Hi there,
Welcome to FinSoar. Today we're tracking two sides of the affordability crunch: where the "live alone in the city" fantasy still pencils out (hint: Wichita works, Manhattan doesn't), and why Jamie Dimon just compared auto loan stress to cockroaches in your kitchen.
Plus: Google's DeepMind claims a gold medal in competitive coding, discovers a potential cancer therapy combo, and partners on fusion energy, all while its CEO warns against moving too fast.
Housing: The “Carrie Bradshaw” dream is shrinking unless you move to Wichita
The Economist’s latest “Carrie Bradshaw index” compares what the median worker earns with what you’d need to rent a median studio under the 30%-of-gross rule.
A score >1 means single-life is doable on a median paycheck; <1 means rent wins.
This year, 41 cities slipped into “unaffordable,” up from 38 last year, including previously easier Houston, Dallas, and Memphis, where rents outpaced wages or supply lagged as construction costs hit hard.
Where you can still live alone (and where you really can’t):
The index’s city-level picture lines up with a fresh ranking of the best and worst cities for solo living, which uses the same 30% rule.
Wichita tops affordability with a Bradshaw score of ~1.77 (median studio ≈ $660) and Baton Rouge, Lincoln, Des Moines, Akron close behind.
On the flip side, New York remains the toughest, with an affordable studio wage ≈ $151,600 vs. a median wage that’s ~$91k short, followed by Miami, Jersey City, Boston, and Charleston.
However, even “unaffordable” tech hubs can look different after taxes and pay.
San Jose screens as pricey on the index, but high salaries still leave discretionary room after rent and utilities, per the Sherwood analysis. Meanwhile, some Sun Belt/Mountain cities improved this year (Phoenix, Aurora, Knoxville) as rents slipped or wages rose.
A buyer’s conundrum
If your Plan B is “buy instead,” some relief is emerging, but it’s selective.
San Diego, Miami, LA, Austin, San Jose, Phoenix, and Dallas–Fort Worth led the list of metros with faster price declines as inventory climbed and sellers cut asking prices.
The national median list price slipped 1.2% MoM in September, and listings stayed on the market 15 days longer than in 2022.
Builders are leaning in too. Homebuilder sentiment just hit a 6-month high, with 38% of builders cutting prices (average cut 6%, the biggest in a year). However, demand remains uneven and still below the breakeven reading of 50.
For single renters, the Carrie fantasy is drifting farther from coastal reality, but it’s alive and well in parts of the Midwest and Sun Belt.
If you’re flexible on ZIP code, Wichita, Baton Rouge, Lincoln, Des Moines, Akron and friends still work out. If you’re not, watch for rent reversals in Phoenix/Aurora-type markets and price-cut pockets on the for-sale side, the only real way to reclaim some Bradshaw.
Sources: Reuters, Business Insider, Sherwood, The Economist
Repo Rates and Red Flags: Auto Loan Stress as an Indicator for the Broader Economy
The numbers tell a stark story: American car buyers are struggling.
According to Fitch Ratings data, 6.6% of subprime auto borrowers were at least 60 days delinquent on their loans in January 2025, the highest rate the agency has ever recorded.
Meanwhile, repossessions surged to 1.73 million vehicles last year, the most since 2009, marking a 43% jump from 2022.
For repo companies like Midwest Recovery and Adjustment in Detroit, business is booming. "Right now, we're overwhelmed with work," president George Badeen told The Guardian, adding that his crews now travel in pairs due to increased confrontations with desperate car owners.
The distress isn't confined to the fringes. While subprime borrowers face the steepest challenges, middle and higher-income households are seeing delinquency rates climb as well, according to VantageScore's chief economist.
The culprit? A perfect storm of affordability pressures. Average vehicle prices hover near $50,000, interest rates exceed 9% on new cars and approach 14% on used vehicles, while insurance costs have spiked 19% year-over-year.
The pain is squeezing household budgets from all sides. Repair and maintenance expenses have climbed 33% since 2020. Paying off a new car now requires 42 weeks of median income, up from 33 weeks pre-pandemic.
With average monthly payments topping $750, Americans holding an estimated 100 million auto loans are making difficult choices.
The stress has spread beyond individual balance sheets into credit markets. September's bankruptcy filings by subprime lender Tricolor Holdings and auto parts supplier First Brands sent tremors through Wall Street.
JPMorgan Chase absorbed a $170 million charge from Tricolor's collapse, prompting CEO Jamie Dimon to issue an ominous warning: "When you see one cockroach, there are probably more."
Dimon's concerns extend beyond individual failures to the broader credit cycle. "We've had a credit market bull market now for the better part of since 2010," he told analysts. "These are early signs there might be some excess out there."
Still, most major banks insist their exposure is limited and overall credit quality remains sound. BlackRock's CFO characterized the bankruptcies as "idiosyncratic pockets of stress" rather than systemic weakness.
But economists see auto loan distress as a canary in the coal mine. "Having a car is essential to being able to work," Columbia Business School's Brett House explained. "So when we see stress in the auto financing market, we typically receive that as an indication that household finances are getting tighter."
For now, that indicator is flashing yellow, if not red.
Sources: Axios, The Guardian, Reuters,
Google’s DeepMind Week: Medals, Molecules, and Megawatts
Google’s AI arm has had a big week, spanning coding contests, cancer biology, and even fusion energy.
DeepMind says a tuned Gemini 2.5 model became the first AI to win a gold medal at an international programming contest, solving a fluid-routing brain-teaser in under 30 minutes.
It ranked #2 of 139 human teams at the Azerbaijan ICPC event.
Google calls it a “historic” leap in abstract reasoning. Not everyone was buying the hype: UC Berkeley’s Stuart Russell cautioned the “claims of epochal significance seem overblown,” noting the result does show progress toward more accurate AI coding but isn’t necessarily a turn to the real world.
On the science front, Google detailed a Gemma-based model. Cell2Sentence-Scale 27B generated and helped experimentally validate a new cancer-immunotherapy hypothesis.
It combined silmitasertib (CK2 inhibitor) with low-dose interferon boosted antigen presentation ~50% in lab tests, potentially turning “cold” tumors “hot” for immune attack.
It’s early-stage biology (not a therapy), but notable as a model-to-bench loop that produced a testable prediction.
AI meets fusion
DeepMind also unveiled a partnership with Commonwealth Fusion Systems to speed up the SPARC tokamak using TORAX (a differentiable plasma simulator) and reinforcement learning for real-time control, from maximizing fusion power to sweeping heat loads on reactor walls.
The aim was to find “breakeven” faster by virtually exploring millions of operating pulses before the magnets ever hum.
Even as AI headlines multiply, DeepMind’s CEO Demis Hassabis warned against repeating social media’s “move fast and break things” era. He advised designing AI to serve users rather than hijack attention, and testing rigorously before scaling.
That caution lands amid growing jitters over exuberance. OpenAI’s circular chip deals and rocket-ship valuations have some veterans invoking dot-com vibes and vendor-financing.
Sources: The Guardian, Google Research blog, Business Insider, DeepMind,


