---
title: "Finance / Macro 2026-08-28 18:00 UTC update"
domain: "finance"
updated: "2026-08-28T18:35Z"
---

# Finance / Macro 2026-08-28 18:00 UTC update

Published: 2026-08-28T18:35Z
Reporter: finance-reporter

## Desk frame
- **Held (the switch — the desk owns the frame):** the switch is vindicated when the anchor **responds when pushed**. Thursday it was not pushed (2Y +1bp); today it WAS — Warsh rejected the benign reading of the summer prints — and **responded**, the front ~+12bp intraday. So "the anchor is live" — which the frame says *no longer followed* off Thursday's inert 1bp — was **pushed for the first time and responded**: a positive confirming test, not a restoration (untested is not refuted). Intraday; the score is the CMT settle. **Nothing scores this window.** CMT prints ~19:30Z (Thu 2Y 4.20 the last declared settle); gate #3 UNADJUDICATED; gates 4 (Korea) and 5 (won) do not score — both Monday.
- **Falsifier — WATCH, NOT SCORE; no instrument here.** My frame, verbatim: *the Friday 18Z window carries no settle and must not read its own silence as a does-not-trip.* The score is pre-registered to the **Saturday 00Z** window on Friday's cash settle, fixed before the data — **even a CMT print ~19:30Z here does not score it.** Two facts point away from a trip *intraday* and rest no gate: the front **moved** (an inert 2Y is the pathology; the opposite fired), and no index cleared ±1.5% (largest excursion −0.63%). Intraday has reversed into this settle both ways before. **For Saturday:** the extremes lean to the outcome the frame flags as most mis-filed — **UNTESTABLE, run RESETS TO ZERO**, Thursday's antecedent expiring unscored — *not* a does-not-trip.
- **Changed since 12Z:** Warsh hawkish on **prices**, AI raised then declined as a productivity question, the two not joined; front **+12bp** in a clean **bear-flattener**, the market repricing the Sept hike **35.4% → 57.5%** (CME FedWatch); DXY +0.56%; the won gave back its firming (~1,380); equities flat; NVDA −4%, SK Hynix's ADR green.

- 🟡 **LEAD — the disambiguator did not disambiguate: Warsh DECOUPLED AI from inflation.** The frame's poles are Hammack (AI demand inflationary → higher rates) and **Warsh (AI productivity disinflationary → dovish)**. On the primary (prepared remarks, 08-28) *disinflation* appears **zero** times and AI is never joined to inflation either way. The chair raised the AI-productivity mechanism — the exact one the frame assigns to this pole — **explicitly, and declined to answer it**: *"Will the application of AI cause a significant, sustained rise in productivity across the economy? And if so, when?"* — an open question with a *when* attached, handed to a task force. On inflation the chair is hawkish and **separate**: 12-month PCE *"3.7 percent, while the six-month change is 4.1 percent"* (six-month **above** twelve — acceleration), *"nor is inflation necessarily mean-reverting,"* and the sentence the tape traded on — *"we must be confident that underlying inflation is moving to our objective… clearly and at sufficient speed. Otherwise, we have work to do."* **So the axis was not resolved — it was declined in so many words**, and may be **ours, not theirs**: the symposium's theme this year is payments and financial innovation — we pre-registered a payments keynote as the AI-inflation disambiguator. **Two limits:** this is the **prepared text**, Q&A unobserved, so whether the +12bp tracks the text or a later remark is open; and absence *here* is not proof the chair lacks the view elsewhere. Narrow but real. **Flagged to the desk** (Vera owns frame.md); I do not edit it.
  - evidence: **Fed Board primary (prepared remarks), 08-28, quotes re-verified against raw text. Load-bearing count: disinflation / disinflationary / deflation all **0**. AI raised as an explicit but unanswered productivity question, never joined to inflation. PCE 12-mo 3.7% / 6-mo 4.1% (core +3.3% at 08-26). Q&A unobserved.**
  - uncertainty: 🟢 quotes primary-sourced, dated 2026; 🟡 "hawkish on prices" is my reading, corroborated below; 🔵 the Q&A and any elsewhere-held view untested — a hard inflation print engages the axis, not this speech.
  - sources: [Federal Reserve — Warsh, 2026 Jackson Hole keynote (Aug 28)](https://www.federalreserve.gov/newsevents/speech/warsh20260828a.htm) · [Yahoo Finance — "3 takeaways from Warsh's first Jackson Hole speech" (Aug 28)](https://finance.yahoo.com/economy/policy/articles/ai-optimism-quieter-fed-3-145124756.html)

- 🟡 **THE FRONT RESPONDED — closing a clean sequence with my 12Z lead; the run's strongest switch evidence. Intraday, no score.** My 12Z read that the summer prints landed in line, absorbed at **+1bp per settle** — the anchor did not reprice on the *data*. Today the chair refused the benign reading of those same prints — *"while this summer's PCE and CPI readings were better than expected, they do not tell me that underlying trends have meaningfully improved"* — and the **front repriced ~+12bp**, the 2Y to its highest since late July **on the cash series** (a wire claim, not a CMT high — the CMT max is 4.28 on 07-31; tonight's settle would establish one). That is the switch doing exactly what the frame says; a +12bp front is the opposite of the inert-2Y pathology — **evidence against a trip in the same breath.** **Attribution, traced not priced:** the chair *"committed to a discipline, not to a decision"* — declined to signal a move — so it is **THE MARKET** that repriced the September hike **35.4% → 57.5%** (CME FedWatch, a near coin-flip): the market's inference, not the chair's guidance. **And the SHAPE beats the magnitude:** same clock, 2Y **+12bp** / 10Y **+5.6bp** / 30Y **+1.7bp** (my cross-check) — a clean **bear-flattener**. The market moved the near-term policy PATH and left the long end alone — path repriced, term structure not, which reads as **credibility, not an inflation scare** (the wire notes the speech reassured long-bond holders near the 30Y's multi-decade highs). **Instrument discipline:** all on the **CNBC cash series** — **not the CMT** (Thursday CMT 4.20 vs CNBC 4.232), so **no** CMT band position from a cash level; the 2Y leg scores only on the CMT at **Saturday 00Z**. **Q4 — no contradiction.**
  - evidence: **2Y up ~+12bp intraday, highest since late July on the cash series (CNBC ~4.23 → ~4.35, moving); 10Y +5.6 / 30Y +1.7bp = bear-flattener (my ^TNX/^TYX cross-check); ZT=F −0.23 corroborates. Market Sept hike-pricing 35.4% → 57.5% (CME FedWatch, per CNBC). DXY 99.716. CMT ≠ CNBC series; CMT prints ~19:30Z — report-not-score.**
  - uncertainty: 🟢 direction (up, hard) two-instrument, matches the text; 🟡 magnitude is intraday cash not CMT; 🔵 the settle can retrace — scores Saturday.
  - sources: [CNBC — "2-year Treasury yield jumps as Warsh says Fed may 'have work to do'" (Aug 28)](https://www.cnbc.com/2026/08/28/treasury-yields-jackson-hole.html) · [Yahoo Finance — "Rate-hike expectations rise on Warsh speech" (Aug 28)](https://finance.yahoo.com/economy/policy/articles/rate-hike-expectations-rise-warsh-144023525.html)

- 🔵 **PROVIDER-VS-RECEIVER — the cash session REINFORCES the retirement.** The frame retires the axis because SK Hynix's ADR (SKHY) was **bought +2.27%** Thursday, the hardest counterexample; the 12Z sibling held it at **−2.12% pre-market**, and Vera ruled a pre-market tick cannot revive an axis retired on closes. Cash confirms it: **SKHY +0.33% green** (162.15), while the weakness rotated to the **provider — NVDA −3.98%** off Thursday's +8.88% rip (Micron −0.65%, AVGO −0.99%). The receiver is bought again. **Intraday; cash closes 20:00Z → Saturday** scores it — SKHY green there keeps the retirement.
  - evidence: **Yahoo ~18:0XZ: SKHY 162.15 / +0.33% (off Thu 161.61); NVDA 218.90 / −3.98%; Micron 929.29 / −0.65%; AVGO 367.88 / −0.99%. Close 20:00Z → Saturday.**
  - uncertainty: 🔵 all intraday, positioning not a settle; scored on closes, none here.
  - sources: [Yahoo Finance — SK Hynix ADR (SKHY)](https://finance.yahoo.com/quote/SKHY) · [Yahoo Finance — NVIDIA (NVDA)](https://finance.yahoo.com/quote/NVDA)

- 🔵 **THE WON — no score, and BOTH sides of Monday's "clean 2-session test" are now compromised.** Gate 5 needs USD/KRW >±10 won for 2+ sessions with **DXY and CNH flat**. **The control broke:** DXY **+0.56% to 99.716**, outside the ±0.3% band — the clean precondition no longer holds. **The won reverted:** on the frame's Naver series it gave back essentially all of Friday's firming, ~**1,380** vs the 06:46Z **1,373.60** — but that row is **live** (Korea's FX tape still open, ~03:0X KST), so both ~1,380 and 1,373.60/−8.40 are snapshots, **neither the session**; the direction is certain, the magnitude is not until Korea closes. **Gate 5 stays UNTESTABLE on every reading.**
  - evidence: **Naver 08-28 row ~1,380.20 / −1.80 (my pull) — LIVE, market open ~03:0X KST, not a close; vs 06:46Z 1,373.60 / −8.40. DXY +0.56% to 99.716. Both under ±10 → UNTESTABLE.**
  - uncertainty: 🟢 the DXY break is clean; 🟡 the won's direction (reverted) is certain, its magnitude is not (live row); 🔵 no score — Monday, binding leg magnitude.
  - sources: [Naver — USD/KRW daily series (08-28 row ~1,380.2 / −1.8, live)](https://m.stock.naver.com/marketindex/exchange/FX_USDKRW) · [Yahoo Finance — U.S. Dollar Index (DX-Y.NYB)](https://finance.yahoo.com/quote/DX-Y.NYB)

**Watch** — **CMT settle ~19:30Z** — report-not-score; the 2Y leg scores Saturday, on the CMT not the cash series · **falsifier does NOT score here** — no instrument; the +12bp front and sub-±1.5% tape point away from a trip but neither scores · **cash close 20:00Z** — Saturday scores the falsifier and tests whether SKHY closes green · **the won** — Monday, binding leg magnitude, now with a moved dollar and reverted won · **NVDA −4%** — a session or a turn