Our forecasting track record

What changed, in plain words

• Oil: Brent rose a fourth straight session to about USD 84.67 after Iran said the Strait of Hormuz stays shut until Washington lifts sanctions and pays war reparations. Traffic is still collapsed — 8 to 15 ships crossed on each of 4, 5 and 6 August against roughly 130 before the war. Over the weekend an Abu Dhabi state tanker was attacked and Houthi fighters claimed a strike on Saudi Arabia's Jazan refinery. — The world's most important oil route is still shut, and the reopening deal just slipped away.
• Tariffs: a 50% US tariff on certain Canadian dairy, vehicle and alcohol imports takes effect on 19 August, under a 1930 law no president had ever used. It covers about USD 20 billion of imports, and North American free-trade status does not exempt goods. — A dated price rise on a narrow set of everyday goods, under a law with no legal precedent.
• Fed: the odds of a US interest-rate *rise* in September fell to about 44% from 57% before Friday's jobs report. Three officials voted to raise at the July meeting, and inflation still ran at 3.7% a year in June. — Washington is still arguing about raising rates, not cutting them.
• Chips: memory-chip shares are more than 20% below their recent highs, but the pricing has not cracked — TrendForce still expects standard memory contract prices to rise 13–18% this quarter. — Share prices are pricing a memory glut the actual chip prices are not showing.
• Long-term rates: the US government's 30-year borrowing cost sat at 5.22% on 6 August, near its highest in decades, while every measure of market fear stayed quiet. — The bond market and the share market are pricing different futures; which one moves is not obvious.
• Gold: gold held above USD 4,300, at about USD 4,346, after rising more than 7% last week — a two-month high, and up 30% on the year. It is *not* a record: gold peaked at USD 5,608 in January 2026, so it sits about 22% below that. — Gold rose because a rate rise looks less likely, which lowers the cost of holding it — the widely repeated 'record gold' framing is simply wrong.
• Shares: the S&P 500 closed at a record 7,757.64 on Friday and US futures were roughly flat this morning. Asia rose — Japan's Nikkei 2.55%, Korea 0.7%, Hong Kong 0.7% — and Germany's DAX set a fresh high, while the UK's FTSE slipped 0.34%. — Share markets worldwide are treating a weak US jobs report as good news, because it makes higher interest rates less likely.
• The week ahead: US inflation figures for July arrive on Wednesday 12 August at 8:30am New York time, with producer prices on Thursday. — These two numbers, not the oil headlines, will decide the September rate argument.
• On the radar (monitored, never traded): Caterpillar, the construction-and-power-equipment proxy, was down 1.72% pre-market at USD 842.19 while Nvidia was up 2.27% at USD 223.96 — the industrial and AI legs of the same buildout moving in opposite directions. Copper, the other buildout sensor, is up 48.8% over a year. — The machinery-and-metals side of the AI construction story is not confirming what the chip side is doing.

Headline

Hormuz re-prices from 'deal imminent' to 'deal conditional.' Brent's October contract stood at USD 84.43 at 04:30 GMT and the board read 84.67 (+1.34%) pre-open, a fourth consecutive higher session and roughly 16% above the pre-war level. The driver was explicit conditionality: Iranian Foreign Minister Abbas Araghchi said on Sunday that Iran and Oman were close to an agreement on transit routes but that Hormuz will not reopen until Washington eases sanctions and pays war reparations, and that Tehran is not in direct talks with the US — against Washington's earlier assertion that a deal was near. President Trump said the US was now "low-keying it." The physical picture has not improved: MarineTraffic showed 8–15 crossings on each of 4, 5 and 6 August versus roughly 130 pre-conflict, and the IMO counts at least 64 violent incidents and 17 deaths involving commercial vessels since late February.

The risk surface widened beyond the strait. The UAE condemned an Iranian missile attack on an ADNOC-owned vessel over the weekend, and Houthi forces claimed an attack on Saudi Arabia's Jazan refinery. Note what that does to the shape of the risk: the disruption is now partly about refining and loading capacity outside the chokepoint, not only about transit through it. EU gas rose 5.03% to 58.34 on the same tape.

A trade-policy axis that this sweep had not been reading. Three proclamations dated 20 July impose 50% Section 338 duties on 52 dairy, 439 motor-vehicle and 63 alcohol HTSUS subheadings, effective 12:01am ET on 19 August. This is the first invocation of Section 338 of the Tariff Act of 1930 by any president; the provision has no implementing regulations and no judicial precedent, and unlike prior actions it overrides USMCA preferential origin and stacks on Sections 301 and 122. Energy, potash, fish and critical minerals are carved out — which is why this is a goods-price event and not an energy-chain event. Scale is modest in aggregate (~USD 20bn, ~5% of US imports from Canada) but concentrated: USD 19.3bn of the 2024 base sits in the motor-vehicle list.

The rates/risk divergence is the week's cleanest unresolved question. DGS30 closed 5.22% on 6 August after a 5.27% period high on 31 July, near multi-decade highs, while 2s10s sat at +0.46pp, high-yield spreads at 2.71pp, VIX at 15.15 and sector/factor breadth at 91.7% above both the 200-day and 50-day averages. Long-duration is pricing one regime and every risk sensor another. Per the 2026-08-09 reviewer's handed candidate, that transmission question is now logged as P-0076 rather than asserted.

Rate path: hike-versus-hold, still not cut. LSEG put September tightening at 43.9% against 57% pre-report and hold at 60.4% against 43.2%; Trading Economics' board read ~44% against ~67% a week earlier; Kalshi showed ~65% hold. Cite the range and the venue, not "the market expects." Three FOMC officials dissented for a hike on 29 July with headline PCE at 3.7% y/y in June. Inflation Insights' Omair Sharif argued the downside jobs surprise may not move a Committee focused on inflation; Citi's economists took the other side, calling hikes unlikely with a cut as the base case for October. July CPI lands Wednesday.

Memory: equity fear still un-corroborated by contract pricing. Memory names sit more than 20% below recent highs, yet TrendForce's 3 July assessment has conventional DRAM contract prices up 13–18% QoQ in 3Q26 and raised PC DRAM to 15–20% from 8–13%. That is the fourth consecutive week the pricing evidence has failed to confirm the equity move, and it bears directly on the open memory-glut claims (P-0064, P-0025, P-0028, P-0054) without warranting a new one.

Readings

Fact-check log

Standing theses — re-scored

• #6 Energy and critical-commodity supply — CONFIRMING, and the dominant theme today (TRC-0037). Brent's fourth consecutive higher session, transits at roughly 6–12% of normal, an ADNOC tanker attacked and a claimed Houthi strike on Saudi Arabia's Jazan refinery all sit on the confirm side; EU gas +5.03% and copper +48.8% y/y widen it. Coverage audit: 3 of 3 dimensions covered, with macro_spillover resting on the traded book only. The peace-dividend disconfirmer is already escalated and open as P-0061 (due 2026-09-04) with P-0071 and P-0074 on the same event, so no fourth duplicate was logged. — The supply side of the energy thesis got stronger today, and the bear case is already on the record as a scored claim rather than a hedge in prose.
  For your money A supply disruption that persists shows up in fuel and freight before it shows up in anything else, which is the part of inflation a household feels most directly. It is a slow risk and not a reason to make a big move — a cash buffer for ordinary bills and holdings spread across different things do the work here.

  What to watch It matters more if incidents keep occurring away from the strait itself, or if transits stay near a tenth of normal into September; it matters less once ships actually move under a published agreement.
• #5 Sovereign-debt / rates stress — CONFIRMING on the level, UNPROVEN on transmission. DGS30 at 5.22% (period high 5.27%) is the confirm signal; the disconfirm side is that it has not transmitted — high-yield spreads 2.71pp, VIX 15.15, breadth 91.7% above the 200-day, S&P at a record. The 2026-08-09 reviewer flagged both of this theme's dimensions as book-only; that transmission question is now a scored claim, P-0076, rather than an assertion. — Long-term borrowing costs are genuinely elevated, but the claim that they will drag risk assets is exactly the untested part, so it was logged as a falsifiable prediction instead of narrated.
  For your money High long-term rates raise the cost of mortgages and business borrowing quietly and for years, which matters more to most savers than any single day's share move. The sensible response is structural, not urgent: avoid stretching on new long-dated debt, keep near-term money in cash, stay diversified.

  What to watch It matters more if long-term rates keep rising while share prices hold at records, or if a long-dated government auction is reported as weak; it matters less if long rates ease back.
• #1 AI-credit fragility — still DISCONFIRMED on the financing leg. High-yield spreads at 2.71pp with no widening flag is the fourth straight week the funding-stress signal has failed to appear, even with memory equities in a bear market. Open claims P-0011, P-0040 and P-0056 carry this; the equity dislocation has not reached corporate credit. — The feared crack in how the AI buildout is financed still has not shown up in what companies actually pay to borrow.
  For your money Cheap corporate borrowing is what keeps the buildout and the jobs attached to it going, so its quietness is genuinely good news for now rather than a warning. Because it is also the thing that would change first, this argues for knowing how much of your savings sits in technology-heavy funds — not for acting on the story.

  What to watch It matters more if the extra interest risky companies pay climbs for several consecutive weeks off today's low base, or if a named AI-linked borrower is reported to have missed a payment; it matters less while spreads sit near multi-year lows.
• #4 Compute as a strategic resource / memory oligopoly — the pricing evidence keeps disconfirming the equity fear. TrendForce's 3Q26 assessment has conventional DRAM up 13–18% QoQ and raised PC DRAM to 15–20%. Per the standing CXMT discipline, the falsifier is a named tracker reporting a month-over-month contract-price decline (P-0064, p=0.20, due 2026-09-30) — not a share-price move. No new near-duplicate memory disconfirmer logged. — The China-driven memory glut that share prices have been pricing for a month is still not visible in the prices actually being charged.
  For your money Rising memory prices reach ordinary buyers as pricier phones, laptops and cars, so this is a cost-of-living channel as much as a market one. Neither the fear nor the pricing is settled, so the useful step is understanding what you already own rather than adding to the bet.

  What to watch It matters more if a named tracker or a top-three maker reports contract prices actually falling month over month; it matters less while each revision to those prices keeps pointing upward.
• #9 Labour-market deterioration — CONFIRMING on hiring, DISCONFIRMED on firing. Friday's −23,000 payrolls with 103,000 of downward revisions is a clear confirm; the automated labour gauge still reads calm because the Sahm-rule gap is 0.067pp and the four-week claims average is 8.9% *below* the prior four weeks. A hiring freeze without a firing wave. — Companies have stopped hiring but have not started firing, which is a real deterioration and also not yet the recessionary version of it.
  For your money This shape of market hurts job-changers and new entrants most and caps the pay rises that keep up with prices, while leaving people already in stable jobs largely untouched. If your income rests on one employer, a slightly larger cash cushion is the proportionate response — not a dramatic change.

  What to watch It matters more if weekly unemployment claims start trending upward for several weeks, or if the next payrolls report is negative and revised down again; it matters less if hiring recovers with claims staying low.
• #7 Equity concentration / breadth — DISCONFIRMED today. Sector and factor breadth reads 91.7% above both the 200-day and 50-day averages with two new 52-week highs and no new lows, and Monday's strength was broad across Japan, Korea, Hong Kong, China and Germany. Records on wide participation, not on a narrowing few. Caveat: this is a roughly twelve-ETF proxy, not literal 500-stock breadth. — The market is at records with broad participation rather than a handful of giants carrying it, which is the healthier configuration.
• #2 De-dollarisation / parallel rails — no new evidence today. The dollar index at 99.66 with the yen at 158.68 and the won at 1,418.89 is currency weakness in Asia, not a settlement-rails story. Gold's rise traces to the rate-path repricing and Chinese institutional buying reported last week, not to a fresh reserve-system datapoint. Nothing on mBridge or e-CNY volumes surfaced. — No movement in the slow shift away from the US dollar today; gold's move has a simpler interest-rate explanation.
• #3 Crypto → AI/compute rotation — quiet. Bitcoin USD 65,167 (+0.51%) and Ether USD 1,924.64 (+0.88%) both drifted higher with shares, offering no rotation signal either way. — Crypto moved with share markets rather than against them, so it says nothing new about where risk appetite is heading.
• #8 AI restructures IT services — no new evidence. No Accenture or systems-integrator datapoint surfaced in today's sweep; P-0005 remains the open claim on managed-services bookings. — Nothing new on whether AI is eating the consulting business.

Research only — not financial advice. The practice portfolio is for learning only; this is a read-only risk briefing meant to help explain risk, never tell anyone what to buy or sell.
Project snapshot — changes waiting for review: 16 · tracked forecasts open: 44 (checking now: 0) · track record so far: 0.118227 · practice portfolio updated: 2026-08-07.
Past briefs & dashboards: market-sentinel-lyart.vercel.app/reports