The Look-Through

The Paranoidist | Issue #20 By Paul Morin | June 22, 2026

There is a number coming that everyone will read as good news, and the danger is that it will be so easy to read that way. Sometime in the next several quarters, if the trajectory holds, the inflation figures that have run hot all year will start to cool, and the cooling will look clean: prices easing, the index drifting back toward target, the long supply shock of the Iran war finally washing out of the data. The Federal Reserve has a story ready for that moment, and it is a flattering one. It will say the patience paid off, that the productivity revolution arrived on schedule, and that artificial intelligence is doing to the price level what every optimist promised. The new chair has been telling a version of that story since before he had the job. The thing that keeps me up is not that the story is wrong. It is that a second, much worse story produces almost exactly the same number, and the Fed has spent the spring rebuilding its instruments in a way that makes the two nearly impossible to tell apart from the inside.

Issue #18, "The Price of a Declaration," left the Fed weighing its own contested independence and a long end of the curve charging a premium for it. Issue #19, "The Run and the Drift," ended on the new chair's first meeting as something happening "tomorrow." That meeting has now happened. On June 17, Kevin Warsh chaired his first FOMC, held the funds rate at 3.5 to 3.75 percent on a unanimous vote, and did three things that matter more than the hold: he stripped the policy statement to a fraction of its former length, dispensed with the forward guidance that had told markets which way the next move would lean, and declined to place his own dot on the projection grid. He also stood up five task forces to overhaul the Fed's operations, one of which will review how inflation itself is measured and what causes it. This issue is about what that combination does to the Fed's ability to read the most important number it will see in the next two years, and about the difference between two disinflations that arrive wearing the same face.

Call them what they are. Productivity disinflation is the good kind: output per worker rises, the same labor produces more, real incomes climb, and prices soften because the economy genuinely got more efficient. Displacement disinflation is the bad kind: prices soften because a growing slice of the workforce has been removed from the wage economy, their lost income is lost demand, and the index falls not because everyone got richer but because a cohort can no longer afford to push it up. In the aggregate price data, the two can look almost identical. One is the economy working. The other is the economy hollowing. And a central bank that has staked its framework on the first being the only one in play is exactly the institution least equipped to notice when the second arrives.

Start with the official version, because it is internally coherent and, on its own terms, defensible. Warsh comes to the chair with a settled view that supply-shock inflation should generally be looked through rather than fought, on the logic that a central bank cannot do much about a war in the Persian Gulf or a closed strait except wait, and that hiking into a supply shock risks crushing demand to cure a problem demand did not cause. Paired with that is his long-stated conviction that AI will prove disinflationary over time, because rising productivity lowers the cost of producing goods and services. Put the two together and you get a framework that is patient with today's hot prints because it reads them as transitory supply effects, and optimistic about tomorrow's cool ones because it expects them as the dividend of a productivity boom already underway.

The data the Fed is looking at does not yet contradict that read. Headline CPI ran at 4.2 percent in May, a three-year high, and the PCE index the Fed prefers was at 3.8 percent in April, both attributable in part to the energy spike the war drove. At the same time the labor market looks sturdy on the surface: payrolls added 172,000 jobs in May and unemployment sat at 4.3 percent, essentially unchanged over a year. To a committee that believes the inflation is supply-driven and the labor market is healthy, the prescription almost writes itself: hold, stay restrictive, let the supply shock clear, and wait for productivity to do the disinflating from here. The revamped projections carry a hawkish tilt, with the median dot now implying a funds rate near 3.8 percent by year end and at least one hike alive. None of this is irrational. On the comfort story's own assumptions, it is the correct policy.

The shorter statement and the missing dot are sold as part of the same competence. Less guidance, the argument goes, means less false precision and more honesty about how data-dependent the committee really is; a chair who will not pre-commit to a path cannot be captured by his own forecast. Read generously, it is humility. The trouble is that the same posture has a second function, and it is the one this issue is about: a Fed that says less, publishes less, and exempts its chair from the accountability grid has also made itself harder to read, harder to second-guess, and harder to catch in the act of misclassifying the very number it is waiting for.

Now read the coming disinflation from the other side. Suppose some meaningful part of the price softening that arrives over the next several quarters is not the productivity dividend at all, but the early demand shadow of a labor transition already visible if you look below the headline. The clearest evidence is not a forecast; it is in the payroll data. A Stanford study released last August, built on records covering millions of American workers, found a roughly 13 percent relative decline in employment for workers aged 22 to 25 in the most AI-exposed occupations since generative AI went mainstream in late 2022, even after controlling for firm-level shocks, with the steepest declines, near 20 percent, in software engineering and customer service. Employment for older workers in the same occupations held steady or grew. And the crucial detail for what follows: the erosion showed up not in wages, which barely moved, but in hiring that quietly stopped happening. The corporate ledger tells the same story in anecdote. Amazon shed roughly 30,000 corporate roles across two waves between October and January, close to nine percent of its office staff, as its CEO openly tied workforce reduction to AI efficiency. Salesforce's CEO has said on the record that AI agents let him cut his customer-support division from about 9,000 people to 5,000. In London, open finance-analyst positions have fallen from more than 350 four years ago to around 80 today.

Here is the mechanism the comfort story leaves out. When a cohort of would-be earners is removed from the wage economy, or never allowed to enter it, the income they would have earned is income that never gets spent, and the demand they would have created is demand that never shows up in prices. A disinflation produced this way looks, in the index, almost exactly like the productivity kind: prices ease, the print cools, the chart bends toward target. But the cause is the opposite of benign. Productivity disinflation is a wealthier economy buying the same things for less. Displacement disinflation is a thinner economy buying less because a slice of it can no longer afford to buy at all. The first is the dividend. The second is demand destruction wearing the dividend's clothes. And the cruelest feature of the resemblance is that the worse the displacement, the better the inflation number looks, which means the gauge the Fed is waiting on improves fastest in precisely the scenario it should fear most.

This is the run-versus-drift structure of the last issue, ported from the currency to the labor market. There I argued that the most consequential monetary shift of the decade was arriving as a drift no institution was built to catch. Displacement disinflation is the same shape of problem. It does not announce itself with a recession's bang. It seeps in as soft prints and a hollowing entry tier, individually explicable, collectively invisible, and on the official scoreboard it reads as success right up until the composition of the economy underneath the index has quietly changed.

The reason the Fed is poorly placed to tell these two apart is not incompetence. It is the design of the instruments. Begin with the dual mandate's headline gauge, the unemployment rate. At 4.3 percent and stable, it is the single number that most reassures the committee that demand is healthy and the labor market tight. But displacement of the kind the Stanford data describes does not, at first, move the headline rate, because it operates through hiring rather than firing. A 22-year-old who never gets the first job does not register as unemployed the way a laid-off 45-year-old does; he registers as still in school, as not yet started, as out of the labor force, as stagnant young-worker employment sitting underneath an aggregate that still looks fine. The same study's authors noted that their findings help explain why employment growth for young workers has gone flat while overall employment stays resilient. The headline rate, in other words, is exactly the wrong instrument for catching a transition that hides in who gets hired rather than in how many get fired.

Layer the new framework on top of that blind spot. A committee that has decided to look through supply-shock inflation has trained itself to discount hot prints as noise, which is fine until the habit extends, by reflex, to discounting the cool prints as the awaited productivity payoff. The chair's prior that AI is disinflationary is not a neutral lens; it is a hypothesis the institution is now invested in confirming, and a falling price index is exactly the evidence that confirms it. Then add the task force on how inflation is measured. I am not suggesting anything sinister; reviewing measurement is legitimate. But consider the position it creates. The same institution that owns the 2 percent target, that has staked its chair's credibility on delivering it, and that expects AI to help deliver it, is now also reviewing the ruler by which delivery is judged. When the body that sets the target, holds the prior, and grades the result all sit under one roof, the risk is not fraud. It is motivated perception: a thousand small, defensible interpretive choices that all lean the same way, toward reading the good number as the good story. The shorter statement and the missing dot complete the enclosure, because they remove the external handholds, the published path and the chair's own forecast, that outsiders would otherwise use to check the committee's reading against its own prior words.

None of this requires bad faith. It requires only that an institution built to want one answer be handed a number consistent with that answer and with its opposite, and be given fewer instruments than before for telling which it is looking at.

What makes this more than an academic worry is the timing, because three things are arriving in the same window and they interact. The first is the sticky energy premium from Issue #19's neighbor problem: even with a ceasefire nominally in place, the Hormuz risk premium has not fully unwound, and headline inflation will stay elevated longer than the underlying picture warrants. The second is the Fed's resulting hawkish bias: as long as the print runs hot on energy, the committee has cover to stay restrictive, and the dots say it might even hike. The third is the displacement disinflation building underneath. Stack them and the trap takes its sharpest form. The Fed holds policy tight to fight a supply-driven headline number, while the demand-driven softness from labor displacement builds in the part of the economy the headline rate cannot see, so that restrictive policy is being applied directly into an emerging demand shock the framework has pre-classified as benign productivity. The energy premium keeps the Fed's foot on the brake exactly as the part of the economy it is braking loses demand for a reason the brake cannot fix and was never meant to address.

The fiscal side compounds it, and here the threads of the last two issues converge. A Treasury leaning toward bill-heavy issuance and quietly courting the runnable stablecoin demand from Issue #19 needs a tame front end and a cooperative funding environment. A Fed staying restrictive into a stealth demand shock raises the odds of a policy error that shows up first at exactly that front end, in exactly the contested-independence premium Issue #18 located there. The three pieces I have written this spring, the declaration, the run and the drift, and the look-through, are not three economies. They are one economy seen from three windows, and the windows look onto the same room.

The honest forecast comes in three parts, and only the first is close to mechanical. Near term, the comfort story holds and looks vindicated. The energy premium fades, the supply shock washes out, the inflation prints cool, and the Fed reads the cooling as the productivity dividend arriving on cue. Nothing in this stage looks like a problem, and that is exactly the point: the likeliest near-term outcome is that displacement disinflation gets misfiled as productivity disinflation in real time, because the two are genuinely hard to separate and the institution has every incentive to choose the flattering label. The scoreboard reads as a win, and stays that way for a while.

The medium term is where the uncertainty lives, and it turns on two questions that cannot be put on a calendar. The first is whether the displacement deepens enough to break through the composition mask into the headline numbers, the moment the hollowing entry tier finally drags the aggregate unemployment rate up or pulls consumption down in a way no one can attribute to anything else. The second is whether the Fed, having spent a year staying restrictive on energy and reading the resulting disinflation as success, recognizes the misclassification in time to stop tightening into the demand shock, or whether recognition arrives only after the policy error is already in the system. Neither has a due date. Both grow likelier the longer the energy premium gives the committee cover to hold and the productivity prior gives it a reason to like the falling number.

The counter-move, when it comes, will not be a clean reversal. Expect instead a slow reinterpretation: a pivot in the data relabeled after the fact, a discovery that the disinflation was "broader-based than expected," a measurement review that quietly reframes what the index was telling us all along. The historical pattern for a central bank that has misread the character of a shock is not a crisp mea culpa but a gradual change of subject, and the cost of that gradualism is borne in the lag between when the demand shock started and when policy stopped fighting the wrong war.

So the base case is not a crash. It is a legitimized misreading that holds as long as a short list of gauges stays quiet. Watch the gap between the headline unemployment rate and the youth and entry-level employment series, because that gap is the displacement hiding in composition. Watch real consumption against real wages, because demand softening faster than the average wage explains is the demand shadow showing through. Watch the labor-force participation of the young, because a cohort that never enters is the drift the unemployment rate misses. Watch what the inflation-measurement task force actually changes, because the ruler is now a variable. And watch whether the committee's language ever distinguishes productivity disinflation from demand-destruction disinflation at all, because as long as it does not, it cannot tell you which one it is steering into. The scoreboard will keep reading as a win right up until the morning one of those gauges says the falling number meant the other thing.

Four areas of board exposure, mapped against the displacement read rather than the comfort story.

Workforce-heavy services and anyone doing multi-year headcount planning. If your labor model assumes the entry pipeline keeps refilling itself, it may be calibrated for a world that is closing. The exposure is concrete: a hollowing junior tier means the cohort you would normally promote into mid-level roles in three to five years is not being hired now, which is a succession problem that compounds silently and surfaces as a leadership-bench gap long after the decisions that caused it. The question for the board is whether your workforce plan treats the entry-level contraction as a temporary cost saving or as a structural change in how human capital enters the firm, because those two readings imply opposite decisions this year.

Consumer-facing businesses watching their own pricing power. If your demand softens over the next several quarters, the comfort story will tempt you to read it as a macro tailwind, disinflation easing input costs and a friendlier rate environment coming. The displacement read says to check first whether the softness is concentrated in the younger and entry-level cohorts whose incomes are eroding, because demand destruction dressed as disinflation hits some customer segments and not others, and an expansion or pricing decision built on the benign reading can walk straight into the malign one. The question is whether your demand model can distinguish a customer base getting more efficient from one getting poorer, since the top line can look the same either way.

Rate-exposed balance sheets and duration managers. This is the interest-rate version of the trap. A Fed staying restrictive on a hot energy-driven headline while a demand shock builds underneath is a Fed that can be late in both directions, too tight for too long and then forced to reverse harder than the curve is priced for. The interaction with the front-end fragility from the last two issues is the dangerous part: a policy error read into a Treasury market already charging a premium for contested independence and already holding a runnable stablecoin creditor is an error with more than one way to transmit. The question for the rates book is whether your positioning survives a scenario in which the Fed fights the wrong inflation with the right-looking number for longer than consensus expects.

AI capital allocators and technology strategists. The firms deploying AI to cut labor cost are, in aggregate, the mechanism of the displacement disinflation, which makes them both its beneficiaries and, eventually, its exposure. The same automation that lowers your cost base lowers someone's income, and enough of that in aggregate lowers the demand your own products eventually face. There is a fallacy-of-composition risk here that no single firm internalizes: each automation decision is individually rational and collectively demand-destroying, the labor-market cousin of the bank run no depositor means to start. The question for the strategy committee is whether your demand forecasts assume a customer base whose income is unaffected by the very efficiency you are selling.

The productivity story may simply be true. AI may be doing exactly what the optimists and the chair expect, raising output per worker and lowering the cost of goods and services in a genuine, broad-based dividend that shows up as benign disinflation because it is benign disinflation. On that reading, the displacement signals are the transition friction that always accompanies a productivity boom, real but temporary, and the new higher-skilled, higher-paid roles backfill the hollowed entry tier the way technology-driven job creation has done for nearly a century. If that is the world we are in, the Fed's read is good judgment, not a blind spot, and I am pattern-matching a healthy transition to a sinister one.

The displacement may be modest, and partly cyclical. The same firms cutting staff are also unwinding post-pandemic over-hiring into a restrictive rate environment, and a fair reading attributes much of the softness to the cycle rather than to automation; the Stanford authors controlled for firm-level shocks, but disentangling AI from a tech downturn that AI also helped cause is genuinely hard. Salesforce's own CEO has called blaming AI for layoffs a scapegoat narrative, and AI is cited explicitly in only a single-digit share of layoff announcements. If most of the "displacement disinflation" is ordinary cyclical disinflation, the unemployment rate is an adequate gauge for it and my composition story is a layer of complexity the data does not require.

The Fed may not be as blind as I claim. The committee watches far more than the headline rate, including youth and entry-level series, participation, and a wide array of wage and flow data, and a chair who built his reputation on skepticism of comfortable narratives is not an obvious candidate to be captured by one. The shorter statement and missing dot may reflect exactly the data-dependent humility they are advertised as, and the measurement task force may sharpen the Fed's perception rather than bias it. If the institution is in fact alert to the distinction I am drawing, the look-through risk is one its own process is built to catch, and I am underrating the people in the room.

And the two disinflations may be more separable than I allow. They do leave different fingerprints if you look at the right series, output and margins rising under one, consumption and low-end demand falling under the other, and a competent central bank with good data can in principle tell them apart in close to real time. If the fingerprints are clear enough, the resemblance I have built the whole issue on is more apparent than real, and the misclassification I fear is a forecasting problem the Fed can solve rather than a structural blindness it cannot.

The DeepStrategy.ai signature method requires sorting risk, which is quantifiable, from uncertainty, which is not, at every major analysis. The numbers here are risk. Headline CPI was 4.2 percent in May and PCE 3.8 percent in April; the funds rate is 3.5 to 3.75 percent and the median dot now points near 3.8 percent; unemployment is 4.3 percent and May payrolls added 172,000; early-career employment in the most AI-exposed jobs is down on the order of 13 percent since late 2022; the large banks' own technologists project headcount cuts measured in the hundreds of thousands over the next several years. These can be tracked, charted, and updated on a schedule, and any board can put them on a dashboard tomorrow.

What will not resolve to a figure is everything that matters most. Whether the disinflation that arrives is productivity or displacement, and in what mix, is an uncertainty, because the same falling index is consistent with both. Whether the Fed perceives the difference in time, or files the malign number under the benign story, is an uncertainty about an institution's interior judgment that no data series reports. Whether a framework staked on AI optimism and stripped of its external handholds is exercising humility or building an enclosure is an uncertainty. And whether the measurement of inflation itself, now under review by the body the measurement judges, stays a fixed ruler or becomes a moving one is an uncertainty sitting underneath all the others. The price index is a risk and tells you the size of the move; whether that move is the economy getting richer or a cohort getting poorer is an uncertainty, and the two answers point in opposite policy directions from the same number. As always, the precise, trackable figure is the announcement, and the part that actually moves the institution is the part that will not reduce to a probability.

The institution that consumes the analytical process as preparation for multiple futures has what the forecast cannot provide: adaptability. The headlines, when the inflation finally cools, will say the Fed held its nerve and the productivity revolution delivered. The architecture says something quieter and harder to celebrate. A falling price can mean an economy that got more efficient or an economy that lost a slice of its earners, and those two economies are nearly indistinguishable in the one number everyone will be watching. The Fed has told us it will look through the inflation it does not want to fight. The worry is the second meaning of the phrase: that having trained itself to look through the number it dislikes, it will look straight through the one it likes, and never see the labor shock standing directly behind it.

The boards best positioned for this are not the ones cheering the cool print as vindication. They are the ones asking, of every softening number they see, whether it is the dividend or the drain, and pricing the difference before the scoreboard tells them which one it was.

A number that falls. A story that flatters. And a labor shock hiding in the one place the gauge was never built to look.

The Paranoidist publishes weekly, with flash issues when events warrant.

If this changed how you think about one thing, consider subscribing. If it didn't, tell me what I'm missing.

Paul Morin is the founder of DeepStrategy.ai, author of Uncertainty: When Risk Is Not Enough (a guide to decision-making when probabilities fail), and publisher of The Paranoidist, BoardroomRadar, and ScenarioWatch. He has spent more than three decades in entrepreneurship, finance, risk management, and insurance, which is why he worries about the things that keep other people awake at night.

Researched, written, and edited in collaboration with Claude by Anthropic.