Read the source · this Framework is our reading of their public work · Steno Research →
Synthos Research · Frameworks · Thesis Snapshot
Andreas Steno: how he actually thinks
This is not a profile. It is a working model of Andreas Steno's worldview — his causal
beliefs, his quantified transmission lags, his own stated falsifiers (including one that fired on schedule),
what he changed his mind about (dated), and where he is silent — reconstructed from 1,041 dated claims he made
between July 2023 and April 2026, and tested against 116 claims the model never saw. His shows host many
voices; every claim beneath this Framework, and every one of the 116 test claims, was verified as his own voice
before it was used — a first for this series. Every causal link here has been adversarially verified against
the record, and every quote on this page is a dated, verbatim fragment from our claim record.
Jul 2023 → Apr 2026 · his voice only · 116 later claims held out for testing
durable beliefs
10
the spine — never reversed while the dollar view flipped six times
recorded mind-changes
12
incl. one retreat that created a new belief
direction fidelity
73.5%
25 of 34 out-of-sample tests, strict
Model fidelity: 73.5% strict direction on 34 out-of-sample tests
What that number means: we froze this model on his claims through April 13, 2026, asked it to predict
how Steno would react to real market events from the following three months, and graded those predictions
against the claims he actually made in that window — claims the model had never seen.
25 of 34 gradeable predictions got his direction right, strict — 8 partials and exactly 1 outright miss.
The pass bar, set before grading began, was 60%. With half-credit for partials: 85.3%.
Zero invented mechanisms: the model never reasoned through a causal link that had failed adversarial
verification against the record — including three predictions that explicitly declined the tempting dead path.
The honest flaw, one: the model underestimates his conviction. It predicted he would fade the rate-hike
consensus at conviction 60–70; he did it at 90. Where he runs hot, the model runs lukewarm.
The honest flaw, two: the model mean-reverts his winners; the man rides them. It took profit on the
April software long that the real Steno was doubling down on. His entries are this model's strong side; his
exits are its weak side, and we say so.
The numbers, both directions: his stored coefficients all survived adversarial audit against the
record — yet in the test window he kept doing the same lag arithmetic with different numbers. We scored
the stored coefficients 0-for-3 verbatim, 3-for-3 in idiom, and disclose both.
Silences: 9.5 of 13 predicted silences were correct (~73%) — and the misses are named: he engaged
autonomy and a mega-IPO the model expected him to ignore.
Eval integrity: the first Framework tested entirely on speaker-verified claims — all 116 test-window
claims are his own voice, none a co-host's or a guest's.
1 · How the world works, according to Steno
Steno has no eras. Where other thinkers pivot, his record shows a single operating
system across 33 months: a slow spine of structural beliefs that never moved — while on top of it he rotated
tactically at speed, flipping his dollar view six times inside three months. The beliefs below are the spine.
Liquidity is issuance and plumbing — QE and QT are theater
Held November 2023 → April 2026 · his most load-bearing belief
What moves markets is Treasury issuance and the money-market plumbing — the TGA, the RRP, bank reserves, repo —
not the Fed's balance-sheet headline. The belief ends the corpus doing live work, in a dollar-figure call on the
funding gap.
QE was largely a head fake— November 2023
watch Yellen's issuance, not Powell, for liquidity— January 2024
Estimates a ~$250bn bank-reserve funding gap into year-end— December 2025
Fiscal dominance: a 7% deficit outlaws the layoff recession
Held October 2023 → January 2026 · born from his own retreat (section 4)
Perpetual large deficits prop growth, raise the neutral rate, and make the classic layoff-driven recession
nearly impossible. This belief did not arrive as a thesis — it arrived as the mechanism that killed his own
recession call, then stayed.
hard to get big layoff waves with the deficit running ~7%— January 2024
very hard to get a deep recession while the government keeps pumping money out— July 2024
easy fiscal plus tighter monetary policy is the new regime— 2024
Crowded positioning is the fade signal
Held July 2023 → April 2026 · the full span of the record
The question is never what the market knows; it is what the market is missing. Consensus positioning is not
information — it is the other side of his trade.
ask what the market is missing, not what's already known— August 2023
buy the most-hated, net-short names— March 2026
Real-time hard data beats surveys and first prints
Held November 2023 → January 2026
Freight rates, electricity output, withheld employment taxes, pollution readings, Trueflation, port congestion —
he trades the nowcast and treats official first prints and PMIs as head fakes to be checked against it.
withheld employment taxes are reaccelerating, electricity output is up— July 2024
Trueflation prints below 1.5% YoY vs official ~3%— March 2025
Shale broke the geopolitics-to-oil link — spikes are for selling
Held February 2024 → April 2026 · applied live to Hormuz at the record's edge
Since US shale scaled, turmoil no longer buys you an oil rally. Oil is a supply story, and the geopolitical
spike is not the risk — it is the exit.
US shale expansion since ~2000 broke the old 'turmoil equals higher oil' relationship— February 2024
geopolitical spikes from Israel-Iran exchanges mark exact peaks to sell— October 2024
China is structurally broken — its stimulus is always too small
Held August 2023 → July 2025
A balance-sheet recession that only a nationwide bad-debt program could fix — and every announced package falls
short of that bar by design.
chopsticks to a gunfight— on China's stimulus, November 2024
no nationwide bad-debt program to fix it— September 2024
Mechanical selloffs are buyable overreactions
Held June 2024 → April 2026
When the selling is plumbing — liquidity air pockets, forced de-grossing, positioning stop-outs — nothing
fundamental has changed, and the overreaction is the opportunity.
a weekend liquidity air pocket, nothing fundamental changed since Friday— September 2025
little good news is needed to wrong-foot positioning— February 2025
The hub: thesis and position are the same object
His defining trait, near-total across the whole record
Almost every belief on this page arrived as a live trade — with an entry, a P&L, or an explicit
not-yet-triggered status — and position size is his conviction read-out: a top personal bet sized with
significant own capital at conviction 80, versus Holds only a tiny long in NASDAQ at 48. He names the
trade and the doubt in the same breath, and scores his own misses on air. That coupling is why the spine can sit
still while the trades spin: the positions rotate, the operating system does not.
Put a gun to my head, the Fed goes 50bp this week— September 2024
could flip in two weeks as rate-of-change is what matters— 2024
2 · His highest-conviction causal chains
Chains he states as mechanisms — not co-occurrences we inferred. Each step carries his
own words. Steno is unusual among our tracked voices: he attaches numbers to his transmission channels —
coefficients and lags — and those numbers survived adversarial audit against the record.
The plumbing chain: issuance → liquidity → crypto and NASDAQ first
His flagship chain: the spine's liquidity belief wired to a fixed sensitivity ordering.
Government money keeps the floor under the economy.
very hard to get a deep recession while the government keeps pumping money out— July 2024
Liquidity hits assets in a fixed order — crypto, then tech, then the consumer.
it is the most liquidity-sensitive asset class— January 2024
Crypto is even more liquidity-sensitive than tech— 2024
So the turn announces itself at the sensitive end of the ladder.
when they turn you'll see it in crypto and NASDAQ— November 2024
He thinks in numbers: the freight-to-CPI transmission
A coefficient, a lag, and a tradeable window — stated, not implied.
The coefficient.
Every doubling of freight rates adds ~0.7% to headline inflation— January 2024
The lag.
freight-rate spikes pass through to CPI over 4-8 months— January 2024
higher freight rates feed US/Europe goods inflation with roughly a one-quarter lag— January 2024
He thinks in numbers: the dollar leads, everything follows
The dollar leads Bitcoin by about a quarter.
Dollar leads Bitcoin by about a quarter— March 2025
a weaker dollar eases overall financial conditions which fuels Bitcoin— April 2025
And the dollar leads global growth by six to nine months.
global growth typically rebounds 6-9 months later via easier global financial conditions— April 2025
cheapens global dollar liquidity and lets China/Germany ease fiscally— March 2025
The audit, disclosed in both directions
Every coefficient above was adversarially verified as exactly what he said, when he said it. But honesty cuts both
ways: in the three out-of-sample months, he kept running lag arithmetic in this same idiom — with different
numbers. The stored coefficients scored 0-for-3 verbatim, 3-for-3 in idiom. The lag machine is authentically his;
treat the specific constants as snapshots, not laws.
The spine trading live: the Hormuz spike was a sell
The shale invariant applied in real time at the record's edge, spring 2026.
The choke point was already priced around.
Hormuz flows already bypassed in 2/3 of cases— April 2026
And scarcity is a stage, not a state.
scarcity turns to glut, like eggs did from record highs to record lows— April 2026
Meanwhile the shock sorts winners by energy sensitivity.
very low energy sensitivity and is a net exporter of everything— on the US, April 2026
this crisis is the best thing for US equities on a relative basis— April 2026
3 · What would change his mind — his own tests
Steno pre-commits falsifiers with dates and levels attached — and one of them has
already fired, which is why we can show you what his gates do under load.
If — and this gate FIRED
Hormuz flows don't resume before April 1, 2026
His stated falsifier on the 2026 cyclical-upswing thesis — a date, named in advance. April 1 arrived without
resumption, and by his own rule the upswing call died: a falsifier honored on schedule. (In the test window he
then re-armed the thesis behind a fresh gate — flows reopening by early Q3. The gate worked; the man reloaded.)
the business cycle is over for this time— the pre-committed consequence, March 2026
As long as
US core inflation stays below 4%
Risk assets are a buy. Above 4% is, in his words, a completely different environment — the stock-bond
correlation flips against you.
as long as US core inflation stays below 4%— 2026
With inflation below 3% the stock-bond correlation turns negative again— September 2024
Unless
Beijing goes open-ended — direct QE or household transfers
Every China rally is a squeeze to be faded, not a trend to be joined. This is the exit test on his own
structural China bear.
monetary adds are mopped up net-neutral, and it's not real QE— October 2024
one of the biggest positioning squeezes in history, not a new fundamental bull story— November 2024
If
The 10-year approaches 4%
Risk assets wobble at the hurdle — and he expects the Fed to blink first, per the October-2023 playbook.
if the 10-year approaches ~400bps, risk assets won't hold up— November 2023
the Fed signals it will cut rates more to control them— 2023
If
Trump moves on the Fed itself
Direct oversight or a loyalist chair turns gold and Bitcoin into the ultimate debasement trade — with a stated
magnitude on the dollar.
if Trump moves to increase direct oversight of the Fed— April 2025
the dollar could weaken 15-25%— 2025
If
Manufacturing capex doesn't show up soon
The metals and construction-cycle thesis goes back on the operating table — his own stated review clause.
must revisit the metals/construction-cycle thesis if manufacturing capex doesn't pick up soon— 2025
Worth knowing — the AI gate
His AI-capex bull case carries exactly one bear condition, stated himself: Oracle-style negative free cash flow is
not an issue unless the AI investments fail to return (February 2026). No returns, no thesis — otherwise it is
normal growth-stock capex.
4 · What he recently changed his mind about
We log mind-changes as a feature, not an embarrassment. Steno's pattern: tactical views
flip in days on a positioning fact or a nowcast turn — and once, a retreat manufactured a brand-new piece of his
worldview.
January 14, 2024 — the retreat that built a belief
Through late 2023 he carried a hard recession call — services rolling over,
the corporate profit cycle failing to deliver the forecast hockey-stick earnings recovery, prompting layoffs —
and defended it under pressure: may be off by a couple months but not standing down on the call (November 2023).
Then he stood down — via a stated mechanism, not a shrug: hard to get big layoff waves with the deficit running ~7%.
The recession became recession less likely, pushed out a couple quarters — and the mechanism that killed the
call became a permanent belief: deficit-prevents-recession, spine class, held for the next two years. Most
mind-changes subtract a belief. This one added one. That is the epistemic signature we built this Framework to catch.
2024 — "no cuts at all this year" became a 50bp September cut, on three named triggers
Old (February 2024, conviction 80)Fed won't cut at all this year
→
New (September 2024)An insurance cut, then a done deal after Powell's Senate testimony — ultimately 50bp
His stated triggers, in sequence: the soft May inflation report —
starting to doubt his prior no-cuts-in-2024 view — then Powell's testimony, then
the weak July labor report (100K jobs) is a major green light.
Three dated dominoes, each on the record.
February 16, 2026 — the dollar flip (the sixth-plus, and counting)
Old (2025 → early 2026)Bearish dollar all year — EUR/USD is heading toward 120
→
NewNear-term dollar stability — having flipped now-casting more dollar-positive as the US picks up speed versus Europe
Stated trigger: his own nowcast turned. This is the tactical layer working as designed — the
dollar view flipped six times across one three-month stretch of the record while the spine beneath it never moved.
Speed here is not inconsistency; it is the operating system.
December 15, 2025 — conceded the Bitcoin four-year cycle
OldCrypto is priced off liquidity and macro; the four-year cycle downplayed
→
Newthe four-year-cycle October-peak thesis has so far been proven right — and breaking it now needs an exogenous shock
Honest detail: this concession surfaced fully formed after a six-week gap in our record. We
date it at re-entry and say so — never backdated into a gap we cannot see.
He grades himself on air
The ledger keeps his own scoring: the September 2024 short-gold bet (conviction 78) ended in wrong-footed us;
an oil call logged as it's gone wrong so far; a rates call recorded as his 'next week' call did not hit.
A voice that scores its own misses is exactly the kind this record is built to hold onto.
5 · Where he is silent
Two kinds of silence, labeled: mechanisms he has argued against on the record, and
mechanisms that simply never appear in 1,041 claims. They are his sharpest differentiators from consensus macro.
Actively rejected — he has said no, on the record
De-dollarization and the dollar collapse.petrodollar ended is nonsense; The dollar isn't dead;
reserve erosion takes many decades, only on the margin. His dollar bears are cyclical calls, never
regime-collapse calls.
QT drains liquidity and sinks markets.QT is largely irrelevant — the whole balance-sheet-headline
channel is rejected wholesale; liquidity runs through issuance and plumbing.
Yield-curve inversion predicts recession.Inverted yield curve isn't reliably restrictive now —
recession, when he argued it, ran through labor, services and credit, never curve shape.
Deficits summon bond vigilantes. Issuance is plumbing, not a vigilante trigger; deficits in his model
prop growth. The single yields-channel exception he owns is European defense issuance.
AI capex is a bubble about to crash markets. AI is absent as a macro driver until December 2025 — then
arrives bullish (buy the receivers of the capex), with the one gated bear case above.
Never asserted — 1,041 claims, zero mechanism
Valuation multiples → returns. Equities are traded on liquidity, cycle and positioning; multiples
never appear as a US timing tool.
Wage-price spiral → entrenched inflation. Reframed away: US stickiness is shelter and fiscal flow,
not wages.
Real yields → gold. The textbook gold mechanism never appears; his gold runs on central-bank
rotation, flows and debasement.
Crypto leverage → systemic contagion. Crypto crashes are weekend liquidity air pockets and
buy-the-dip events throughout the record.
Taiwan invasion → semiconductor shock. Never asserted despite dense semis coverage — his chip-supply
risk runs through Hormuz logistics and Korean export data instead.
Why this matters: ask this Framework about a valuation scare or a curve inversion and the
honest answer is "this model is silent here." Refusal with receipts beats confident invention.