The Commoner

Marty's Bent

Let's Check Our Priors

I've spent a lot of time in this rag over the years writing about the debt problem and wondering when the bond market would finally force the government to deal with it. I still think that reckoning is coming. I've also clearly underestimated how long governments can postpone it. When you spend as much time as I do looking at the things that are broken, it's worth stopping occasionally to ask whether you've become too quick to dismiss anything that suggests the economy is doing better than you expected.

Jordi Visser's post this morning got me thinking about this again. Markets climbing a wall of worry while everyone has a very good explanation for why they shouldn't be climbing. There is plenty to worry about. The debt hasn't disappeared. Crude and diesel prices are painful. The Fed just raised rates. But there are also people building an incredible amount of productive capacity, particularly around AI, and I don't want my view of the government's balance sheet to make me blind to what those people are accomplishing.

Take the Atlanta Fed's GDPNow estimate. The model was projecting 5.1% annualized real growth for the third quarter yesterday and kept that headline estimate in its update this morning. If it comes in at that level, it would be the strongest quarter since the end of 2021. That deserves some attention.

I saw people suggesting that oil exports were doing a lot of the work. I haven't found the evidence for that. July's petroleum exports fell in real terms, and yesterday's model breakdown had exports overall subtracting from growth. Inventories were adding roughly two percentage points, which is a reason to be careful with the headline. Businesses accumulating inventory doesn't necessarily tell you how much demand they'll have for it down the road. Even after taking inventories, trade and government out of the picture, however, yesterday's model had private domestic final demand growing at roughly 4.7%. There is enough consumer spending and private investment underneath the headline that I wouldn't wave it away.

Of course, an aggregate number doesn't tell us how the family trying to pay its bills is doing. We've spent enough time discussing the K-shaped economy to know that the fortunes of asset owners and people living paycheck to paycheck can diverge considerably. I wanted to know whether the consumer was spending because he was earning more or because he was putting more of his life on a credit card.

The data doesn't give us a clean percentage for that. A purchase made with a credit card and paid off when the bill arrives is very different from a purchase added to a balance someone is already struggling to pay down. Looking at the amount spent on cards alone won't tell us which is happening.

What we can see is that credit-card balances reached $1.263 trillion in the second quarter, up $21 billion from the first. The New York Fed's data also show 12.92% of outstanding card balances were at least 90 days delinquent. That's an uncomfortable number. It was down slightly from 13.12% in the first quarter, but above where it stood a year earlier.

There is an important wrinkle here. That measure includes old, charged-off debts that are still being reported to the credit bureaus. The New York Fed looked into the divergence and found that those debts remaining in the data longer help explain why the total delinquent share has risen even as the pace of new delinquencies has stabilized. Its annualized rate of card balances newly moving into serious delinquency was 6.97% in the second quarter, almost unchanged from a year earlier. People are having trouble paying their bills. The latest numbers don't show that trouble suddenly accelerating.

Nor does the latest monthly spending report look like an unambiguous borrowing-fueled boom. In July, inflation-adjusted disposable income rose 0.4% while real consumer spending was essentially flat. The saving rate was 3%. Meanwhile, revolving credit grew at a 2.5% annualized pace, slower than in June. That leaves a fairly thin saving margin out of current income, but I don't think we can honestly look at these numbers and dismiss all of the consumer's resilience as people maxing out their cards either.

Wages don't give us much reason to declare victory for the middle class. Real hourly earnings were down 0.3% from a year earlier in August. Real weekly earnings were up 0.3%, with longer hours helping make up the difference. There are encouraging developments in the labor market, including fewer people working part time because they couldn't find full-time work. I had also wondered whether the change in immigration policy was beginning to produce broad gains for native-born workers. I'd be encouraged to see that, but the numbers so far haven't convinced me that it's happening broadly.

Energy is another area where I want to resist an overly simple answer. Higher crude and diesel prices are a real cost for households and the businesses that move the things they buy. American producers can earn more and reinvest some of those earnings. Over time, higher prices can bring out more supply and encourage people to use energy more efficiently. That doesn't get the trucking company or the commuter through next month's bills. The timing matters, and the people collecting the additional revenue aren't necessarily the people absorbing the higher costs.

Then there's the bond market, which remains very much on my mind. The ten-year Treasury yield was around 4.95% this morning. Coming back below 5% doesn't make borrowing cheap or remove the government's refinancing problem. It does make me reluctant to treat every move higher as the moment when everything finally breaks. I've been waiting for that moment for a long time. In the meantime, businesses keep adapting and people keep finding ways to build.

This is where I come back to bitcoin. I've never thought we needed to root for the economy to fall apart for bitcoin to make sense. I'd much rather see people build useful things, become more productive and have more left over at the end of the month. My concern is what happens to the value of that surplus when it sits in a money whose supply can be expanded to accommodate a government that refuses to live within its means. Stronger growth would help the fiscal situation. It wouldn't give me much confidence that Washington had suddenly acquired spending discipline.

Bitcoin gives people a way to save outside of that arrangement. Whether they choose to do so, and what they're willing to pay for it today, are separate questions. The recent bitcoin ETF outflows are a reminder that a compelling long-term reason to own something doesn't stop people from selling it. I remain very bullish on bitcoin. I also want to be careful about turning every encouraging economic development into a prediction about what its price must do next.

For now, I'm taking the growth estimate seriously and paying attention to the people building the infrastructure behind it. I still want to see that progress show up in real wages and healthier household finances. There's a lot left to fix. But I'd like to leave myself enough room to be pleasantly surprised, particularly if the alternative is becoming so attached to my expectation of a debt crisis that I miss the good things happening while we wait.

SIGNAL

Flock needs to show its work

The joint investigation from 404 Media and WIRED should force a serious conversation in every town using Flock cameras. After physically obtaining one camera, hackers recovered an encryption key that unlocked stored footage. Other sensitive storage remained encrypted. Reporters found logs spanning roughly 21 days across several periods that recorded about 1.6 million generated images on that one camera. These devices are sitting above public roads collecting the movements of people who never agreed to participate. Physical access should be an obvious part of the security assumptions.

Particularly as AI makes attacks easier to carry out, I have very little patience for asking Americans to accept this collection on faith. Flock promised a summary of independent testing and remediation work this month. Let's see which devices were tested, how the keys and footage are protected, and whether fixes actually reached the cameras in our neighborhoods. The people being monitored deserve those answers.

Spending the bitcoin was the easy part

Gary Brode's Deep Knowledge Investing report on El Salvador, coauthored with Alex Petrou, is worth your time. Gary went through the process of getting bitcoin ready to spend and then actually buying things with it in El Zonte. He describes a fair amount of frustration getting payment apps funded and usable. Once the funds were available, the Lightning payments worked. Anyone who has tried to introduce someone to bitcoin should recognize the distinction.

I also appreciate his point about prices being quoted in dollars. A merchant charging you a dollar price and accepting the equivalent in sats has still accepted bitcoin as payment. We don't need to pretend bitcoin has displaced the dollar throughout El Salvador to recognize that people are using it to buy things. Getting from an interested newcomer to that first successful payment is where we still have work to do.

Crusoe is putting numbers behind the buildout

Crusoe published its AMD inference results yesterday, running gpt-oss-120b and DeepSeek-R1 across 512 MI355X GPUs. The company reports roughly 5.75 million output tokens per second for gpt-oss in the offline benchmark, measured across the whole system, and has released the code and deployment configuration. That is a much more useful starting point than another vague announcement about how much compute someone intends to buy.

I'm interested in the competition this represents. There are people figuring out how to serve open-weight models at scale on hardware from companies other than NVIDIA. If I were buying inference capacity, I'd want to run my own workload on this setup and see what it cost. Crusoe has published enough of its work to make that a much more concrete exercise than comparing marketing claims.

Update your Alby extension

If you use Alby's browser extension, check that you're running version 3.15.0. The maintainers recommend updating after addressing security reports, including tighter handling of payment requests supplied by websites and additional confirmation requirements for signing. One fix changes when a website's automatic-payment allowance is debited, preventing overlapping payment requests from reusing the same available allowance. Giving a website permission to interact with a wallet is a meaningful decision. You should know what you've authorized and periodically remove permissions you no longer need. The fix is specific to payment-allowance accounting; routing fees remain a separate consideration. If you run the connected node, check its fee limits too.

The rare earth that has me paying attention

I watched Maxinomics' video on China's rare-earth leverage the other day and highly recommend it. The material that stuck with me is samarium, used in samarium-cobalt magnets that retain their strength in the high temperatures encountered in demanding military and aerospace applications. Finding the mineral is only the beginning. You need to separate it, turn it into usable metal and alloy, manufacture the magnets, and qualify them for the equipment they're going into. We gave China an incredible amount of leverage by allowing so much of that work to concentrate there.

There is progress. Lynas produced its first samarium oxide in Malaysia in March, giving the supply chain another source outside China. The metal and magnet stages still matter.

I'm encouraged by the desire to build coming out of El Segundo and Austin. We need to turn that enthusiasm into reliable output quickly. China can exert pressure while we're still getting the factories ready.