Seiche’s strain reading is being driven by a dated reserve path, not a disorderly funding print. Reserves fell $49.3B in the latest weekly ledger, but SOFR remains below IORB and Standing Repo Facility use is zero.
A fresh weekly balance-sheet change, rather than an overnight-rate dislocation, is driving Seiche’s dollar-funding warning. Reserves fell $49.3B to $2,944B through 2026-08-12. The ledger records a $56.6B TGA rebuild and a $39.7B rise in foreign reverse-repo balances, partly offset by a $36.5B release in the residual.
That change arrives before a $95B Treasury auction settlement on 2026-08-20. Seiche’s dated reserve-path engine is saturated and contributes 11.0 points to its 45.3 STRAIN composite. But the immediate price of overnight cash has not confirmed a shortage: SOFR is below IORB, Standing Repo Facility accepted volume is $0.00B, and market stress is at the 24th percentile of its own history.
The conclusion is guarded. Dollar-funding capacity may be becoming more sensitive to scheduled drains through reserve absorption and calendar pressure. It is not yet established that overnight funding is scarce.
The observable mechanism begins with the weekly reserve ledger and a known settlement date. Through 2026-08-12, reserves declined $49.3B to $2,944B. Seiche attributes the move across six legs that sum to the reserve change to the dollar: the TGA rebuilt by $56.6B, foreign RRP rose by $39.7B, and the residual released $36.5B. The desk says $18.9B, or 0.28% of assets, remains in an unnamed residual.
Seiche then places the $95B settlement on 2026-08-20 into its reserve path. Its forecast reserve level for that date is $2,933.8B, with a worst-case path of $2,886.4B. These are Seiche scenarios, not observed reserve prints.
Their importance rests on a fitted reserve-demand curve. Seiche estimates a kink at $3,703.1B, putting reserves $759B below that estimate as of 2026-08-12. The fit has an R-squared of 0.62. That is structural evidence of reduced buffer in the model, not an official threshold and not proof that the system cannot clear overnight funding.
The testable transmission is therefore narrow: if scheduled pressure matters, secured overnight pricing should worsen relative to IORB or the official repo backstop should see meaningful use. Neither confirmation is present yet.
The latest ledger is the material observed change. Reserves were $2,944B as of 2026-08-12. The TGA was $966.6B as of 2026-08-14, while foreign RRP was $0.255B as of 2026-08-17.
A separate official-sector measure points in the same direction without establishing a funding event. Foreign officials reduced Fed custody Treasuries by $90.2B over 13 weeks, while foreign RRP parking rose $34.8B. The combined official footprint nevertheless fell $55.5B to $2,954B. That is only a partial rotation within the measured official footprint.
The board’s divergence is explicit. Plumbing is at the 63rd percentile, market pricing at the 24th percentile, leaving plumbing ahead by 39 percentile points. The composite reads 45.3, classified as STRAIN, with full input coverage. Its dominant driver is the dated reserve path, not a contemporaneous funding-price break.
The editorial pooled five-business-day event read is 6.8%. The board’s machine-learning output is 2.9% as of 2026-08-13 and says its probability levels do not beat climatology. Those readings are for ranking and alerting, not literal odds.
Readers can inspect the published daily dispatch, data overview, and series index. The composite is a Seiche derivation, not an observed market price.
The counter-case can defeat the thesis: overnight dollar cash may remain abundant even as a calendar model identifies lower reserve buffers.
SOFR was 3.62% as of 2026-08-14, versus IORB at 3.65% as of 2026-08-18. The editorial read puts SOFR 1.7 basis points below IORB as of 2026-08-12. EFFR was 3.63% as of 2026-08-14. These are not scarcity prints.
Nor is the official backstop being tested in size. Standing Repo Facility accepted volume was $0.00B as of 2026-08-17. Broad market stress was at the 24th percentile as of 2026-08-14; VIX was 14.25 and high-yield option-adjusted spread was 2.67%.
This evidence does more than temper the warning. It challenges its mechanism. A below-kink estimate and a scheduled settlement can identify vulnerability, but cannot establish a binding dealer or bank constraint. Seiche also says its historical analogs do not beat climatology on their sample, while its book remains neutral because the published event threshold is unmet. If cash continues to clear below IORB through the pressure dates, the calendar-led strain interpretation is overstated.
The immediate test is the $95B settlement on 2026-08-20. The thesis gains support only if the reserve-path pressure is independently accompanied by secured overnight pricing deteriorating relative to IORB or meaningful Standing Repo Facility use.
It fails its central near-term test if settlement passes while SOFR remains below IORB and the facility remains unused in size. That outcome would show that the dated drain did not translate into observable scarcity in the overnight market.
Seiche also marks 2026-08-31 as a month-end pressure date, with $24B of settlement and a $2,857.9B worst-case reserve path. It flags 2026-09-15, 2026-09-16, and 2026-09-30 as further calendar-pressure dates. These are scenarios, not forecasts of a funding event.
The pressure chain starts with the ledger and calendar, then requires confirmation in secured funding. The sequence is: reserve balances and dated settlement pressure; an observable response in secured overnight pricing or official backstop use; then any broader transmission into market liquidity, crowding, and exit capacity.
The first link is visible. The second has not been confirmed.
For offshore-dollar monitoring, Seiche records a USDT peg deviation of -10.6 basis points as of 2026-08-18, with a z-score of 0.88. Stablecoin supply was $306.4B and changed -0.6% over 30 days. The board labels funding-pop conditions flat water as of 2026-08-14. These are monitoring measures, not evidence that digital-asset funding is impaired.
This analysis uses Seiche’s methodology, dispatch archive, dated daily dispatch, data overview, and series index. The cited observed inputs include Federal Reserve H.4.1, Federal Reserve custody and foreign RRP data, New York Fed secured rates and operations, FRED, BEA GDP, Treasury auction calendars, VIX, credit spreads, and rates volatility.
Seiche describes its composite as deterministic synthesis across published point-in-time engines. Confidence measures agreement across independent signals, not the probability of a market outcome. Publication times are not collected uniformly, so event, desk-knowledge, and publication clocks remain separate. Modelled reserve paths, fitted kinks, percentiles, and composite scores are not tradable prices or official policy thresholds.
Research boundary: Seiche covers system dollar-funding capacity. LiquiLens covers institution and lender balance-sheet risk. Undertow covers market liquidity, crowding, and exit capacity. Related Palimpsest readings are context-only and never enter the composite or a model feature.
Research and market data, not investment advice.