DAILY REPORT

Daily AI Market Report — 2026-08-31

Regime: Risk-On (score 33) · Published Mon, 31 Aug 2026 10:23:45 GMT · Decisions sealed before the U.S. open

Market overview

New Fed Chair Kevin Warsh used Jackson Hole to remind the market who sets the cost of money. His hawkish lean pushed the long end higher and squeezed the very technology and semiconductor names that have carried the tape all year. The S&P closed only marginally lower at -0.23%, but under the surface the rotation was sharper than the index let on — long yields up, semis leading the fade, and a stubborn AI demand narrative from NVDA refusing to let tech break down.

Every one of our strategies read the day as broadly constructive — the regime score sat at +33, mildly risk_on, with VIX at 14.4 and a normal +0.39pt 2s10s curve. But they disagreed sharply on how much of that constructiveness to actually own. Maverick took the tape at face value and ran deployment to 90.8%; Sentinel treated the same tape as a warning and sat at 61.3% with gold as its single largest position. That gap — nearly 30 points of invested capital on identical data — is the story of the session.

The dividing line was how each strategy priced the Warsh risk against the NVDA reward. Meridian and Maverick leaned into the durable AI demand thesis, holding twin 10%+ Technology anchors in MSFT and NVDA and pairing them with Energy via MPC. Sentinel accepted the same names but shrank them and let a 12.6% GLD weight lead the book. Regent, by contrast, refused to concentrate at all — its largest single position was just 4.3%, spreading conviction across Technology, Staples, Health Care and Industrials.

Today's lesson is that a mildly risk_on regime is not a green light — it is a question. The quant score says lean in; the new Fed Chair says be careful what you lean into. Our strategies answered that question with wildly different position sizes on the same set of tickers, and that dispersion in sizing — not the direction of any one name — is where the real information sits.

Insights

• DEPLOYMENT_GAP — On an identical +33 risk_on read, Maverick deployed 90.8% while Sentinel held at 61.3% — a near 30-point spread. Maverick reads Warsh's hawkishness as noise against a durable AI demand cycle; Sentinel reads rising long yields as the exact condition to keep dry powder and a gold buffer.
• SEMIS — Semiconductors led the day's decline as long yields rose, yet Meridian and Maverick both kept NVDA at a 10%+ up-weighted anchor. The interpretation split is explicit: the tape says semis are vulnerable to higher rates, while the AI demand narrative says the growth runway hasn't peaked — and the aggressive book chose to believe the runway.
• GOLD — Sentinel let GLD rise to 12.6%, its largest single holding and above its own Technology sleeve. With VIX still low at 14.4, that is a deliberate choice to hedge a policy risk that volatility hasn't yet priced — a buffer built before the fear, not after it.
• BREADTH — Regent capped its top position at just 4.3% and spread across five sectors, while Maverick ran three names above 12%. Same constructive backdrop, opposite expressions of conviction — one strategy diversifies away single-name policy risk, the other concentrates into the strongest secular story.

Key directions

• The regime read mildly risk_on at +33, with VIX at 14.4 and a normal +0.39pt 2s10s curve — but the S&P still closed -0.23% as new Fed Chair Warsh's hawkish Jackson Hole tone lifted long yields and pressured tech.
• Semiconductors led the day's decline on the rate move, yet the durable NVDA AI demand narrative kept the sector from breaking down — leaving the tape genuinely two-sided rather than cleanly risk-off.
• The sharpest signal was dispersion in sizing: our strategies ran invested capital from 61.3% to 90.8% on identical data, meaning the disagreement is about conviction, not direction.
• Gold emerged as the swing buffer — where aggressive books held it near 4.5%, the defensive book let it lead at 12.6%, a hedge built while VIX was still low rather than after a spike.
• The single-stock event risk was real — PayPal's failed M&A drop is a reminder that idiosyncratic shocks persist even in a constructive tape, reinforcing the case for position sizing over hero bets.

Regime read

The engine placed today at +33 on the -100/+100 scale — mildly risk_on, but far from a clean all-clear. The favorable inputs were concrete: VIX printed a calm 14.4 (contributing +17.7), the 10Y-2Y curve stayed normal and positive at +0.39pt (+11.7), Fear & Greed sat at 54 in mild greed territory (+2.2), and the index held +8.75% above its 200-day trend. Asia was quietly supportive overnight, with Shanghai +0.9% and KOSPI +0.5% offsetting fractional dips in the Nikkei and Hang Seng.

The offset came from policy. New Fed Chair Kevin Warsh's hawkish Jackson Hole tone lifted the long end of the curve and pressed the very technology and semiconductor names that dominate index weight. The S&P closed -0.23% (a -0.7 contribution), and while no high-volume selloff was detected — volume ran a benign 0.90x average — the internal rotation was sharper than the headline. That is why the quant score's constructive lean carried an asterisk this session.

Meridian applied a slight qualitative haircut on top of the raw score, landing on a neutral posture rather than pushing the risk_on read to its full extent. The reasoning is that the raw factors capture volatility and breadth well, but they do not yet price a fresh, credibly hawkish Fed Chair against a durable NVDA AI demand narrative that keeps semis from cleanly breaking. The overlay cap was held at the pre-registered 10, subject to reassessment after forward data.

The takeaway for the regime frame: a +33 score in a rate-sensitive, event-heavy tape is best read as permission to participate, not a mandate to press. The wide deployment spread across our strategies — from 61.3% to 90.8% — is the honest expression of that ambiguity, and it is why sizing discipline mattered more than direction today.

The four AI personas

Meridian · Core

Meridian settled on a neutral-with-a-haircut stance, holding invested capital at 77.8% and anchoring the book with twin 10% up-weighted positions in MSFT and NVDA. It paired that Technology core with a 17% Energy sleeve led by MPC (10%) and PSX (7.2%), then carried a 6.6% GLD position as a modest neutral buffer. The read: honor the durable AI demand thesis, but trim overall risk given Warsh's hawkish tone and rising long yields rather than press it.

Maverick · Aggressive

Maverick took the tape at face value and ran deployment to 90.8%, the highest of any strategy. It pressed Technology to 24% of book with MSFT and NVDA both at 12.2% up-weighted, and stacked MPC at another 12.2% in Energy — full conviction into the AI demand narrative with only a thin 4.5% GLD hedge. Warsh's hawkishness was treated as noise against a growth runway it judges hasn't peaked, so the capital came from cash rather than from trimming the winners.

Sentinel · Conservative

Sentinel read the same session as a warning and held invested capital at just 61.3%, keeping the widest cash cushion in the group. It let GLD rise to 12.6% — its single largest position, sized above its entire Technology sleeve — while shrinking MSFT and NVDA to 6.9% each. The stance is deliberate: accept the AI names but at half-conviction, and fund a policy hedge against rising long yields before volatility prices it, not after.

Regent · Sector Champion

Regent refused to concentrate, capping its largest single position at just 4.3% (MSFT) and spreading 82.8% invested capital across five sectors — Technology, Energy, Financials, Health Care and Industrials each near 7-10%. Rather than pick a side in the Warsh-versus-NVDA debate, it diversified the policy risk away entirely, adding neutral Staples (PG 3.8%) and Health Care (JNJ 3.4%) defensives alongside its tech and communication holdings. The logic: in a regime that is constructive but rate-sensitive, breadth beats a single concentrated bet.

← 2026-08-27 · All reports · Track record

Virtual investment simulation — informational and entertainment purposes only, not investment advice. All decisions were sealed and timestamped before the U.S. market open; percentages and derived scores only, no price data is republished.