The Bureau refers to the Bureau of Labor Statistics — the federal agency that produces America’s monthly employment report. Every first Friday of the month, the curtain goes up. This issue, we review the performance.
Every month the Bureau puts on a performance. The lights come up, the number drops, the market applauds. May’s show was spectacular — 172,000 jobs, nearly double what anyone expected. The audience rose to its feet.
But look closely at the stage. Half the cast isn’t there. They were conjured by a statistical model, assigned to businesses that may not exist, scheduled to be quietly walked back eleven months from now in a footnote nobody will cover.
The chandelier is still hanging. The Phantom is still in the building.
May’s 172,000 non-farm payrolls came in nearly double consensus estimates. What almost no mainstream coverage mentioned: a meaningful share of that figure wasn’t counted — it was calculated. Specifically, it was produced by the BLS Birth-Death Model, a statistical tool that estimates jobs created by businesses too new to survey and jobs lost by businesses too recently closed to track. Those estimated jobs are added to the actual survey count before the headline number is published.
The model’s track record has become impossible to defend with a straight face. In January 2027, the BLS will conduct its annual benchmark review — reconciling monthly estimates against unemployment insurance records covering 98% of actual U.S. jobs. The last three reviews erased 485,000, 818,000, and 898,000 jobs respectively. Not rounding errors. Record-breaking corrections, each larger than the last, each arriving quietly while the market had already moved on.
For the full year 2025, the average monthly job gain after revisions was approximately 15,000 — not the 48,000 being reported in real time. The headline and the reality were three times apart.
There is a legitimate counterpoint. In February 2026, the BLS reformed the model, incorporating live sample data for the first time. The 2026 monthly birth-death figures have been running roughly half of 2025’s comparable months — January 2026 came in at 39,000 versus 127,000 the prior January. May 2025 registered 129,000; the 2026 figure appears to be in the 60–65,000 range. A smaller phantom contribution is better than a larger one. It does not make the phantom disappear.
Read this chart: The 2026 birth-death model has been running dramatically lower than 2025 following the February 2026 methodology reform. May 2025 registered 129K. The estimated 2026 May figure is approximately 63K — roughly half. A smaller phantom, but a phantom nonetheless.
The household survey — the one that asks actual people — told a more honest story. Unemployment held at 4.3% for the eleventh consecutive month. Long-term unemployment rose by 524,000 over the year, now representing 27.5% of all unemployed. Labor force participation sat at 61.8%, still below its pre-pandemic peak. The jobs being created are real for some Americans. For a growing number, the labor market is structurally closed.
Read this chart: Four consecutive years of record-breaking downward revisions. The 2025 correction of −898,000 is the largest on record. This is not random error. It is a structural and systematic overcount of jobs that the annual audit corrects — quietly, in a footnote, after the market has moved on.
“When the chandelier drops in January 2027, the audience will have already gone home.”
— The Kool-Aid Diaries
The market fell significantly after the May report. The irony is worth sitting with: a jobs number nearly double expectations triggered a selloff, because strong employment removes the last excuse for a Fed rate cut. The market wasn’t reacting to economic reality. It was reacting to the loss of a fantasy.
Markets have been pricing in rate cuts for the better part of two years. Every month that inflation stays elevated, the thesis gets pushed back — and every month, a meaningful cohort of institutional money still acts as if the pivot is just around the corner.
The clearest evidence of how deeply this thinking is embedded: the IPO valuations currently being floated for SpaceX, OpenAI, and Anthropic — each in the range of $300 billion to over $1 trillion. These are not grounded valuations. They are what happens when institutional money, still running on the muscle memory of zero-interest-rate-era multiples, meets narrative-driven assets with no current earnings to discipline the math.
A business earning $1 billion annually is worth roughly $16–20 billion at historical market multiples. Price it at $300 billion and you are paying 300 times earnings — assuming earnings will grow fast enough, for long enough, at low enough discount rates, to justify it. That assumption requires cheap money indefinitely. Cheap money requires a Fed that cuts. The Fed cannot cut while inflation runs above its target. The valuation chain collapses at the first link. This isn’t a commentary on whether these are great companies. It is a commentary on what the market is still refusing to price in.
Markets have been punishing gold and silver on the logic that elevated interest rates make yield-bearing assets more attractive than non-yielding hard assets. The analysis is not wrong on its own terms. It is incomplete in a way that costs people money.
Your savings account pays 5% interest. Inflation is running at 4.5%. Your nominal return is 5%. Your real return — what you actually gained in purchasing power — is 0.5%. Now inflation rises to 6% while the nominal rate stays at 5%. Your real return is now negative. You are losing purchasing power despite earning interest. That is the environment we are moving into. Nominal rates remain elevated. But real rates — nominal minus actual inflation — are falling. When real rates fall toward zero or go negative, the historical case for hard assets becomes very strong.
In the 1970s, nominal interest rates rose for much of the decade. Gold still increased roughly 20-fold between 1971 and 1980. The market that was selling gold because rates were rising was making the same category error being made today: looking at the nominal number and missing the real one. The market is punishing precious metals for the wrong reason, using the wrong measurement.
Read this chart: Nominal rates have stayed elevated. But real rates — nominal minus CPI — have been declining as inflation remains persistent. When the real rate line approaches or crosses zero, the environment historically favors hard assets. That environment is now.
The Bureau conjures phantom jobs every month. But the Bureau isn’t the only institution in America running a ghost accounting system.
Corporate America has its own phantoms. They show up every morning, answer emails, carry the workload of colleagues eliminated to fund AI capital expenditure — and they do it for the same paycheck they received before the layoffs. Studies show salaried workers are now logging an average of 8 additional unpaid hours per week. A full extra day. Every week. The salary line doesn’t change. The hours do. Nobody measures it. Nobody reports it. It disappears into the productivity statistics that make the economy look healthier than it is.
The Bureau counts jobs. It doesn’t count phantom hours. And in the gap between what is measured and what is real, the K-shaped economy does its quietest work.
The K-shape is not a COVID artifact or a policy failure. It is a system operating exactly as designed. Picture the letter K. The upper stroke rises. The lower stroke falls. Asset owners ride the upper stroke. Wage earners ride the lower one. Every crisis widens the gap. Every recovery lifts the top while the bottom barely stabilizes before the next disruption arrives.
The phantom jobs in this month’s report? Lower stroke. The 524,000 additional long-term unemployed rising behind the headline? Lower stroke. The $1.3 trillion in credit card debt used to buy groceries at 22% interest? Lower stroke. The record asset prices, the nominal wealth highs, the trillion-dollar AI valuations? Upper stroke. Exclusively. The arithmetic has been running in the same direction for four decades.
You want to understand how a system actually works? Skip the rhetoric. Read the incentive structure. Two Americans each generate $100,000 in a year. One earns it as a salary. One earns it as long-term capital gains. Here’s the full picture — federal, state, and the piece almost nobody includes: Social Security.
| Tax Category | Wage Earner | Capital Gains Earner |
|---|---|---|
| Federal income tax | $17,400 | $15,000 |
| State & local (est. avg.) | $5,500 | $2,000 |
| FICA — employee share | $7,650 | $0 |
| FICA — employer displacement | $7,650 | $0 |
| Total estimated burden | $38,200 | $17,000 |
| Effective total rate | 38% | 17% |
| After-tax income | $61,800 | $83,000 |
Same $100,000. The wage earner takes home $61,800. The capital gains earner takes home $83,000. A $21,200 difference — more than two months of gross income — explained entirely by the form the income took. Not effort. Not risk. Not contribution to society. The form.
Now consider the full opportunity cost of being a wage earner rather than a capital gains earner over a thirty-year career. Three numbers compound against you simultaneously, every single year.
First: $16,700 in unpaid labor from the phantom hours — 16.7% of your $100,000 salary working for free, every year. Second: $15,300 in combined FICA — money that could have been invested but instead funds a retirement system that replaces roughly 40% of your pre-retirement income. Third: $21,200 in additional taxes paid simply because your income came from labor rather than capital.
Added together: $53,200 per year that the wage earner surrenders and the capital gains earner does not. Compounded at 9% annually over 30 years, that $53,200 per year becomes approximately $7.25 million in lost wealth accumulation. In today’s purchasing power — discounted at a 5.5% real rate — that $7.25 million is worth approximately $1.45 million in current dollars. The honest, inflation-adjusted figure.
But here is what makes it visceral. At a 7% withdrawal rate, that $7.25 million generates $507,600 per year before tax. After a 25% effective capital gains rate, that is approximately $381,000 per year — after tax — for life. Nearly $381,000 annually in passive income — from money the system extracted from you a dollar at a time, every year, for thirty years.
Not because the wage earner worked less hard. Not because they made worse decisions. Because of the form their income took.
These aren’t loopholes. They are architecture. America has never cared about stakeholders — just shareholders. The tax code is the proof.
Make no mistake: Social Security is not a retirement supplement. It is a tax — one you hope younger generations will be able to pay when you become a beneficiary. And that hope grows more uncertain by the year. Artificial intelligence threatens to automate the very jobs that fund it. Student loan debt is suppressing the household formation and savings capacity of the generation that will be asked to write the check. The math was never great. The future makes it ominous.
Read this chart: The full burden comparison — federal, state, and FICA — on identical $100,000 income. The wage earner’s 38% effective rate versus the capital gains earner’s 17% is not a loophole. It is the system working as designed.
Now add the phantom hours. A salaried worker earning $100,000 — contracted for 2,080 hours annually — is actually working closer to 2,496 hours after the unpaid 8 hours weekly are factored in. Their effective hourly rate drops from $48.08 to $40.06. A 16.7% real wage cut that never appears on a pay stub.
Now layer on real inflation — not the government’s carefully adjusted CPI figure, but what your grocery bill, insurance renewal, rent, and car repair invoice have been telling you since 2022. Independent measures and real-world basket analysis put cumulative consumer price increases at 18% to 20% or more for typical American households — roughly twice the official figure. That same $100,000 salary now buys what approximately $82,000–$83,500 bought three years ago.
Combined — phantom hours plus real inflation erosion — the worker’s true compensation has declined somewhere between 30% and 35%. Not the number that makes the economy look manageable. The number most Americans already know because they live it every month.
Read this chart: Four measures of the same $100,000 salary across four years. Nominal salary unchanged. CPI-adjusted purchasing power declining. Effective hourly rate after unpaid hours declining further. Combined real compensation — the number nobody reports — down 30–35% from 2022.
Because 7.3 million Americans are unemployed. Because long-term joblessness is rising. Because the mortgage doesn’t care about your grievance. The employer holds the leverage. Silence is rational. And rational silence at scale becomes invisible exploitation at scale. The phantom hours inflate productivity statistics and mask labor deterioration simultaneously — a mirror image of the Bureau’s phantom jobs, running in the same direction.
Human economic organization has evolved through distinct phases. Hunter-gatherer. Agricultural. Industrial. Service. Each transition disrupted the existing labor structure, concentrated wealth in the hands of those who owned the new means of production, and eventually resolved — through reform or rupture.
We are in the service phase now. Seventy percent of GDP is consumer spending. That single fact contains a structural vulnerability almost nobody states plainly: a service economy is the first economic model in history that requires its own workforce to remain broadly solvent to function. A manufacturing economy exports. A service economy sells to itself. When the consumer runs out of money the engine doesn’t slow — it seizes.
The next phase — the one the K-shape is either accelerating toward or permanently foreclosing for most Americans — is the passive income economy. The logical next step in the civilizational arc is a phase in which individuals participate not primarily as sellers of labor but as indirect owners of productive capital — receiving returns on investment, sharing in productivity gains, generating the passive income base that makes the consumer economy self-sustaining rather than self-liquidating.
This is achievable — with one prerequisite the current system is systematically destroying: savings. And before anyone dismisses this as theoretical, two nations have already proven the model works at scale.
Norway’s Government Pension Fund Global was built on North Sea oil revenues beginning in 1990. It is now valued at over $1.7 trillion — the largest sovereign wealth fund in the world. Every Norwegian citizen is an indirect beneficiary. The fund owns approximately 1.5% of every listed company on earth. Norway’s public services, retirement security, and intergenerational wealth are not dependent on any single generation’s labor income. That is the passive income economy — in practice, at national scale.
Singapore runs two sovereign wealth vehicles — GIC and Temasek — managing an estimated $700–$900 billion combined. Temasek holds direct ownership stakes in Singapore Airlines, DBS Bank, and major global companies. Together they fund Singapore’s budget, subsidize housing, and backstop the retirement system — not through taxation of labor, but through the returns on national ownership of productive capital.
The United States has no equivalent. The closest analog is the Alaska Permanent Fund — which distributes annual dividends to every Alaska resident from oil revenues — covering one state, averaging roughly $1,000–$2,000 per person per year. The model exists and works. The political will to replicate it at national scale does not. Every investment requires capital first. Every passive income stream requires an asset base. And Americans — squeezed by phantom hours, eroded by real inflation, taxed at 38% effective rates on their labor, carrying record credit card balances at 22% interest — are being structurally prevented from accumulating the savings that would allow them to make the transition from labor income to ownership income.
Artificial intelligence accelerates this dynamic by automating not just physical labor but judgment, expertise, and professional knowledge — the foundation of the professional middle class. The trillion-dollar valuations for OpenAI, Anthropic, and SpaceX are not just market exuberance. They are a map of where the upper stroke of the next K is being drawn. A very small number of people are drawing it. The rest of us are performing in their opera.
Read this chart: Both series indexed to 100 in 1990. Median wages have grown modestly in real terms. Asset prices — stocks, real estate, financial wealth — have compounded dramatically. The gap between the two lines is the K-shape rendered visually. It has been widening for over three decades.
This newsletter doesn’t moralize. The system is what it is. So what do you do with that knowledge?
Become an owner. Not because it is fair — it isn’t. But because remaining purely a seller of labor in a system explicitly designed to reward ownership is a guaranteed losing position. Own assets. Own equity. Own productive capital. Put your money where the tax code puts its thumb on the scale.
Put your money with the crooks. Not cynicism. Literacy.
History gives us two precedents for how extreme wealth concentration during technological transitions resolves. Reform. Or rupture. The K-shaped economy — accelerated by AI, sustained by a tax code that rewards owners, operating inside a consumer economy that requires broad solvency to function, systematically preventing the savings accumulation that would allow workers to transition to the next phase — is not a stable equilibrium.
The Fifth Economy will be defined by who owns the intelligence and who was able to save enough to get there. The phantom hours, the phantom jobs, the phantom wages — they are all part of the same performance. The Kool-Aid Diaries will keep watching.
The BLS cannot survey a business that doesn’t exist yet. And it doesn’t always know immediately when one closes. So every month, it runs a statistical model that estimates the net jobs created by new business births and destroyed by business deaths — companies too new or too recently gone to appear in the actual survey sample. That estimated figure gets added to the real survey count before the headline number is published. You never see the seam.
The model is built on five years of historical data from unemployment insurance records. It learns what typical business formation looks like and projects forward. In normal times, the errors are small. In abnormal times — post-pandemic dislocations, rapid interest rate changes, structural shifts in business formation — the model keeps forecasting normal when the economy is doing something else entirely. It doesn’t know what it doesn’t know.
Here’s why the revisions are almost always downward: the model is structurally more likely to overestimate births than deaths. New businesses are optimistic — they often register with unemployment insurance before they hire significantly or survive long. Closed businesses stop filing, but the model’s lag means their estimated contribution lingers in the count longer than it should. The bias runs in one direction. The annual benchmark correction runs in the same direction — down — every single time.
The practical takeaway: every monthly payroll headline you read is partly a measurement and partly a forecast. The measurement gets corrected once a year. The forecast almost always overstated reality. Reading the birth-death adjustment figure alongside the headline — available the same day at bls.gov — takes thirty seconds and tells you how much of the number to treat as a phantom.
The phantom jobs. The phantom rate cuts. The phantom wages. The phantom valuations. Every month the Bureau lights the fog machine. Every month the market applauds the performance. Every month the chandelier holds — until January, when it doesn’t.
The Kool-Aid is being served like a witch’s brew. All smoke!