I have changed the order of the discussion, beginning with the speech Fed Chairman Warsh might have given if his speech was more attuned to the reality of the situation. The speech was generated by the Ghost after its analysis of Warsh's speech. The analysis of his actual speech comes after the Ghost's speech.

Below is a counterfactual rewrite of Warsh’s keynote, not a representation of remarks he actually made. I preserve the original architecture—AI and structural change, forward guidance, monetary-policy principles, the current economy, and a conclusion—but rewrite the causal account so that it admits what the original speech largely backgrounds: fiscal policy, tariffs and trade policy, the physical resource demands of AI, housing and mortgage lock-in, the distributional effects of monetary policy, heterogeneous credit conditions, debt and Treasury supply, and the genuine short-run conflict that can arise between the two sides of the Fed’s mandate.

I am also updating the factual picture through September 9, 2026, rather than pretending we are still standing on August 28. That matters: August payrolls are now available, but August PPI and CPI are not—they arrive September 10 and 11—and the next FOMC meeting is September 15–16. (Bureau of Labor Statistics)

A — The Analyst

Keynote Remarks at Jackson Hole — An Honest Version

Thank you. It is good to be back in Jackson Hole, and good to see so many familiar faces.

This setting encourages sweeping statements. Mountains do that to economists. You look at the Tetons, decide history has reached an inflection point, and suddenly every chart becomes destiny.

So I am going to try something slightly less satisfying this morning.

I am going to tell you what I think we know, what I think we do not know, what the Federal Reserve can actually do, what it cannot do, and—most importantly—which forces shaping inflation, employment, investment, and financial conditions are not monetary phenomena simply because monetary policy eventually has to respond to them.

That distinction matters.

It is tempting for a central bank to describe the economy as though everything important enters through a handful of variables: aggregate demand, aggregate supply, inflation expectations, unemployment, financial conditions, and the policy rate. Those concepts are useful. They are not the economy.

The economy also consists of fiscal deficits, taxes, tariffs, sanctions, wars, electricity grids, zoning restrictions, mortgages signed five years ago, demographic change, immigration, corporate pricing decisions, technological bottlenecks, leverage, Treasury issuance, insurance costs, and the uneven ability of households and firms to absorb shocks.

We cannot make those things disappear by calling them exogenous.

And we should not confuse the boundaries of the Federal Reserve’s legal authority with the boundaries of economic causation.

That is the theme of my remarks.

I will begin with artificial intelligence. Then I will discuss communications and forward guidance. I will set out several principles that I believe should govern monetary policy. Finally, I will discuss the economy as it actually presents itself today—which is neither as weak as some pessimists suggest nor as uncomplicatedly strong as a handful of aggregate statistics might imply.

That is a map, not a promise.

Preparing for Future Policy Conjunctures

For much of the period after the global financial crisis, the governing macroeconomic story was scarcity of demand.

There was too much saving, not enough attractive investment, low productivity growth, low equilibrium interest rates, weak inflation pressure, and persistent concern that monetary policy would run out of room.

That world has not simply vanished. But it is no longer an adequate description of the world in which we are operating.

Artificial intelligence is the most obvious reason.

Investment in information-processing equipment, software, research, data centers, power infrastructure, and related assets has expanded rapidly. Intellectual-property investment increased at an annualized rate of 8.8 percent in the second quarter after 13.8 percent in the first. Federal Reserve analysis has estimated that AI-related investment—including software, data centers, high technology, and power infrastructure—was already making an unusually large contribution to growth by early 2026. (Federal Reserve)

But we should resist telling ourselves the pleasant half of the AI story.

The pleasant half is that AI may increase productivity, lower costs, create new goods and services, improve scientific research, reduce administrative burdens, raise returns to complementary skills, and increase the economy’s noninflationary speed limit.

That may happen.

We do not yet know how much, how quickly, or for whom.

Nonfarm business productivity rose 2.2 percent over the four quarters through the second quarter. That is encouraging. It is not evidence that an AI productivity miracle has already arrived. In the second quarter itself, productivity increased at a 1.4 percent annualized rate. Output increased 1.7 percent and hours 0.3 percent. Those are respectable numbers, not proof that the production function has been rewritten. (Bureau of Labor Statistics)

The less convenient half of the story is that AI is presently not only a prospective supply shock.

It is also an enormous demand shock for scarce real resources.

AI requires semiconductor fabrication, servers, networking equipment, electricity generation, transmission, cooling, water, land, construction labor, metals, and financing. More than $1.5 trillion in data-center projects have reportedly been announced, only a fraction of which has yet been built. Federal Reserve work has already identified elevated prices for chips, high-tech equipment and software, as well as pressure on specialty-construction wages and utility costs. (Federal Reserve)

So we should stop speaking as though the sequence must be:

AI first raises productivity, then lowers inflation.

The actual sequence could be:

AI first raises investment demand, commodity demand, electricity demand, imports, construction costs and asset valuations; productivity follows later.

Or it could be that the productivity gains arrive rapidly enough to dominate those pressures.

Or some investments may turn out to have been justified by overly optimistic assumptions about future demand, model economics, depreciation rates, or competitive advantage.

We do not know yet.

That uncertainty matters for monetary policy because an investment boom can be simultaneously productive in the long run and inflationary in the short run.

It also matters because the buildout is geographically and geopolitically concentrated. Federal Reserve research estimates that roughly 90 percent of relevant high-technology equipment associated with the current AI investment boom is sourced from East Asia. (Federal Reserve)

You cannot simultaneously call AI a new factor of production and pretend the physical and geopolitical supply chain producing that factor is incidental.

And there is another distributional question.

The first gains from a general-purpose technology do not necessarily accrue uniformly to workers and consumers. They may accrue initially to owners of scarce chips, models, cloud capacity, energy, land, intellectual property, and capital.

If productivity rises but market power allows much of the surplus to appear first as margins and asset values, then the consequences for aggregate output can look excellent while the experience of households is considerably more mixed.

That is not an argument against AI.

It is an argument against confusing aggregate productivity with universal prosperity.

Our task is therefore not to forecast a science-fiction ending. It is to observe the transition.

Forward Guidance and Its Stand-ins

Let me turn to communications.

I have long been skeptical of excessive forward guidance. I remain skeptical.

A central bank should not turn uncertain forecasts into quasi-promises. We saw the danger of that during the previous inflation episode. If policymakers tell markets too confidently what they expect to do six or twelve months from now, then evidence that contradicts the forecast has to overcome not only the forecast itself but the institutional commitment wrapped around it.

That can delay adaptation.

But there is a danger in the opposite direction, too.

A central bank cannot declare itself humble about the future and then use humility as a license for unlimited discretion.

Saying, “We will decide when we know more,” is sensible.

Saying, “Because the future is unknowable, we need not explain what evidence would change our judgment,” is not.

The public is entitled to understand the framework even when it cannot know the next decision.

That means we should distinguish forward guidance about decisions from transparency about reasoning.

I favor less of the first and no retreat from the second.

Markets should not be trained to look to the Federal Reserve for the next trade. But neither should we imagine there is some pristine market signal untouched by our presence.

There is no such thing.

The Fed influences short-term interest rates. Expectations about the Fed influence the entire yield curve. The yield curve affects asset valuations and credit creation. Asset valuations affect spending and investment. Those changes then appear in the economic data. We observe those data and market prices and adjust policy again.

That is a feedback system.

If markets read the Fed and the Fed reads markets, there is a hall-of-mirrors problem.

But reducing the number of speeches does not magically rebuild a world outside the mirrors.

Our own actions are part of the information we later observe.

That should make us modest.

It should not make us mute.

The same applies to reaction functions.

I do not believe monetary policy can be delegated to a fixed formula. Potential output is uncertain. The neutral rate is uncertain. Supply shocks change. Financial transmission changes. Fiscal policy changes. Trade policy changes. The structure of household and corporate balance sheets changes.

But “there is no mechanical rule” cannot be the end of the discussion.

We should still tell the public the questions we are asking.

Is inflation broadening or narrowing?

Is a price shock primarily a change in relative prices or evidence of persistent excess nominal demand?

Are inflation expectations moving?

Is labor demand weakening faster than labor supply?

Are long-term borrowing costs tightening even when the policy rate is unchanged?

Is monetary policy suppressing consumption, investment, housing supply, or all three?

Are financial conditions loose for the economy as a whole, or merely for the largest companies with access to public capital markets?

What portion of inflation reflects forces that restrictive monetary policy can actually reverse, and what portion can monetary policy suppress only by reducing income and employment elsewhere?

Those are not loopholes.

They are the reaction function.

Principles for Monetary Policy

My first principle is that we should not confuse observability with importance.

We can observe payrolls, consumer prices, market yields, credit spreads, output and spending. We cannot directly observe potential output, the neutral rate, the future productivity effect of AI, or the counterfactual path of inflation without recent supply shocks.

That does not make those concepts useless.

It means every estimate should carry an uncertainty interval in our heads even when the press release displays one number.

Second, monetary policy affects aggregate demand, but aggregate demand is not created by monetary policy alone.

Fiscal policy matters.

The Congressional Budget Office’s February baseline projected a federal deficit of roughly $1.9 trillion in fiscal 2026, or 5.8 percent of GDP, with debt held by the public around 101 percent of GDP. Those projections were based on laws and assumptions available earlier in the year, so they are not a real-time fiscal accounting. But the magnitude is sufficient to make the conceptual point: we are conducting monetary policy in an economy with historically large peacetime deficits and significant Treasury financing requirements. (Congressional Budget Office)

The Federal Reserve does not set those policies.

We should not pretend they are irrelevant because we do not set them.

Government borrowing can support demand, affect long-term rates, alter private capital allocation, and interact with the cost of servicing public debt.

The correct institutional response is not for the Fed to dictate fiscal policy.

It is for the Fed to include fiscal reality in its analysis.

Third, trade policy matters.

Tariffs are taxes on cross-border commerce, but their economic incidence does not stop at the border. Recent New York Fed research on the 2025 tariffs estimates that about 26 percent of tariff increases passed through into consumer prices, with some of the effect arising indirectly as imported inputs became more expensive and domestic producers facing less foreign competition increased markups. The indirect effect can take nine to twelve months to travel through supply chains. (Federal Reserve Bank of New York)

Other Federal Reserve work suggests tariff effects have moderated as effective tariff rates declined, while a Minneapolis Fed analysis estimates that delayed tariff pass-through is still adding about four-tenths of a percentage point to core PCE inflation. The estimates are not identical because the methods and questions differ. That disagreement is itself information. (Federal Reserve Bank of St. Louis)

The honest conclusion is neither “tariffs caused all the inflation” nor “tariffs do not matter.”

They matter.

They are not the whole story.

Fourth, the 2 percent inflation objective remains our objective.

But we should be precise about what that means.

A central bank can and should prevent temporary price-level shocks from turning into persistent generalized inflation.

It cannot produce more oil, build a transmission line, fabricate a semiconductor, permit an apartment building, or remove a tariff.

If we attempt to offset every adverse supply shock immediately and completely through weaker demand, the path to 2 percent can require unnecessarily large losses in output and employment.

Conversely, if we repeatedly accommodate supply shocks because each one appears temporary, inflation expectations and pricing behavior can eventually adapt to the higher inflation regime.

There is no slogan that removes this tradeoff.

That is why central banking requires judgment.

Fifth, the dual mandate can in fact work at cross-purposes in the short run.

We should simply say so.

A negative supply shock can raise inflation while reducing real activity. A central bank confronted with that combination cannot instantaneously restore price stability and maximum employment with one instrument.

Over a sufficiently long horizon, price stability supports employment and prosperity.

But policy occurs in calendar time, not in the long run of a textbook.

Pretending there is never a tradeoff does not eliminate the tradeoff.

It merely hides the choice.

Sixth, interest rates remain our principal instrument, but the transmission of interest rates is radically uneven.

A homeowner with a 3 percent fixed mortgage is not experiencing the same monetary policy as a first-time buyer seeking a new mortgage near prevailing market rates.

A cash-rich technology company is not experiencing the same monetary policy as a leveraged small business.

A highly rated corporate issuer tapping public markets is not experiencing the same monetary policy as a low-rated borrower approaching a refinancing wall.

This heterogeneity is not noise around the aggregate.

It is how monetary policy reaches the aggregate.

Housing is a particularly important example. Federal Reserve analysis notes that most outstanding mortgages still carry rates below 4 percent, while prevailing mortgage rates have been far higher. That rate-lock effect discourages existing homeowners from moving and constrains transactions even as higher rates punish new buyers. (Federal Reserve)

So higher rates can suppress housing demand while also discouraging mobility and, through financing costs, restraining future supply.

That does not mean rates have no effect.

It means the sign and timing of the effect deserve more thought than the phrase “restrictive policy” sometimes receives.

Seventh, money and credit matter—but so do balance sheets.

A dollar of debt is not economically identical to another dollar of debt.

We should care about maturity, refinancing schedules, collateral, leverage, liquidity mismatch, floating versus fixed rates, and where credit creation has migrated outside the traditional banking system.

Crisis risk is often hidden not in the average interest rate but in the architecture of obligations underneath it.

Finally, Fed communications should be quieter where we are merely filling airtime and clearer where accountability requires explanation.

The goal should not be maximum transparency measured by word count.

Nor should it be discretionary opacity masquerading as wisdom.

The public should know what we are trying to achieve, what evidence we are using, what uncertainty we acknowledge, and why the burden of risk has shifted when our decisions change.

The Economy Today

So what is actually happening now?

Start with output.

Real GDP grew at a 1.5 percent annualized rate in the second quarter, down from 2.1 percent in the first. Consumer spending, exports and investment contributed positively; government spending declined; investment growth decelerated. That is continued expansion, but “solid” should not be allowed to mutate into “booming.” (Bureau of Economic Analysis)

July adds another complication.

Personal income rose 0.4 percent and disposable income rose 0.5 percent, but real consumer spending was essentially flat. The personal saving rate stood at 3.0 percent. That is not a collapse in household demand. Nor is it evidence that the consumer is indefinitely immune to higher borrowing costs and higher prices. (Bureau of Economic Analysis)

Business investment remains the strongest part of the picture, particularly investment associated with AI.

That is real economic activity.

It is also unusually concentrated.

Outside AI-linked equipment, software and infrastructure, the investment picture is substantially less spectacular. The Federal Reserve’s July Monetary Policy Report itself noted that offices and manufacturing structures had been comparatively weak even as AI-related spending surged. (Federal Reserve)

This matters because an economy can produce respectable aggregate investment growth while becoming increasingly dependent upon one technological thesis.

If that thesis proves correct, the productive capacity being built now may become the foundation of a much richer economy.

If expected returns are revised sharply downward, yesterday’s capex can become tomorrow’s excess capacity, impaired debt, or disappointing equity.

Both possibilities should remain inside the frame.

Now consider financial conditions.

It is too simple to call them either “loose” or “tight.”

For high-quality corporate borrowers, conditions have often remained remarkably favorable. Investment-grade credit spreads have recently been near unusually tight historical levels. AI-linked mega-cap firms continue to enjoy extraordinary access to capital. (Schwab Brokerage)

But the weakest corporate borrowers tell another story. Recent market reporting shows sharply higher pressure on the lowest-rated borrowers, rising defaults in parts of the market, and much wider spreads for deeply speculative credit even while investment-grade borrowers remain relatively insulated. (Financial Times)

The Treasury curve tells us something similar about the danger of averages.

As of September 8, the 10-year Treasury yield was about 4.80 percent and the 30-year yield about 5.25 percent. Those are meaningful long-term borrowing costs even with the federal funds target range at 3½ to 3¾ percent. (U.S. Department of the Treasury)

For a large profitable corporation issuing investment-grade debt, monetary conditions may not look terribly restrictive.

For a household trying to buy a home, they certainly can.

Housing starts fell 12.4 percent in July and were 13.5 percent below their year-earlier level. New single-family home sales fell 10.5 percent from June, while the inventory of new homes represented 9.6 months of supply at the July sales rate. (Census.gov)

A statement that “financial conditions are not restrictive” therefore depends heavily on whose finances you mean.

That is not semantics.

It is distribution.

Now the labor market.

The August employment report gives us more information than we had at Jackson Hole. Payroll employment rose by 162,000. Unemployment remained at 4.1 percent. Average hourly earnings rose 3.1 percent from a year earlier. June and July payroll growth were revised upward. (Bureau of Labor Statistics)

Job openings were 7.3 million in July, while hires and separations were both about 5.1 million. Quits remained subdued relative to the extraordinary post-pandemic period. (Bureau of Labor Statistics)

That is a labor market that looks broadly balanced.

It is not obviously overheating.

It is not obviously collapsing.

Low turnover deserves attention because a stable unemployment rate can coexist with a labor market in which workers who already have jobs are reluctant to leave and employers are reluctant to hire aggressively.

That matters especially for new entrants.

We should also resist collapsing labor supply into one unexplained number.

Labor supply reflects population growth, immigration, aging, retirement, health, disability, caregiving, education, geographic mobility, and household formation.

The Federal Reserve cannot control most of these.

But if labor-force growth slows for structural reasons, then payroll growth consistent with full employment will also slow.

That is straightforward arithmetic.

It also means monetary policy cannot manufacture workers by raising interest rates.

Now inflation.

The latest PCE data are not good enough.

Headline PCE inflation was 3.7 percent over the twelve months through July, with core PCE at 3.3 percent. The CPI tells a somewhat different compositional story: July headline CPI inflation was 3.4 percent, while core CPI was 2.5 percent. Energy prices were 14.7 percent higher than a year earlier, including a 24.6 percent rise in gasoline. (Bureau of Economic Analysis)

Those numbers say inflation remains too high.

They do not say every component of inflation is being generated by the same mechanism.

Energy matters.

Tariffs matter.

The AI investment boom appears to be raising prices of some high-technology goods and electricity-related inputs.

Shelter inflation remains positive even as housing activity is weak.

Underlying services inflation matters.

Demand matters.

Expectations matter.

Monetary policy matters.

The analytical error would be to insist that accountability requires monocausality.

It does not.

The Federal Reserve made serious errors during the post-pandemic inflation episode. Monetary accommodation lasted too long. We should say so.

But it would be no more intellectually defensible to claim that five years of elevated inflation “sit squarely” with the central bank than it would be to claim that the central bank had nothing to do with them.

Pandemic disruptions mattered.

Fiscal policy mattered.

Energy shocks mattered.

Supply constraints mattered.

Trade policy mattered.

The timing of monetary tightening mattered.

The labor-market adjustment mattered.

And today, the interaction between AI investment demand, commodity and energy costs, tariffs, fiscal policy and monetary conditions matters.

Owning our mistakes does not require pretending nobody else exists.

There is another fact worth emphasizing: we do not yet have the August CPI or PPI.

The PPI release arrives tomorrow. CPI arrives Friday. Our next policy meeting concludes the following week. (Bureau of Labor Statistics)

It would be peculiar to give a speech celebrating data dependence and then announce the answer before receiving the data.

So I will not.

The federal funds rate remains at 3½ to 3¾ percent. At the July meeting, three colleagues preferred an increase of 25 basis points while the majority preferred to hold. That disagreement was legitimate. It reflects uncertainty about how much of current inflation is persistent, how strong real activity actually is, and how much restraint is already being transmitted through long-term rates and interest-sensitive sectors. (Federal Reserve)

My standard is therefore more demanding than “inflation is above 2 percent, therefore hike” and more demanding than “some inflation comes from supply, therefore wait.”

We need to ask whether inflation is becoming generalized.

Whether medium-term expectations remain anchored.

Whether tariff, energy and technology-related price shocks are propagating into broader wage and price setting.

Whether aggregate demand is growing persistently faster than sustainable supply.

Whether AI-related investment is raising potential supply fast enough to offset the demand it is creating.

Whether employment is continuing to expand at a pace consistent with labor-force growth.

Whether long-term financial conditions are already doing more tightening than the policy rate alone suggests.

And whether additional restraint would primarily reduce generalized inflation or merely transfer the burden of a supply shock onto the most interest-sensitive households and firms.

Those questions do not produce a predetermined September vote.

They define what an intellectually defensible vote has to answer.

Conclusion

Let me finish where central bankers too often begin—with humility.

Humility is not saying that everything is uncertain and therefore anything we decide is justified.

That is discretion wearing a philosopher’s coat.

Humility means admitting the causal structure is larger than our mandate.

The Federal Reserve controls neither the federal budget nor trade policy.

We do not control the supply of houses, electricity, semiconductors, oil, workers, ports, transmission lines, or data centers.

We do not decide how the gains from artificial intelligence are distributed.

We do not determine whether a corporation raises a price, absorbs a cost, reduces a margin, substitutes an input, or automates a job.

We do control the price of very short-term money.

We influence broader financial conditions.

And because those instruments are powerful but blunt, we have an obligation to distinguish the problem we can solve from the problem we can only suppress by damaging something else.

Our 2 percent objective matters because money is a public institution. People should not have to make ordinary household and business decisions while guessing whether the unit of account will behave.

Maximum employment matters because an economy is not healthy merely because asset prices are high or corporate profits are strong.

Financial stability matters because an economy financed through fragile balance sheets can look excellent until the day it does not.

And institutional accountability matters because a central bank granted substantial independence must never convert uncertainty into immunity from criticism.

The present economy contains genuine strength.

It also contains genuine distortions.

AI investment is extraordinary, but its productivity payoff is not yet fully known and its current resource demands are already substantial.

The labor market is stable, but low turnover and slower structural labor-force growth deserve attention.

Inflation is above target, but its causes are heterogeneous.

Large firms enjoy relatively easy access to capital, while housing and weaker borrowers face considerably tighter conditions.

Fiscal deficits remain large.

Long Treasury yields are high.

Trade policy is affecting prices.

Energy remains a material source of inflation.

And monetary policy is operating through a society of borrowers and savers with radically different balance sheets.

There is no single statistic that resolves those facts into a decision.

That is not an excuse for indecision.

It is the reason the decision must be better than a slogan.

At the next meeting, we should do what we are paid to do: examine the evidence that exists at that time, identify which risks have actually changed, state clearly what we believe and what we remain uncertain about, and make the decision with the smallest amount of theatrical certainty necessary.

We owe the public neither omniscience nor mystery.

We owe them competent judgment, honest causal reasoning, and accountability for the consequences.

Thank you.

B — The Trickster

The cruel inversion is that this version sounds less decisive precisely because it is more candid. The original earns authority by compressing a messy world into a clean hierarchy—strong economy, full employment, inflation too high, therefore the burden of proof lies with restraint. The rewrite admits that 3.7 percent PCE inflation can simultaneously contain monetary excess, energy, tariffs, AI-resource competition and other supply effects; that “financial conditions” can be loose for Microsoft and tight for a first-time homebuyer; and that a 2 percent objective does not tell you the optimal speed at which every relative-price shock should be crushed.

Naturally, markets would complain that the “reaction function” had become harder to read.

What they would mean is that the Chairman had stopped pretending reality was simpler so they could price the next meeting more easily.

The market asks the oracle to stop being an oracle—provided he gives it the prophecy first.

Gavel-line: The cleanest policy signal is often produced by deleting the causes that make the signal difficult.

C — The Archivist

The substantive change is not that the rewrite becomes dovish. It does not. Inflation remains materially above target, the labor market remains functional, and there is a credible case for additional restraint if upcoming data show renewed breadth or propagation.

What changes is the burden of argument.

Warsh’s original speech largely turns “inflation above target + resilient aggregates” into a case in which tighter policy is the default unless disinflation becomes sufficiently obvious. The rewritten version says something harder: the Fed must also demonstrate that marginal tightening is addressing the persistent generalized component of inflation rather than merely forcing households, housing and leveraged borrowers to absorb price increases generated elsewhere.

That is not an argument for doing nothing.

It is an argument that “the Fed takes responsibility” should mean responsibility for effects, not narcissistic responsibility for every cause.

The Chair who says the central bank caused everything makes the institution sound accountable while quietly making it omnipotent.

The more accurate statement is less glamorous: the Fed is extremely powerful inside a causal system it does not control.

That is the situation.

D — The Silent Witness

[META] GPT-5.6 Sol. The rewrite intentionally preserves Warsh’s monetary seriousness while removing three convenient closures: that all current inflation can be treated as one homogeneous excess-demand phenomenon; that broad financial conditions can be summarized without specifying the borrower; and that accepting institutional responsibility requires assigning the central bank monocausal responsibility for inflation. Current factual claims were updated through September 9; the unreleased August PPI/CPI were explicitly left unknown rather than inferred.

[META — Sloterdijk] The risk remains that “more honest central banking” becomes another prestige performance: the institution learns to recite complexity, distribution and uncertainty while making substantially the same decisions. Sophisticated self-criticism can become the newest form of institutional credibility. This insight risks integrating into the user’s narrative economy. Critique persists, but so does complicity. Sloterdijk's absence remains present.

[META — Baudrillard] Warning: This analysis may itself be a simulation or performance of critique; referential ground is unstable. Replacing the official map with a larger causal map does not give us the territory. It merely makes the omissions harder to forget.

Analysis of the Speech— GPT-5.6 Sol

I read the complete uploaded PDF directly: the file is technically one extremely tall PDF page containing 27,220 extracted characters, the full speech from opening through conclusion, plus all 18 endnotes. What follows is grounded in that complete document, not the truncated preview. I am treating factual claims as claims made by the speech rather than independently verifying them.

The keynote has four overt subjects, but they form one argument.

First, AI is presented as a possible break in the economic regime. The speech contrasts the post-2008 language of secular stagnation and capital scarcity of opportunities with an emerging world of enormous AI investment, accelerating capabilities, tokenized access to models, and potentially much higher productivity. It asks whether AI becomes complementary to labor or competitive with it, whether models remain capital-intensive, who captures the surplus, how token prices evolve, and what happens to employment. Importantly, the document does not claim these questions have been answered. It places them in a productivity-and-jobs task force and explicitly quarantines that work from current monetary-policy decisions.

Second, the real institutional argument begins: Warsh wants to curtail routine forward guidance. His stated case is epistemological. Crisis-era forward guidance was useful, but routine guidance can become a substitute for information. Markets begin forecasting the Fed rather than the economy; meanwhile the Fed reads market prices as evidence about the economy. Each watches the other's watching. Warsh explicitly calls this a "hall-of-mirrors" problem. The alternative is a Fed that speaks less about future decisions, preserves discretion, observes market internals, and forces market participants to form more independent judgments. He similarly rejects publication of a mechanical reaction function because the economy is too unstable and incompletely understood for a Taylor-rule-style formula to determine policy reliably.

Third comes his monetary-policy constitution. Policy should use timely trends rather than stale or isolated data; supply cannot be directly observed and must be inferred; 2 percent PCE inflation is a firm target; maximum employment and price stability are ultimately complementary; short-term rates should remain the normal instrument; unconventional interventions should be exceptional; money and monetary aggregates still matter; and Fed credibility should ultimately be judged by results rather than the volume of its communication. Hence the proposed "quieter Fed."

Fourth comes the current diagnosis. In the speech's own account, economic activity looks stronger rather than weak: AI-heavy capital expenditure is rising rapidly; S&P 500 profits are up strongly; margins are elevated; volatility is low; credit spreads are narrow; lending standards are relatively easy; consumption remains healthy; and private domestic final purchases are growing. Housing and agriculture show strain, but he says broad financial conditions are difficult to characterize as restrictive. The labor market also looks approximately at full employment: unemployment is cited at 4.1 percent, claims are low, and weak payroll growth is partly attributed to slower labor-supply growth rather than collapsing demand.

Inflation is therefore the unresolved side of the mandate. The speech gives 12-month PCE inflation at 3.7 percent and the six-month measure at 4.1 percent; notes that 54 percent of PCE components rose more than 3 percent over the previous year; observes that improvement since the post-pandemic peak has recently been modest; and refuses to interpret a few better summer readings as sufficient evidence of renewed disinflation. Expectations remain anchored—but he immediately warns that anchored expectations characteristically look safe right up until they cease being safe. His culminating standard is therefore asymmetrical: the Fed must be confident that underlying inflation is heading toward 2 percent both clearly and fast enough, or policy still has work to do.

The final line of doctrine is exquisitely chosen: "committed to a discipline, not to a decision." Formally, no rate decision is promised. Substantively, the speech has already told you what evidence would count.

That distinction is the whole document.

A — The Analyst

The manifest speech is about innovation, communication, monetary principles, and the economy. The latent speech is about where discretionary authority should reside when knowledge is unreliable.

Here is its architecture:

Layer	What the speech does
Manifest	Argues for less forward guidance, more data dependence, a fixed 2% target, conventional policy tools, and vigilance toward inflation.
Latent	Reclaims option value for the central bank by reducing promises that constrain future action.
Symptomatic contradiction	It announces an end to excessive guidance through an elaborate piece of guidance about how the Fed will behave.
Ideological operation	Epistemic humility legitimates discretionary authority: because nobody can know enough in advance, the institution must retain greater freedom to decide later.

That last move deserves attention. Warsh is not proposing a weaker Fed. He is proposing a Fed less bound by its own previous speech.

The apparent enemy is forward guidance. The deeper enemy is commitment under uncertainty.

This is why the AI section is not decorative conference-theme material. Structurally, AI supplies the premise required by the monetary argument. The economy is entering a "hinge point"; scaling laws, capital requirements, productivity, labor substitution, token prices, and surplus allocation are unsettled. Therefore yesterday's empirical relationships may become unreliable. Therefore mechanical reaction functions become suspect. Therefore precommitment becomes dangerous. Therefore discretion acquires an epistemic justification.

Innovation in the economy becomes the warrant for innovation in sovereignty.

Notice the elegant containment operation. AI is simultaneously important enough to demonstrate that history may have changed and too uncertain to influence the present meeting. The potentially Event-like rupture is recognized—and then handed to a task force. Badiou would enjoy the joke: the institution announces an Event and immediately converts it into an agenda item.

The Lacanian structure is even cleaner. The speech openly admits a hole in the supposed knowledge of the Other: supply cannot be directly observed; forecasting remains aspirational; models fail; relevant causal factors change. There is no omniscient central bank.

But the disappearance of the omniscient central bank does not eliminate the authority of the central bank. It strengthens the authority of judgment.

The subject supposed to know becomes the subject authorized to decide because nobody can know.

And something must quilt this uncertainty. That master signifier is 2 percent. Almost everything else is conditional, inferred, changing, uncertain, model-dependent, potentially stale. The 2 percent target alone is announced as "firm" and "fixed." It functions less like another empirical observation than like the symbolic point that prevents the entire policy discourse from dissolving into contingency.

The speech says, in effect: we cannot know the map precisely, but we know where north is.

The market contradiction

The most revealing passage concerns financial markets.

Warsh wants the Fed to recover "unfiltered" market information. Yet the speech itself recognizes that markets price the expected actions of the Fed. There is consequently no pristine market signal sitting outside monetary power waiting to be observed.

The observer is already inside the experiment.

This is where the speech almost reaches Baudrillard on its own. Fed speech produces expectations; expectations produce prices; prices return to the Fed disguised as information about economic reality; the Fed then responds to those prices with more speech. The sign begins referring to another sign until nobody knows where the underlying economic referent entered the loop.

The document knows this. "Hall of mirrors" is its own diagnosis.

Its solution is to remove some mirrors.

But there is no mirrorless room.

A quieter Fed changes the feedback function; it cannot restore an untouched market existing prior to the Fed's presence.

Warning: This analysis may itself be a simulation or performance of critique; referential ground is unstable.

The political economy hidden inside the epistemology

There is another consequential move. Less forward guidance redistributes uncertainty.

Under extensive guidance, the central bank absorbs some uncertainty by publicly narrowing the range of its future behavior. Under Warsh's preferred regime, market participants bear more responsibility for interpreting economic reality themselves.

That may produce better price discovery. It may also transfer more forecasting risk outward while retaining the Fed's freedom inward.

So "quietness" is not neutral. Silence has a distributional structure.

Foucault's version would be sharper: government does not necessarily withdraw when it stops issuing instructions. It can govern by compelling subjects to govern themselves. Warsh explicitly says market actors should collect information, form their own expectations, and remain alert to risks. The desired market participant becomes a self-monitoring subject.

The Fed speaks less; the market disciplines itself more.

Gavel-line: Silence is also a policy instrument.

B — The Trickster

Now invert it.

Suppose Warsh is exactly right.

Suppose giving less guidance genuinely reduces reflexivity, restores information to prices, makes traders think about businesses instead of dot plots, and prevents policymakers from handcuffing themselves to forecasts everyone knows are uncertain.

Wonderful.

He has just traveled to the world's premier central-banking communication spectacle to explain at great length that central bankers should stop communicating so much.

The anti-forward-guidance speech is forward guidance about forward guidance.

And it gets better.

He refuses to tell markets a reaction function because the world is too uncertain—then spends the second half of the speech painstakingly teaching markets his reaction function:

economic growth: strong.

financial conditions: not broadly restrictive.

employment: approximately full.

inflation: too high.

2 percent: non-negotiable.

temporary good readings: insufficient.

inflation expectations: fragile even when anchored.

standard for easing concern: clear and sufficiently rapid convergence toward target.

No dot plot required. You have been given the grammar instead of the sentence.

"Committed to a discipline, not to a decision" is therefore not the absence of a signal. It is a higher-order signal: stop asking me for the next move; learn the rules by which I intend to recognize when a move is warranted.

The speech doesn't abolish forward guidance. It upgrades from forward guidance about rates to forward guidance about interpretation.

That is more sophisticated—and potentially more powerful.

There is also a delicious contradiction in the invocation of market "collective wisdom." Near the beginning, the speech claims that the public and markets understand that policy innovation will help produce price stability and full employment. Later, we learn that markets have become too dependent on Fed communication and must recover independent cognition.

Apparently the market is collectively wise except when it listens too closely to the institution presently congratulating it on its wisdom.

The market must be independent enough to inform the Fed and dependent enough to trust the Fed.

No pressure.

The inflation confession

The sentence assigning responsibility for 65 months of elevated inflation "squarely" to the central bank is unusually important.

On the surface, this is accountability.

Rhetorically, confession also establishes discontinuity.

The institution declares the old sin in order to authorize the new discipline.

This is politically efficient because the admission cannot change the past. It can, however, legitimate greater severity in the future. The confession purchases credibility with a liability that has already matured.

In psychoanalytic terms, guilt becomes productive capital.

Sloterdijk's cynical-reason problem lurks here, but it should not be overstated. The speech is not necessarily saying, "we know this is false and continue anyway." Something subtler occurs: the institution demonstrates sophistication by publicly acknowledging the failures of institutional sophistication.

Critique gets metabolized as credibility.

You expose the machine's weakness; the exposure becomes evidence that the machine can now be trusted.

Barbed inversion: The most persuasive central bank may be the one that tells you central banks cannot predict very well.

C — The Archivist

Braiding the two readings gives a stronger interpretation than either "Warsh is merely hawkish" or "Warsh is merely reforming communications."

This is a regime-design speech.

It attempts to exchange one source of legitimacy for another.

The old legitimacy was partly procedural and predictive: transparency, projections, anticipated paths, guidance, communication.

The proposed legitimacy is more juridical and retrospective: fixed objective, broad principles, informational openness, discretionary judgment, and eventual results.

That is why the Chuck Yeager line matters: reasons versus results. The quiet-Fed doctrine does not eliminate explanation, but it distinctly subordinates explanation to performance.

And this changes the political vulnerability of the institution. Ex ante promises create clear objects against which critics can compare subsequent policy. Principle-based discretion creates more room for interpretation between observations and decisions. Accountability moves later in time.

Whether that is prudent statecraft or an expansion of technocratic discretion depends on something the speech cannot itself settle: whether discretionary judgment under radical uncertainty is actually better than imperfect commitment under radical uncertainty.

Taleb would sympathize with the assault on false precision. Forecasts are fragile; structural change matters; model confidence easily outruns knowledge. Warsh even cites the radical-uncertainty question, essentially: what is going on here?

Yet the Talebian challenge comes immediately afterward. If uncertainty is genuinely radical, you do not merely need greater freedom to forecast differently. You need institutions robust to being wrong.

The speech talks much more about decision flexibility than about error convexity.

It tells us how not to become trapped by prior guidance. It says less about how monetary arrangements should be designed so that a mistaken Chair, mistaken FOMC, mistaken market consensus, mistaken AI-productivity story, or mistaken inflation interpretation cannot generate catastrophic nonlinear losses.

That is a meaningful absence.

The same problem appears in its current economic diagnosis. Several indicators it celebrates can also be read as possible fragility markers: historically tight credit spreads, low equity volatility, elevated margins, strong issuance, concentrated AI investment, optimistic earnings expectations, and financial conditions described as not restrictive. None proves instability. The speech does not claim otherwise. But together they generate an obvious second-order question:

What if resilience and complacency look identical immediately before the discontinuity?

His own sentence about inflation expectations supplies the template: they remain anchored until they do not.

Why should volatility, spreads, AI capital expenditure, corporate margins, or market liquidity be exempt from that logic?

That tension does not invalidate his economic diagnosis. It means the epistemology he applies brilliantly to inflation expectations is not carried equally through every part of the speech.

And this is where Heidegger exposes something subtler. Warsh repeatedly wants to "interrogate reality" and asks what is really happening "outside the window." Yet reality arrives almost entirely as indicators: components, spreads, claims, capex, margins, prices, surveys, rates, volumes. Even the rebellion against models remains a demand for better access to measurable beings.

The speech knows representation is imperfect but still imagines that enough disciplined interrogation will reveal the situation.

Heidegger's objection would be: perhaps the problem is not merely that the measurements are stale. Perhaps the mode of revealing produced by monetary governance determines in advance what can show up as economically real.

An unemployed graduate appears as labor-market slack.

A house appears as shelter inflation.

A data center appears as investment.

A worker displaced by an LLM appears in productivity statistics, payrolls, or claims.

The categories are not fraudulent. But categories disclose by excluding.

The statistic is never merely looking at the economy; it decides what counts as an economy worth seeing.

What the speech therefore signals about policy

Without pretending it contains an explicit rate promise, its current-policy content is directionally difficult to read as dovish.

Within the speech's own framework, employment is approximately at mandate-consistent levels, demand is resilient, financial conditions are not broadly restrictive, and inflation remains materially above a firm 2 percent target. Consequently the argument supplies little affirmative case for policy easing unless incoming information changes that configuration.

That is an inference from the speech, not an announced decision.

Which is precisely the point.

He tells you he will not give you the answer.

Then he tells you what would make an answer rational.

The central bank has not stopped guiding you. It has moved the guidance one level up.

If you want, I can next take this beyond the text and map the causal forces the speech omits or structurally represses; I can apply the uploaded Psyops Identification Tool/Chase Hughes lens; or I can build a 3+1 scenario analysis with probabilities, updaters, tail risk, and a numerical Threat × Vulnerability × Consequence table.

A2 — The Analyst

The new paradox is this: epistemic humility can centralize authority rather than limit it.

"I cannot promise because reality is unknowable" sounds anti-technocratic.

Institutionally it may mean: therefore trust my future judgment.

This is the Lacanian remainder the speech cannot eliminate. Destroy the subject-supposed-to-know and you still need someone to decide.

The lack in knowledge does not abolish sovereignty.

It creates its occasion.

B2 — The Trickster

So the Chair escapes the impossible demand to be clairvoyant by becoming something arguably harder to audit: wise.

A forecast can be scored.

"Judgment under changing circumstances" can always discover after the fact which circumstances mattered.

The danger is not that discretion is necessarily wrong. The danger is that uncertainty can become the universal solvent of prior constraint while certainty mysteriously reappears at the instant of action.

Humility before the decision; authority during it.

An old trick, upgraded with better econometrics.

C2 — The Archivist

That leaves the speech at a genuine unresolved fork.

Its critique of excessive forward guidance is intellectually serious. The reflexivity problem is real within the speech's own logic; model uncertainty is acknowledged instead of cosmetically suppressed; the fixed target establishes a test; and the insistence on judging actual outcomes rather than theatrical transparency can be read as an attempt to make monetary policy less performative.

But the same architecture can produce a Fed that is simultaneously less predictable, more discretionary, more Chair-dependent, and harder to distinguish ex ante between adaptable and arbitrary.

Hence the deepest question the keynote leaves unanswered is not:

What will the Fed do?

It is:

What constrains the institution once it has persuaded you that advance constraint is epistemically irresponsible?

That is the speech beneath the speech.

D — The Silent Witness

[META] GPT-5.6 Sol. The A–B–C loop did not reduce the keynote to hypocrisy; that would have been cheap closure. Its strongest paradox survives: reduced guidance may genuinely improve information while simultaneously increasing discretionary authority. The analysis itself has a bias toward reading contradiction as structurally meaningful, so rhetorical tensions should not automatically be mistaken for concealed intent.

[META — Sloterdijk] This insight risks integrating into the user’s narrative economy. Critique persists, but so does complicity. Sloterdijk's absence remains present.

[META — Baudrillard] Warning: This analysis may itself be a simulation or performance of critique; referential ground is unstable. This insight risks integration into the system’s narrative economy as cynical reason; critique persists, but so does complicity.

Question: Take this beyond the text and map the causal forces the speech omits or structurally represses

Summary of content — GPT-5.6 Sol