,000 due in ten years is worth about $744 today if you discount it at 3% a year, and about $508 at 7%. The rate changes what every future cash flow is worth today. That covers a stock's future profits, a bond's coupons and the value of a house to someone who has to finance it.
Chancellor spends much of the book criticising the 2 that followed the Great Financial Crisis, when the Federal Reserve held its short-term rate near 0%. He argues that ultra-low rates encourage too much risk-taking, punish savers, inflate asset prices and keep "3" alive. Zombie companies are firms that survive only because borrowing is cheap.
Carlson's pushback: ZIRP didn't cause the crisis people predicted
During the 2010s, many commentators warned that ZIRP would bring hyperinflation, a collapse in the dollar or a speculative crash. Carlson notes that none of that happened. Asset prices did rise. He puts that mostly down to two things: stocks were recovering from a "lost decade" in the 2000s, and the giant tech companies were producing real profits.
He offers several pieces of evidence that the level of rates doesn't drive markets as much as people assume:
The 4 has done better so far in the 2020s than it did in the 2010s, even with rates of 4–5% for several years.
The 5 averaged more than 10% in the 1980s, and stocks still boomed.
Bond yields were above 5–6% for much of the dot-com bubble.
Mortgage rates were low for most of the 2010s, but housing didn't boom until the pandemic made people rethink where they lived.
In a footnote he adds that yields matter much more when stocks are falling. He also notes that complaints never stop. When rates are low the Fed is "punishing savers", and when they're high it's "punishing borrowers".
There is no "normal" rate
Carlson's chart of the 10-year Treasury yield since 1940 shows several distinct eras:
WWII: yields were deliberately held down so the government could fund the war cheaply.
1950s to early 1980s: inflation returned and rates climbed.
Early 1980s to 2020: four decades of falling inflation and falling rates.
Pandemic: yields hit the floor.
Since then: yields rose quickly as inflation came back.
The lesson he draws is that "interest rates are basically never normal". No single resting level exists for rates to return to.
Where ZIRP really did harm: no margin for error
Carlson's main point is that ZIRP failed in a different way than its critics predicted. It left "zero margin for safety" for when rates eventually had to rise.
A 7.5% 30-year mortgage is not unusual historically. Mortgage rates never dropped below 9% in the 1980s and averaged more than 8% in the 1990s. The problem is that home prices in the 2020s were set in a world of 3–4% mortgages. As Carlson puts it, today's buyers face 7.5% rates "with housing prices that only work with 3% rates."
To see the scale (an illustration, not from the article): the monthly principal-and-interest payment on a $400,000 30-year loan is about $1,690 at 3% and about $2,800 at 7.5%. The house is the same, but the payment is roughly 65% higher.
That affordability gap sits on top of the 6. Owners holding 3% mortgages won't sell, because their next loan would cost much more. Together these have, in Carlson's word, "nuked" housing activity.
The numbers
2000: more than 5 million existing homes sold, in a country of 280 million people and 105 million households.
Today: just under 4 million 7, with more than 340 million people and 135 million households.
Simple division (my own, not Carlson's) puts that at roughly 4.8 sales per 100 households in 2000 and about 3 now, a drop of more than a third per household. Carlson's population-adjusted chart shows sales have been below the housing-bust trough for three years running.
This matters beyond would-be buyers. Carlson cites a chart from the housing research firm John Burns showing that housing accounts for roughly 20% of U.S. economic activity once you count construction, the spending that follows a move, and related services. When turnover freezes, a fifth of the economy slows down.
What the argument does and doesn't claim
Carlson is explicit that he isn't saying ZIRP was better than today. His complaint is that policymakers didn't think through what it means to swing "from one extreme to the next" in a short time, or who pays for it. In housing, that's mainly first-time buyers and people who need to move.
The piece leaves several things open. It doesn't suggest what a better path would have looked like, such as raising rates more slowly or never cutting to zero. It doesn't say whether house prices, rates or time will close the affordability gap. It also doesn't weigh the housing damage against the inflation the rate hikes were meant to fight. The S&P 500 comparison is suggestive but loose, because many things besides rates changed between the two decades. The John Burns 20% figure is a broad measure that includes housing-related spending, not just homebuilding. Treat the article as a sharp framing of the problem rather than a policy prescription.
Remember
Today's 7.5% mortgage rates are below the 1980s and 1990s averages; home prices are what's out of line.
Per household, existing home sales have dropped from nearly 5 per 100 in 2000 to about 3.
Stocks have done well under both high and low rates, so rates alone don't explain market booms.
Realtor.com median list prices: peak in 2022, then roughly sideways for four years.1 / 2
Bill McBride of the Calculated Risk blog has published Part 2 of his mid-October 2026 housing review, covering prices. Only the opening section is free, but it gives a clear picture: the market has stalled rather than crashed.
Asking prices are drifting lower
Realtor.com data on 1 (the middle asking price among homes for sale) shows US prices peaked in 2022, when mortgage rates jumped. They have moved mostly sideways since. Recently they've edged down:
Median list prices have been below year-earlier levels for 11 straight months.
September's median was the lowest for a September since 2021.
Two cautions from McBride. First, list prices usually peak in June, so this measure will keep falling through January for seasonal reasons alone. Second, medians "can be distorted by the 2." If more small or cheap homes come up for sale, the median falls even if no house lost value.
Sale prices are still rising, barely
The S&P CoreLogic 3 index avoids the mix problem. It tracks repeat sales of the same homes. For July 2026, not seasonally adjusted (4, the raw figures):
Index
Year over year
5 (10 big metros)
+3.4%
Composite 20 (20 metros)
+2.5%
National
+1.9%
After seasonal adjustment, the National index rose 0.29% from June to July. McBride expects August's report to show a similar annual gain of about 1.9%.
What it adds up to
The two measures differ partly because of timing: Case-Shiller runs about two months behind and averages three months of sales. The overall picture is consistent, though. Prices are roughly flat in nominal terms. With a 1.9% gain likely at or below general inflation, inflation-adjusted home prices are flat to slightly down. That's affordability improving slowly, as incomes rise while prices barely move.
Big cities are holding up better than the nation as a whole, which suggests the softness lies outside the largest metros.
The rest of the post is for paid subscribers. It covers newer price data, mortgage rates and rents, so the free part can't say whether sale prices have softened since July.
Bill McBride of the Calculated Risk blog has published Part 2 of his mid-October 2026 housing review. Part 1 covered inventory and sales. Part 2 turns to prices, mortgage rates and rents, though only the price section is free to read. The rest, including what he calls "more timely house price data," is behind a paywall. What follows covers only the public portion.
Asking prices: four years of going sideways
McBride starts with a Realtor.com chart of 1, the middle asking price among homes for sale. By that measure, US list prices peaked in 2022, when mortgage rates rose sharply. They have moved mostly sideways for four years since.
The recent direction is down:
Median list prices have been lower than a year earlier for eleven months in a row.
The September 2026 median was the lowest for any September since 2021.
McBride adds a seasonal point. Median list prices usually peak in June, so this series will keep falling through January whatever the underlying trend does. A drop over the coming months doesn't by itself mean the market is weakening. The useful comparison is year over year: this September against last September.
The catch with medians
McBride notes that medians "can be distorted by the 2." The median is just the middle listing, so it changes when the kind of home on the market changes, even if no individual home's value moves.
Here's an illustrative example (not from the article). Suppose a town has 50 starter homes listed at $300,000 and 50 larger homes at $600,000. The median sits around the midpoint. Now say builders add 30 small new homes at $280,000 while the big homes stay the same. The median falls, yet no existing house lost value. More condos, smaller new builds, or a shift in activity toward cheaper regions can all pull a median down this way. So falling median list prices hint at softness but don't prove it.
Sale prices: still rising, slowly
Repeat-sales indexes avoid this problem. The best known is the S&P CoreLogic 3 index, which compares the sale prices of the same houses over time. Its latest readings, for July 2026:
Index (not seasonally adjusted)
Change from a year earlier
4 (10 large metro areas)
+3.4%
Composite 20 (20 metro areas)
Adjusted for seasonal swings, the National index rose 0.29% from June to July. McBride expects the August report to show roughly the same 1.9% annual gain, judging from other data he tracks.
"Not seasonally adjusted," or 5, means these are the raw numbers. For year-over-year comparisons that's fine, because the same month is compared with itself. The month-over-month figure needs the seasonal adjustment, because spring and summer are always stronger.
Reading the two measures together
The two sources send somewhat different signals, and the difference itself tells you something:
Asking prices are slipping year over year, and the data is current through September.
Sale prices of the same homes are still rising, but only by about 2% a year, and the data runs only through July.
Some of that gap is timing. Case-Shiller lags by about two months and averages three months of sales. Some is mix. Either way, the picture is a market that has stalled rather than collapsed. A 1.9% nominal gain is likely at or below general consumer inflation, so inflation-adjusted home prices are roughly flat or edging down. That's a slow way for housing to get more affordable: incomes catch up while prices barely move.
The city figures matter too. The 10 largest metros (+3.4%) are rising faster than the nation (+1.9%). That suggests the weakness is concentrated outside the biggest cities. Many Sun Belt and Western markets added lots of inventory after the boom, though the free portion of the article doesn't break this down by region.
What the article leaves open
The paywalled section apparently covers newer price data, mortgage rates and rents. So the free portion can't say whether the softening in asking prices has reached sale prices since July, where mortgage rates stand now, or how rents are moving. McBride's August Case-Shiller estimate rests on "other data" he doesn't name in the public text. Treat this as a snapshot of price momentum, not the full state of the market.
Remember
Falling median asking prices can reflect cheaper homes being listed, not homes losing value.
Repeat-sales indexes like Case-Shiller track the same houses, so they better show whether a given home is worth more.
Prices of the 10 biggest metros rose 3.4%, well above the 1.9% national gain.
This week the six largest US banks reported third-quarter results. Together they released more than 250 pages of financial information and held over eight hours of conference calls. JPMorgan alone gave investors 523 new data points. Marc Rubinstein, a former bank analyst and fund manager who now writes the Net Interest newsletter, didn't read it all himself. He handed it to four AI agents he had trained.
What the agents found and how he built them are behind a paywall. The free part covers why speed matters and what each agent does.
Why earnings days matter
In his 1 years, writing research for a broker's clients, an earnings day meant reading a release of up to 75 pages that might arrive at 6am. He then briefed salespeople and clients, joined the earnings call, updated his model and published a note, all while racing a dozen rival analysts.
The stakes come from a statistic he cites: earnings dates are only 1.6% of trading days but account for 15% of a stock's annual 2. Changes to profit forecasts drive stock prices, and Rubinstein says the market sometimes gets them wrong. Anyone who processes the numbers faster and better can act first.
The team
Ernie reads a folder of each bank's past releases, presentations, call transcripts and annual report. It checks analysts' previews to see what the market expected, then reports where the quarter matched and where it differed. Once the call transcript is available, it writes a five-page report and a two-page brief in about the time it takes to brew an espresso.
Thelma tracks Rubinstein's ongoing themes: AI agents disrupting bank work, lending to fund AI, (loans from non-bank funds) and rising interest rates. For each, it compares what managers said with what the numbers show.
The verdict
Rubinstein says the agents liked the results more than the market did. Banks broadly raised their 4, especially for 5, which is what they earn on loans minus what they pay for deposits and borrowing. Strong equities trading and banking fees mostly offset weaker bond trading. Consumers kept spending, and loan losses stayed in check.
The free text doesn't say how accurate the agents are or how he checks their work. It also doesn't address whether a speed advantage lasts once every analyst has the same tools.
This week the six largest US banks reported third-quarter results. Together they released more than 250 pages of financial information and held over eight hours of conference calls to explain it. JPMorgan alone gave investors 523 new data points. Marc Rubinstein, a former bank analyst and fund manager who now writes the Net Interest newsletter, didn't read it all on his own. He used four AI agents he had trained for the job.
The free part of his post explains why earnings days matter, what each agent does, and his headline view of the quarter. What the agents found and how he trained them are behind the paywall, so this summary covers the setup and not the results.
Why earnings days are a crunch
Rubinstein describes his old 1 routine, meaning his years as a broker's research analyst writing for clients. Releases arrived as early as 6am and filings ran to 75 pages. He had to brief the firm's internal salesforce and institutional clients, update his financial model, dial in to the earnings call, ask three questions, revise the model again and publish a note. A dozen rival analysts at other firms were doing the same thing at the same time, and some days several banks reported at once.
On the 2, managing money himself, the work was less frantic but more was riding on it. He cites a statistic: earnings release dates are only 1.6% of trading days but account for 15% of a stock's annual 3. In other words, a large share of a year's price movement is packed into a few days. 4, meaning changes to analysts' profit forecasts, still drive stock performance, and Rubinstein says the market doesn't always get them right. Trading is heavy around announcements, so there is enough liquidity to buy or sell size quickly. He says he sometimes traded on his own analysis before the market opened or while the call was still going.
That is the problem the agents are meant to solve. The raw information is public, so the edge comes from processing it quickly and accurately.
The four agents
Ernie does the core analysis. Rubinstein keeps a desktop folder for each bank with its past earnings releases, investor presentations, call transcripts and latest annual report. When a new release comes out, Ernie reads the folder and works out what the market expects from analysts' preview notes. It then reports where the quarter confirms that view and where it departs from it. Once the call transcript is available, Ernie adds it and produces a five-page report and a two-page brief "in the time it takes me to brew an espresso."
Thelma tracks the themes Rubinstein follows across banks and collects new evidence as it appears. The current list:
agentic disruption, meaning AI agents doing work that banks' staff or products used to do
AI financing, meaning lending to fund AI infrastructure
5, meaning loans made by non-bank funds rather than banks
the effect of rising rates, which is back on the list
For each theme, Thelma flags what management said and what the underlying numbers show. That gap can be revealing.
The Mike Mayo bot predicts the question Mike Mayo is most likely to ask on each call. Mayo is a bank analyst, now at Wells Fargo Securities, who has covered more than 140 earnings seasons. Rubinstein says his blunt questions "invariably surface the most pressing issues of the day." Modelling one well-known sceptic is a cheap way to find the weakest point in a quarter before management has to answer for it.
The earnings modeller updates Rubinstein's financial model directly in Excel. He first tested AI modelling in February, in an earlier post called "Excel Forever," and says it has improved a lot since.
What the quarter showed
According to Rubinstein, the agents liked the results more than the market seemed to. Banks broadly raised their 6, especially for 7, the gap between what they earn on loans and securities and what they pay on deposits and borrowing. Fees from equities trading and investment banking were mostly strong enough to make up for a weaker quarter in fixed-income trading. Consumers kept spending and credit quality held up, meaning loan losses did not rise in a worrying way.
He says the agents "turned up things I couldn't have got to as quickly." The free excerpt doesn't say what those were. Paid subscribers can email him for the agents' instructions and outputs.
What to make of it
The post is mainly a picture of how one experienced analyst has split his workflow among AI tools. The design follows how the work is actually done: one agent compares results with expectations, one tracks themes across companies, one plays the adversarial questioner, and one handles the spreadsheet. It also relies on a carefully curated folder of past documents for each bank, so the agent's context is chosen by someone who knows what matters.
The free text doesn't cover accuracy, how often the agents make mistakes, or how Rubinstein checks their output. Those are the obvious questions for anyone who might copy the approach. It's also fair to ask how long any speed advantage lasts if every analyst has similar tools. If they do, the edge may move back to knowing which questions to ask, which the Mike Mayo bot is built to capture.
Remember
Earnings dates are 1.6% of trading days but account for 15% of a stock's annual price variance.
The agents are split by job: compare results with expectations, track themes, predict hard questions, update the model.
Banks raised net interest income guidance, and credit quality held up, though the market reacted coolly.
A new federal tax break starts on January 1, 2027: the 1. Bogdan Sheremeta flags it on HumbleDollar, noting that few people know about it yet. It was created by the 2025 federal tax law.
How it works
You donate cash to a 2, a public charity that spends at least 90% of its money on K-12 scholarships. The scholarships can pay for private school tuition, tutoring, special-needs services, books and supplies. In return you get a federal income tax credit of up to $1,700 per person, or $3,400 on a joint return, according to the article.
It's a credit, not a deduction, and that's the key point. A deduction saves only your tax rate times the gift. A credit comes straight off your tax bill.
Here's an illustrative example, assuming each donated dollar earns a dollar of credit up to the cap. A filer in the 22% bracket who gives $1,700 to a normal charity saves about $374, and only if they 3. Give the same $1,700 to an SGO and their federal tax falls by $1,700. In effect, they are steering money that would have gone to the Treasury toward scholarships, at little or no cost to themselves.
The state opt-in
States must 4 for their SGOs to qualify. Thirty have so far, and the deadline is February 15, 2027. You can give to an SGO in any participating state, even if your own state hasn't joined. Each participating state will identify its SGOs, and the IRS will publish a list to check before you give.
Paperwork
When you give, tell the SGO the gift is meant for the credit.
Keep the 5 with your donor number. The SGO must send it by January 31 of the next year.
You don't need to give the SGO your Social Security number.
A 2027 gift goes on your 2027 return, filed in 2028.
What's not covered
The short post doesn't say whether unused credit carries forward to later years, whether income limits apply, or how the credit interacts with state credits or the charitable deduction. The IRS list of SGOs isn't out yet either. Check the final IRS guidance before counting on the full amount.
From January 1, 2027, a new federal tax credit will reward cash gifts to nonprofits that pay for K-12 scholarships. Bogdan Sheremeta, writing on HumbleDollar, points out that few people have heard of the 1. It was created by the 2025 federal tax law. Sheremeta's post is short and practical. This piece covers what it says and what it leaves out.
What the credit is
You give cash to a 2, a public charity that spends at least 90% of its money on scholarships. In return you get a credit against your federal income tax. A credit is not a deduction. The scholarships can pay for:
private school tuition
tutoring
special-needs services
books and supplies
The cap is $1,700 per person, or $3,400 on a joint return, according to the article.
Why "credit, not deduction" matters
The difference decides how much the gift really costs you. Here is an illustrative example, assuming the credit matches each donated dollar up to the cap, which is how the article describes it. A single filer in the 22% bracket owes $10,000 in federal tax.
As an ordinary deductible gift: $1,700 to a charity saves $1,700 × 22% = $374 if they 3. The gift really costs them $1,326.
Under this credit: the same $1,700 cuts their tax bill to $8,300. Their own net cost is close to zero. The money goes to the SGO that would otherwise have gone to the Treasury.
So the credit lets taxpayers steer part of their federal tax to scholarship funds. Many people who don't itemize get no tax benefit at all from ordinary charitable gifts. For them the gap is even bigger.
How the state opt-in works
States must 4 for SGOs based there to qualify. Thirty have done so, and the deadline is February 15, 2027. Sheremeta notes you can give to an SGO in any participating state, even if your own state has stayed out. Many donors will still prefer a local SGO, so your state's decision affects where your money can go close to home.
The process runs in steps:
Each participating state identifies its qualifying SGOs.
The IRS publishes a list of them.
You check that list before giving.
Paperwork
When you give, tell the SGO the gift is meant for the credit.
The SGO must send a 5 with your donor number by January 31 of the next year. Keep it.
You don't have to give the SGO your Social Security number.
A 2027 gift is claimed on your 2027 return, filed in 2028.
What the article leaves open
The post is a heads-up, not a full guide. It doesn't cover:
whether unused credit can be carried forward to later years or refunded
how the credit interacts with state school-choice credits or the federal charitable deduction for the same gift
whether income limits apply
which 30 states have opted in (it links to a list rather than naming them)
There is a timing question too. The IRS list of approved SGOs doesn't exist yet. Anyone planning a gift early in 2027 may have to wait for it. Check the IRS's final guidance, or ask a tax preparer, before counting on the full $1,700 or $3,400.
The article takes no position on school-choice policy itself. It treats the credit as a planning opportunity. Readers will form their own views on whether federal tax money should be routed toward private-school scholarships.
Remember
A $1,700 credit can wipe out $1,700 of tax; a $1,700 deduction saves only a fraction of that.
You can give to an SGO in any participating state, even if yours hasn't opted in.
Check the IRS list of approved SGOs and keep the January 31 acknowledgement.
Round up. Inflation can stretch budgets, and your family's needs may change.
+2.5%
National
+1.9%
3
The Mike Mayo bot predicts the question Mike Mayo, a blunt bank analyst at Wells Fargo Securities with more than 140 earnings seasons behind him, is most likely to ask on each call.
The earnings modeller updates his financial model directly in Excel.
