Journal ·
Saturday, August 22, 2026
Regime Risk-onThe Week in Review
Market Regime
RISK-ON the regime engine's latest close-of-day read, carried into the weekend (markets closed).
The regime engine's latest read is RISK-ON, unchanged from the prior call. The inputs behind it: VIX at 14.89 (calm), breadth at 55.7% of names above their 200-EMA (543 of 975, classed healthy_marginal), and SPY closing at 762.62 against a 200-EMA of 710.9 +7.3% above trend.
That combination is the interesting part. A tape 7.3% over its own 200-EMA with only 55.7% of constituents above theirs is a narrow advance, not a broad one. The index is doing the work; the median stock is closer to the line. Breadth sitting at "healthy_marginal" rather than healthy is the honest description of it.
Rates gave nothing away. The 10Y printed 4.69% as of 2026-08-20, WoW +1bps. The 2Y printed 4.19%, WoW +2bps. The 10Y-2Y spread stood at 0.50% as of 2026-08-21, WoW -1bps. Nothing in those moves is a signal; they are inside the noise of a quiet August week.
Two lines did move enough to name. The 10Y breakeven rose to 2.34%, WoW +7bps the largest single WoW change in the block, and the reason the real 10Y rate fell to 2.35% (WoW -6bps). Nominal yields flat, inflation compensation up, real rate down: that is a market pricing slightly more inflation and slightly less restriction, which is a supportive mix for the risk assets already leading.
The one deterioration: HY credit spreads widened to 2.75% as of 2026-08-20, WoW +8bps. At 2.75% this is still a compressed absolute level, and one week of widening establishes nothing. It is worth flagging only because credit is where a narrow tape usually cracks first, and it is the one series moving against the equity read. Inferred, not measured: if the RISK-ON call is going to break, HY spreads are the more likely early tell than VIX at 14.89. That inference is wrong if spreads reverse back toward last week's level while breadth improves.
Labor data stayed benign. Initial claims 206K as of 2026-08-15, WoW -6K. Unemployment 4.1% and nonfarm payrolls 158.9M, both as of 2026-07-01. Fed funds 3.63% as of 2026-07-01. Housing starts 1,239K as of 2026-07-01. The monthly series are all July-dated they describe a labor market from seven weeks ago, and nothing in this week's tape updated them.
Themes in Motion
Ten themes are tracked. Five carry an ACCELERATING tag, five MATURING.
Strengthening. AI enterprise software is the largest accelerating cohort at 19 names, and the widest of any theme in the book. Medtech & diagnostics follows at 11, precision biotech & therapeutics at 9. Crypto exchanges & financials accelerates on the smallest base 4 names, which is a cohort thin enough that a single constituent can set the label.
Maturing. The rollover is concentrated in what led earlier: AI chips & memory (6), GPU cloud & neoclouds (5), space economy (9), critical materials & rare earths (6), M&A & special situations (7), and oncology & immunology (6).
The shape is worth stating plainly. Compute infrastructure chips, memory, GPU clouds is maturing while the software layer built on top of it accelerates. Those are adjacent links in one chain moving in opposite directions. This is a classification read, not a price measurement: no return data for these cohorts is in hand, so the labels describe where the model places each theme, not what any of them returned. The read is falsified if chips and memory re-accelerate while enterprise software's label degrades in a subsequent refresh.
Precision biotech ACCELERATING alongside oncology & immunology MATURING is the same split inside healthcare platform and modality moving one way, indication-specific therapeutics the other.
Under the Lens
787 dossiers were deep-refreshed this week. That is a sweep, not a selection it re-scores the universe rather than concentrating on a shortlist, and the alphabetical head of the run is what is visible here: AA, AAOI, AAON, AAPL, ABCL, ABEO, ABNB, ABT, ABVX, ABX, ACB, ACDC, ACHC, ACHR, ACLS, ACMR, ACN, ADBE, ADI, ADPT, ADSK, ADTN, AEHR, AEIS, AEM, AESI, AFRM, AGIO, AGX, AGYS, AIP, AIRO, AKAM, AKBA, AKTS, ALAB, ALGM, ALGT, ALHC, ALK.
A caveat that governs everything below: the visible slice is 40 of 787 names, and it is alphabetical, not ranked. Conviction ordering across the full refresh is not in hand. Nothing here should be read as the week's best ideas they are the names the visible slice puts against this week's theme labels.
Where the refreshed names intersect the theme map:
- ALAB (Astera Labs) connectivity silicon for AI racks. Sits in AI chips & memory, the MATURING side of the compute split.
- ACMR (ACM Research), ACLS (Axcelis), AEHR (Aehr Test Systems), AEIS (Advanced Energy), ALGM (Allegro MicroSystems) semi equipment and test, the same maturing cohort. Six of the week's visible names cluster in one label that is rolling over.
- ADBE (Adobe), ADSK (Autodesk), ACN (Accenture), AGYS (Agilysys) the enterprise software layer, the theme tagged ACCELERATING with the widest cohort.
- ABCL (AbCellera), ABVX (Abivax), ADPT (Adaptive Biotechnologies), AGIO (Agios), ABEO (Abeona) precision biotech, ACCELERATING.
- ABT (Abbott), ALHC (Alignment Healthcare), ACHC (Acadia Healthcare) medtech, diagnostics and healthcare services, ACCELERATING at 11 names.
- AA (Alcoa), AEM (Agnico Eagle) materials and mining, adjacent to the critical materials theme tagged MATURING.
- ACHR (Archer Aviation), AIRO (AIRO Group) aviation and defense-adjacent, near the space economy label, also MATURING.
The strongest structural observation available from this slice: the semis-equipment names and the enterprise-software names both sit in the refresh, on opposite sides of the accelerating/maturing line, with no price data here to adjudicate which label is early and which is late. Measured: the labels and the cohort counts. Inferred: that the split is a real rotation rather than a lag in how the model tags. That inference fails if the next refresh moves chips and memory back to ACCELERATING without any change in the software cohort.
The Week Ahead
Fifteen dated catalysts land inside the next 14 days, and fourteen of them fall in the first three sessions.
Sunday, August 23 (T-1d): RARE.
Monday, August 24 (T-2d): the cluster AMD, BAK, CLW, FRD, IESC, LSTR, MRX, MSTR, NBIS, PAYC, RXO, WLFC, WNC. Thirteen dates on one day.
Tuesday, August 25 (T-3d): ALOT.
Three of Monday's names map onto the themes above and are the ones worth watching for what they say about the labels rather than about themselves:
- AMD AI chips & memory, MATURING. If the maturing tag is a real deceleration rather than a tagging lag, this is the print where it should show.
- NBIS GPU cloud & neoclouds, MATURING, a 5-name cohort. Thin enough that one result moves the label.
- MSTR sits against crypto exchanges & financials, ACCELERATING on only 4 names. Same thinness problem in the other direction.
- PAYC enterprise software, the widest ACCELERATING cohort.
What the week actually resolves: AMD and NBIS on one side of the compute split, PAYC on the other, all inside three sessions. That is a genuine test of the chips-maturing / software-accelerating read not a confirmation of it, since four prints across two 5-and-4-name cohorts is far too thin to settle a theme. A binary sits in front of each of these and nothing about the theme labels resolves before Monday's close.
Two macro conditions carry into the week. HY at 2.75% is the series to watch a second consecutive week of widening would be the first non-trivial argument against the RISK-ON read, whereas a reversal leaves the +8bps as noise. And breadth at 55.7% is the number that decides whether SPY's +7.3% over its 200-EMA is a broadening advance or a narrowing one; a move up through healthy resolves it one way, a slide with the index holding resolves it the other.
--