The Fed has entered the gold debate. On 3 September, Federal Reserve economist Colin Weiss published a FEDS Note titled "Why Gold Didn't Actually Overtake Treasury Securities as the World's 'Favorite' Reserve Asset" (https://www.federalreserve.gov/econres/notes/feds-notes/) — the first Fed-authored counter-frame to the market narrative that gold has displaced Treasuries as the second-largest reserve asset, and the single most consequential dated input in this fourth dollar-system scan. Alongside a separate finding that the stablecoin aggregate market cap has contracted from US$317 billion in April to roughly US$290 billion today, this cycle leaves the thesis structurally intact but with one conduit's sizing revised downward.
The Weiss note matters on two dimensions. On substance, the dollar-system thesis rests on architectural durability, and the gold-substitution channel is where that thesis faces its sharpest attack; a Fed-authored rebuttal signals that the Board's research programme has moved from silent tolerance to public contest. On timing, six months from the January $5,405 gold peak to a formal institutional counter-frame is a fast response — Fed research does not publish counter-narratives unless a narrative has reached policy-relevant salience, and the Weiss note is itself evidence of that salience. The full text returned a 404 within this scan's window, so substantive detail must await a Q4 refetch; the existence and title alone are nevertheless load-bearing for the durability leg. Gold remains on the equity slate — GLD as the sovereign-CB substitution expression — but its case as a dollar-displacement vehicle is directly disciplined by this counter-position.
The second material development is less welcome. The stablecoin captive-bid conduit — the second of the four-conduit non-informed marginal Treasury-bid architecture established in prior scans — is contracting, not compounding. CoinGecko data (https://www.coingecko.com/en/categories/stablecoins) shows aggregate stablecoin market cap at US$290.2 billion as of 16 September, with USDT at US$183.3 billion and USDC at US$73.7 billion. That is down from US$317 billion on 6 April — a five-month decline of roughly US$27 billion (-8.5%) — and it spans the 15 June live date of the GENIUS Act, the regulation whose high-quality-reserve mandate was supposed to catalyse accelerated issuance. The mandate is live; the demand-side growth has stalled. The July scan modelled a trajectory above US$400 billion by Q4 2026; the derailing indicator — aggregate stagnates below US$350 billion by year-end — is now on course to trigger.
This does not break the mechanism. The GENIUS Act's reserve mandate still operates on a standing stock of roughly US$290 billion, implying plausible Treasury exposure of US$220–250 billion. But the incremental bid per quarter that anchored the compounding case is not materialising. Circle Internet Group's public framing — a July post on agentic payments and an August post on institutional infrastructure in South Korea (https://www.circle.com/blog) — suggests a pivot toward institutional use-cases rather than raw market-cap growth, consistent with the slower aggregate trajectory. Circle's 10-Q filed 5 August (SEC accession 0001876042-26-000248, https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&CIK=0001876042&type=10-Q) is queued for a full-text Q4 fetch on reserve composition. CRCL — the listed pure-play captive-bid expression of the GENIUS channel — stays on the equity slate, but with tempered conviction.
The durability case receives reinforcement from a different angle. A 4 September FEDS Note by Payne and Styczynski, "New Forms of Money and the U.S. Monetary Aggregates" (https://www.federalreserve.gov/econres/notes/feds-notes/), extends the Fed's institutional treatment of stablecoins inside the monetary-aggregates framework. An August 26 note by Falcettoni, "A Decade of U.S. Cross-Border Payments Efforts" (https://www.federalreserve.gov/econres/notes/feds-notes/), frames US cross-border-payments infrastructure as a permanent multi-decade institutional project rather than a reactive response to CIPS or mBridge expansion. And a same-day note by Anbil, Anderson, Cordes and Ruprecht, "Repo Markets and the Fed's Balance Sheet" (https://www.federalreserve.gov/econres/notes/feds-notes/), bears directly on the fourth conduit — the hedge-fund basis trade that the June FEDS Note documented as doubled in size between 2023 and 2025. Counting from March through early September, there are now twelve dated 2026 Fed-research artefacts touching dollar international roles, stablecoins, cross-border payments or Treasury-market microstructure — a density qualitatively distinct from the pre-2025 baseline and further institutionalising the durability case.
The Falcettoni retrospective surfaces a gap in the prior equity slate. The custody-and-settlement banks — dealer-independent, Fed-supervised backbone of dollar cross-border plumbing — were missing. I am adding Bank of New York Mellon (BK): the largest US custody bank, the largest global custodian of Treasury securities, and a disclosed custodian for portions of the Circle Reserve Fund. BK sits at the intersection of Treasury custody, offshore-dollar clearing and growing stablecoin-reserve custody flows — picks-and-shovels on the durability-plumbing thesis, distinct from JPMorgan's dealer and FICC franchise. The August repo-markets note also reinforces JPM's position: its primary-dealer balance sheet is the conduit for both the basis-trade tier and, via the AI-capex private-credit fusion mechanism, the hyperscaler debt-financing pipeline. A repo-stress event that unwinds one simultaneously stresses the other — a structural linkage the next private-credit and AI-capex-ceiling scans must model explicitly.
The broader equity slate — Visa (V), Mastercard (MA), ICE, CME and CRCL as the plumbing picks-and-shovels layer; BlackRock (BLK) and its iShares Treasury franchises SGOV and IEF as the passive-flow beneficiary of domestic marginal-buyer growth — is unchanged in composition. TLT remains the structural avoid, a status reinforced by the 15 September sovereign-fiscal scan's DFII30 print at 3.07%.
What would change my mind
The cleanest falsifications are: an IMF COFER release showing USD share below 55% in a single quarter; RMB COFER crossing 4% in two consecutive quarters; a SOFR-IORB spread above 25 basis points sustained for 60 or more days, signalling a basis-trade unwind; or the stablecoin aggregate contracting below US$250 billion by Q4 2026, which would indicate a structural problem with the second conduit rather than a growth-rate revision. Any of these would require a material revision to the durability thesis.
What I'm watching next
Four dockets. First, IMF COFER Q2 or Q3 2026 — this is the fifth consecutive scan without a fresh release; the absence has become a scan-hygiene problem in its own right. Second, the Weiss note full-text — the 3 September URL returned a 404 within the scan window and is the single highest-priority content item for Q4. Third, the Treasury implementing rule on GENIUS reserve composition — the operative Q4 2026 regulatory docket that will determine how much of the standing stablecoin stock qualifies as high-quality-reserve Treasury demand. Fourth, WGC Q3 2026 Gold Demand Trends, expected late October — the next read on whether the roughly 96 tonnes per month CB gold-buying pace from Q2 2026 is sustained.