Market scan, August 2026

Purpose. An outside-in check on what has changed in the real world, and what this repository’s audited shelf and research corpus would now get wrong. Written for the public website rewrite.

Read date for everything below is 2026-08-22 or 2026-08-23 unless a line says otherwise. Where a figure was computed rather than quoted, the computation is stated.

Headline finding, stated up front so the rest is read correctly. This repository is not stale. The shelf’s asOf dates are 2026-08-17 to 2026-08-23, and independent verification of all 59 shelf entries’ expense ratios found two discrepancies, both under 3 bp. The pricing page’s FRED figures reproduce to the decimal. The tax page’s 2026 limits match the IRS notices. What the repository is missing is not accuracy — it is coverage: products that exist and are not on the shelf, costs that are not in the cost model, and two term-premium series it declared unavailable that are in fact available.

---

Section 0 — What the repository already covers

The shelf, as committed (src/content/shelf.ts, 59 entries)

Also priced or discussed in prose but not on the typed shelf: the eight US spot bitcoin ETPs (BTC, EZBC, BITB, HODL, ARKB, IBIT, FBTC, GBTC), the ether staking trusts (ETHE, ETH, ETHB, ETHA), and the cat bond ETF.

Docs read before searching

factor-products.md, alternative-sleeves-audit.md, capital-efficiency-and-breadth.md, live-managed-futures.md, untested-tilt-candidates.md, current-regime-and-pricing.md, valuation-and-the-allocation.md, structural-and-tax-edges.md, harvesting-and-direct-indexing.md, src/content/placement.ts.

---

Section 1 — Fee and product changes to funds already on the shelf

Verified against issuer pages and StockAnalysis fund pages, read 2026-08-22/23. Fees are net expense ratios. AUM is the figure each source displays; the shelf carries no AUM field at all, which is itself a gap (§CORRECTIONS C-9).

Priority holdings

Sources: Vanguard VTI, Avantis AVUV, Avantis AVDV, Return Stacked RSST, Simplify CTAP, Simplify CTAP summary prospectus, iMGP DBMF factsheet, Man MATE, StockAnalysis fund pages (e.g. AVUV), all read 2026-08-22/23.

Second tier — all verified, all matching

VOO 0.03% / $1.04tn · ITOT 0.03% / $97.17bn · SPY 0.09% / $820.19bn · DFUV 0.21% / $15.89bn · DFLV 0.21% / $7.23bn · DFSV 0.30% / $8.40bn · DFAT 0.28% / $14.65bn · VBR 0.05% / $38.19bn · QVAL 0.28% / $631.54m · AVSC 0.25% / $3.21bn · DFAS 0.26% / $15.79bn · VB 0.03% / $82.69bn · QUAL 0.15% / $47.68bn · SPHQ 0.15% / $19.10bn · DUHP 0.20% / $12.75bn · VEA 0.03% / $236.45bn · SPDW 0.03% / $41.92bn · IEFA 0.07% / $195.14bn · AVIV 0.25% / $2.10bn · DISV 0.42% / $5.17bn · SCZ 0.40% / $13.28bn · GWX 0.40% / $932.41m · IDMO 0.25% / $4.33bn · IMTM 0.30% / $4.31bn · VWO 0.06% / $125.07bn · IEMG 0.09% / $160.00bn · DFEV 0.43% / $2.07bn · BND 0.03% / $162.86bn · SPAB 0.03% / $10.28bn · FMF 0.98% / $263.27m · WTMF 0.66% / $258.43m · NTSX 0.20% / $1.39bn · GDE 0.20% / $481.87m · SCHD 0.06% / $112.00bn · VNQ 0.13% / $39.05bn · TIP 0.18% / $15.07bn · SCHP 0.03% / $16.50bn.

No closures, liquidations or mergers were found among the 59.

The one real product change: Vanguard’s funds were renamed

The shelf’s odd-looking names — “Vanguard Morningstar Total Stock Market ETF”, “Vanguard Morningstar Small-Cap Value ETF” — are correct, and this was worth verifying because it looks like an error.

• Morningstar completed its $375m acquisition of CRSP in February 2026 and rebranded the CRSP Market Indexes as Morningstar Indexes (Morningstar newsroom, read 2026-08-22).
• Vanguard announced on 2026-04-29 that it would update the names of 13 US equity index funds (all share classes, mutual fund and ETF) tracking those indexes (Vanguard press release).
• The renaming took effect 2026-07-29. Vanguard’s own VTI profile page now reads “VTI — Vanguard Morningstar Total Stock Market ETF”.
• Methodology, holdings, fee and ticker are unchanged. This is a branding event, not an index change. The shelf is right and its verdict prose is right.

Caveat the shelf should carry anyway: three shelf names now embed a benchmark-provider brand that could change again. A fund’s identity is its ticker and its CIK, not its marketing name.

JPFP’s first Form N-PORT is still not filed

Checked EDGAR directly (J.P. Morgan Exchange-Traded Fund Trust, CIK 0001485894). The trust’s most recent N-PORT-P batches were filed 2026-06-26 and 2026-07-28 (24 filings), and filings run through 2026-08-18 with no JPFP series report. capital-efficiency-and-breadth.md predicted the first filing is due 2026-08-29 or 2026-09-29; that is still the state, and the earlier of those two dates is six days away. delta for JPFP remains uncomputable. This is the single cheapest open item in the whole corpus and it resolves within a month.

---

Section 2 — New products since roughly mid-2025

2.1 The Return Stacked family has grown to eight funds; the shelf carries two

This is the largest single coverage gap. From returnstackedetfs.com, read 2026-08-22, AUM as of 2026-08-20:

Partner funds on the same page: BTGD (STKd 100% Bitcoin & 100% Gold, 2024-10-15, 0.99%, $51.05m), ISBG and ISSB (IncomeSTKd, both 2026-01-20, 1.14% net, $6.98m and $1.97m).

RSIT is the one that matters and it is three and a half months old. Per the issuer page, the equity leg holds roughly 75% SPDW plus 25% MSCI EAFE index futures — a base leg of approximately 1.00 per dollar of capital and a trend leg of 1.00, which puts delta = (1 − b)/d ≈ 0.00 and makes it structurally the international twin of RSST: it keeps the entire funding-rule gap. This repository has just recommended adding AVDV (untested-tilt-candidates.md §“AVDV: add it”) and separately finds trend to be the best-supported financed diversifier. RSIT is the first vehicle that stacks those two legs in one wrapper. It has no N-PORT and no measurable loading, so nothing can be promoted from it — but it belongs on the shelf as not filed, exactly as JPFP does.

RSSX matters for a different reason: it is a financed gold-and-bitcoin sleeve on top of US equity, obtained through IBIT and CME bitcoin futures. alternative-sleeves-audit.md concludes bitcoin should be funded from the speculation budget by selling equity; RSSX is the construction in which it is not. That does not make the sleeve better — the funding algebra improves the hurdle, it does not create a premium where the return evidence is a price expectation — but the repository’s crypto verdict is currently written against a pro-rata construction only, and a financed one now exists at 0.67%.

2.2 Managed futures and trend

• FFUT — Fidelity Managed Futures ETF. Inception 2025-06-05, $346.23m, and 0.80% — which is an expense cap running to 2027-05-31, not a fee: the table reads 0.80% management plus 0.02% other expenses for 0.82% gross, and FDS may recoup within the fiscal year what it reimburses (summary prospectus dated 2026-05-30, read 2026-08-24). Trend-following across equity, rates, FX and commodities. (etf.com, StockAnalysis, read 2026-08-22.) Not on the shelf, and it is now the fourth- or fifth-largest managed-futures ETF. Verdict: repackaging of a known mechanism, but from an issuer with distribution — its survival probability is higher than most of the shelf’s small trend funds, which is the variable live-managed-futures.md shows actually decides outcomes (52% of the 2019 cohort stopped filing inside 6.5 years).
• JPFP — JPMorgan Managed Futures Plus. Inception 2026-05-27, 0.59%, $32.75m. Already on the shelf.
• MATE — Man Active Trend Enhanced. Inception 2025-12-16, 0.97%, $39.76m. Already on the shelf.
• Category size. Managed-futures ETFs reached roughly $6bn by March 2026 across about 13 funds, against a CTA industry above $300bn (HedgeNordic, read 2026-08-22). The category is growing, not consolidating — but DBMF alone is $4.0bn of the $6bn, so the rest of the shelf’s trend funds are competing for about $2bn. Concentration risk in a category whose main failure mode is closure.

2.3 Systematic value, momentum, quality

Nothing that changes an allocation. Avantis has continued to ship one-ticker fund-of-funds solutions rather than new engines: AVGV (All Equity Markets Value, 0.26%, $458.26m), AVNV (All International Markets Value, 0.34%, $69.28m), AVNM (All International Markets Equity, 0.31%, $742.86m), AVMA (Moderate Allocation, 0.21%, $86.52m). Verdict: repackaging — they hold the same underlying Avantis sleeves the shelf already prices, at a wrapper fee. They are worth a line on the shelf only as an implementation convenience for a small taxable account, not as candidates.

2.4 Defensive, tail-risk, options-based

No launch found that changes alternative-sleeves-audit.md §4’s conclusion (reject explicit tail hedges; the cheaper substitutes win). I could not run an exhaustive census of the buffered/derivative-income category within the search budget available — this is a stated coverage gap, not a null result.

2.5 Catastrophe bonds — the vehicle is now three times the size the audit recorded

The Brookmont Catastrophic Bond ETF trades as ILS, which is the ticker alternative-sleeves-audit.md §5 already uses throughout. ROAR appeared in pre-launch coverage and was never used; it is recorded here only so that a reader who meets it in an article knows which fund it means. The fund listed on NYSE in April 2025, the first US-listed cat bond fund, advised by Brookmont with King Ridge Capital Advisors as sub-adviser (ETFGI, ilsetf.com).

What is new here is the size, and only the size.

• AUM was ~$12m in August 2025, $25.3m at 2025-11-26 (Artemis), and $88.44m at 2026-08-22 (StockAnalysis). The audit’s $88.2m is dated 2026-08-20; the two readings are two days apart and both are current for their date.
• Average daily trading volume rose ~672% between September and November 2025.

The fee belongs to the audit and is not restated here. The audit’s figure is not a flat 1.58%: it is 2.65% gross and 1.58% net under a cap running to 2027-04-30, against costs actually paid of 2.00% annualised in the fund’s own tailored shareholder report. Three numbers, all of them true at once, and quoting the middle one alone understates what a holder has actually paid. Nothing in this scan revises any of them.

The audit’s access finding is now understated: the vehicle has survived eighteen months and tripled in nine. That does not touch the audit’s actual objection, which is price — the risk spread has compressed by roughly half since 2023 — and the reopening condition there is a spread level, not an AUM level. But “the retail history is short” is nine months shorter than it was.

2.6 Crypto ETPs

The regulatory change is the story, not any individual fund. In September 2025 the SEC approved generic listing standards for commodity-based trust shares, replacing case-by-case 19b-4 review with a rules-based path and cutting time-to-list from 240+ days to roughly 75. An asset qualifies if it trades on a regulated market, has had CFTC-regulated futures for six months, or an existing ETF already holds 40% of its assets in it (CNBC, read 2026-08-22). Spot Solana and XRP ETPs reached the market from November 2025.

Current fees and net assets, from StockAnalysis fund pages read 2026-08-22 — except HODL’s and EZBC’s fees, which are from those trusts’ own Q2-2026 Forms 10-Q, read 2026-08-24, because the aggregator’s readings for both were wrong:

The two divergences from the audit’s Q2-2026 10-Q table are the aggregator’s, not the audit’s. The table above shows HODL at 0.25% and EZBC at 0.29%; both trusts’ own Q2-2026 Forms 10-Q, read 2026-08-24, say 0.20% and 0.19% respectively, and HODL’s filing states in terms that after 2026-07-31 — the waiver expiry the audit already flagged — “the Sponsor Fee will be 0.20%”. The audit’s figures stand, the table above carries the filed ones, and the aggregator’s 0.25% and 0.29% appear nowhere in either trust’s filings; see C-16. The generalisation is the useful part: a spot-ETP sponsor fee under waiver is the single field a data aggregator here got wrong twice out of sixteen, and it is cheap to check in the trust’s own quarterly filing.

IBIT went from $43.4bn at 2026-06-30 to $58.78bn — roughly +35%, against bitcoin’s +31.6% price move over the same stretch, so most of that is mark-to-market rather than flow.

Verdict on the new altcoin ETPs: repackaging with worse economics. Solana and XRP ETPs are the same 1933-Act grantor-trust wrapper around an asset with a shorter history, thinner benchmark infrastructure, and no cash-flow claim. Nothing in the repository’s crypto argument turns on which coin is in the trust.

2.7 Structural changes to what is buyable

• ETF share classes of mutual funds. The SEC granted Dimensional’s exemptive relief on 2025-11-17, the first for an active manager and the second ever after Vanguard’s expired patent (Dimensional, Seward & Kissel 40 Act Blog). Substantially identical applications from other managers follow. Why it matters to this repository: the tax-efficiency argument in structural-and-tax-edges.md and placement.ts rests partly on the ETF in-kind shield. If a Dimensional mutual fund acquires an ETF share class, the mutual fund’s capital-gain distributions fall toward the ETF’s. That changes the placement arithmetic for anyone holding DFA mutual funds in a taxable account, and it changes the “ETF or mutual fund” question from a structural one to an empirical one. No shelf fund is affected yet.
• Generic crypto listing standards (September 2025) — above.

---

Section 3 — The current pricing environment

Every FRED figure below was refetched directly from fredgraph.csv on 2026-08-23 for this scan, independently of the repository’s cache. All of them reproduce the values in current-regime-and-pricing.md exactly. That page is correct as written.

3.1 Rates and inflation

A wrinkle worth noting that the repository’s page does not: headline CPI YoY is 3.30% while core CPI YoY is 2.47%. Headline is running above core, which is the reverse of the usual configuration and means the 3.54% trailing-CPI figure the pricing page uses for the ex-post real bill is being driven by non-core components. The real-cash-rate conclusion does not change, but the “+0.19 pp real bill” is more volatile than it looks.

3.2 Term premium — the repository says this is unavailable; it is not

current-regime-and-pricing.md §1.3 states: “No term-premium estimate is supported by this cache. The Adrian–Crump–Moench and Kim–Wright decompositions are not in it… unresolved, and honestly so.”

Both are obtainable:

• Kim–Wright 10-year term premium: +0.8393%, as of 2026-08-14, from FRED THREEFYTP10 — 9,555 daily rows back to 1990-01-02, fetched 2026-08-23. This is a Federal Reserve Board series distributed on FRED. It is in the same data contract the pricing page already uses.
• NY Fed ACM 10-year term premium: approximately +0.80% as of 2026-08-13, against 0.513% in June 2026 and 0.667% in May 2026 (secondary aggregators citing the NY Fed release; primary table at newyorkfed.org term premia). Treat the ACM figure as medium confidence — I read it from aggregators, not from the NY Fed’s own CSV.

The two estimates agree at roughly +0.8 pp, which is materially positive for the first time in about five years. That is a substantive addition, not a bookkeeping one: it says the +0.82 pp 10y−3m slope is roughly all term premium and roughly no expected rate change, which the pricing page explicitly declines to conclude because it lacked the decomposition. The honest label moves from unresolved to measured, two independent estimates, both ≈+0.8.

3.3 Credit — the data-contract finding is confirmed and still binding

Fetched 2026-08-23:

• ICE BofA US High Yield OAS = 2.75%, latest 2026-08-20. 796 daily rows, series begins 2023-08-22. (FRED BAMLH0A0HYM2)
• ICE BofA US Corporate (IG) OAS = 0.82%, latest 2026-08-20. 796 daily rows, begins 2023-08-22. (FRED BAMLC0A0CM)

The truncation is real and current. The pricing page’s finding — that no percentile of a current OAS against its own history can be computed from FRED — reproduces exactly, three years to the day. Anyone rewriting the public site should not quietly quote a “high yield spreads are in the Nth percentile since 1997” line, because the underlying series no longer supports it from this source.

3.4 US equity valuation

The repository’s CAPE of 41.18 and forward P/E of 20.0 are both current and both correctly sourced. shillerdata.com itself does not display the value — it only serves ie_data.xls — so the repository’s practice of reading it from its own cache and cross-checking against aggregators is the right one.

The FactSet detail worth carrying into the site rewrite: a forward P/E of 20.0 that is only 0.1 above its own five-year average, resting on +30% consensus CY2026 earnings growth. The multiple is not the extreme; the earnings denominator is. A CAPE of 41 at the 98.9th percentile and a forward P/E of 20 at roughly the 55th percentile of five years are both true, and the pricing page is right to call the tension unresolved.

3.5 US versus international — this is where the year actually happened

Computed from FRED SP500, fetched 2026-08-23 (price return, dividends excluded):

• S&P 500 calendar 2025: +16.4% price (≈+17.9% total return).
• S&P 500 2026 YTD to 2026-08-21: +12.1% (6,845.50 → 7,674.37).
• S&P 500 first half 2026: +9.6%.

Against that:

• MSCI EAFE returned ~+31.2% and MSCI EM ~+33.6% in calendar 2025, against the S&P 500’s +17.9% — roughly 13 to 16 percentage points of international outperformance in one year, of which a little over 10 pp came from dollar weakness (Yahoo Finance / MSCI coverage, The London Company, read 2026-08-22).
• 2026 has been closer. Some coverage claims the S&P 500 “slipped about 2% in 2026” — that is stale or wrong; the index is +12.1% YTD on primary data. Treat mid-year commentary on 2026 relative performance as unreliable and re-measure it.

Valuation gap (repository’s own figures, Siblis Research at 2026-06-30, read 2026-08-22): US CAPE 35.82, global ex-US 21.02, EM 19.36 — a 1.70× US premium over developed ex-US and 1.85× over EM. Secondary coverage puts developed international and EM at roughly 19× and 18× forward earnings against the S&P 500 near 29× trailing.

The dollar. Fetched 2026-08-23 from FRED DTWEXBGS:

• Broad trade-weighted dollar index 118.9028 at 2026-08-14.
• −1.44% over one year, −0.97% over three years.
• 57th percentile of its own ten-year range (10y min 106.49, max 130.04).

That last line is the one that should temper the story. The dollar delivered 10 pp of EAFE’s 2025 return, and it is still at the 57th percentile of the last decade — mid-range, not cheap. The currency tailwind is not obviously spent, and it is not obviously repeatable. currency-and-the-international-sleeve.md owns this; the repository does not currently carry a dollar level anywhere I could find, and given that a tenth of last year’s international outperformance was currency, it should.

3.6 Gold

• Spot ~$4,587/oz at 2026-08-21; another read $4,607.35 the same day (Fortune, 2026-08-21). The repository’s “2026-08-21 fix of $4,582” is consistent.
• But gold peaked in January. Gold futures hit a record $5,542.40 on 2026-01-29. Gold is therefore roughly −17% from its high while sitting at the 98.5th percentile of its own real price since 1975.

Both facts are true and the pricing page carries only the second. “Most expensive thing on the page relative to its own history, and 17% below its own January high” is a more complete and more useful sentence than either half.

3.7 Capital market assumptions

Vanguard Capital Markets Model, 10-year annualised nominal, June 30 2026 running (Vanguard VEMO return forecasts, read 2026-08-23):

Vanguard notes value stocks continue to offer the most attractive expected-return profile within US equities. I could not extract the bond, small-cap or REIT rows from the page.

Read this carefully, because it cuts against the obvious story. Vanguard now forecasts emerging markets below US equities — 2.0–4.0% against 4.2–6.2% — after EM returned +33.6% in 2025. Developed ex-US retains only a 0.3 pp median edge over the US. A model that conditions on valuation has already priced in most of the 2025 international rally and cut its forward EM number by 1.6 pp in a single quarter. The valuation gap is 1.70×; the model-implied return gap is 0.3 pp. Anyone rewriting the site should not present the valuation gap as though it were an expected-return gap.

I could not reach GMO’s or Research Affiliates’ current vintages within the search budget. Stated gap.

3.8 Does the evidence support changing an allocation on valuation? Yes to the framing the

repository already uses, and the last twelve months are a live test of it

I could not complete a fresh literature sweep — the session’s search budget was exhausted before Task B. What I can do is check the repository’s position against what actually happened, which is the more informative test.

valuation-and-the-allocation.md’s position, as written: CAPE’s out-of-sample record does not support timing the equity share; the relative US/international call is “a different and better-posed question”; the cross-country relation is undetectable after 1990; and the recommendation is a fixed split with the decision framed as regret rather than forecast. current-regime-and-pricing.md reinforces it: of twenty-seven conditioning relations, none predicts the equity premium out of sample, and the term spread’s five-year equity regression carries a Hodrick t of 2.32 with an out-of-sample R² of −0.175.

Three checks against 2025–2026:

1. The relative call paid, once. International beat the US by 13–16 pp in 2025. A valuation-motivated tilt would have been rewarded. One year is one observation, and the repository’s own §5.2 already measures the confounds.
2. Ten pp of the 13–16 was currency, not re-rating. That is the confound the repository names and it dominated the result. The valuation gap barely closed: US CAPE 35.82 vs ex-US 21.02 at 2026-06-30 is still a 1.70× premium.
3. The forward-looking models did not extrapolate. Vanguard cut its EM forecast to 2.0–4.0% and left developed ex-US only 0.3 pp ahead of the US. A model that both conditions on valuation and is disciplined about it says the gap is mostly already collected.

The repository’s view holds. Nothing in the last twelve months supports moving the equity share on CAPE, and the relative call — which the repository already treats as better-posed — delivered its payoff in a form (currency) that its own currency page identifies as the least persistent driver. The one thing I would add to the page: a dollar level with a percentile, because the 2025 result is unreadable without it.

Not verified in this scan (stated gaps): Goyal, Welch & Zafirov’s updated paper; the Siegel / Straehl–Ibbotson / Philosophical Economics accounting-comparability arguments; the post-publication out-of-sample record of valuation-timing rules. The repository’s own §2.3 and §2.4 already cover the last two and were not contradicted by anything found here.

---

Section 4 — Crypto, honestly

All correlation, beta, drawdown and worst-decile figures in this section were computed by me on 2026-08-23 from FRED CBBTCUSD (Coinbase BTC/USD) and FRED SP500, using log returns. The SP500 series is limited to ten years, so the monthly panel is 120 months (2016-08 to 2026-08), not the repository’s 137.

4.1 Is the repository’s verdict still true?

Substantially yes, and one figure has moved in bitcoin’s favour.

The full worst-decile table (worst 12 of 120 months by S&P 500 return):

Two positives in twelve, and one of them (+62.2% in May 2019) is a bull-market month that happened to coincide with an equity dip. The 2026-03 observation is new since the audit was written and it is a genuine, if small, point in bitcoin’s favour.

4.2 Trailing returns, which are the part that has genuinely changed

Computed 2026-08-23:

Over five years, bitcoin has underperformed the S&P 500 at roughly four times the volatility (42.2% realised over the last year against the index’s 12.9%). That is not a selective window: it is the longest horizon the daily panel supports cleanly, and it covers the spot-ETF launch, the 2024 halving, and the entire institutional-adoption story.

4.3 What has structurally changed, and whether it matters

Genuinely changed:

• Flows are large and institutional. IBIT alone holds $58.78bn. US spot bitcoin ETFs took $517m in a single day on 2026-08-19, the largest since early May, and about $853m in the strongest week since April 2026 (CoinDesk, 2026-08-20).
• Volatility has collapsed. 30-day realised volatility fell to an annualised 42%, the narrowest volatility gap to the S&P 500 on record (Coincall market note, 2026-08-19).
• Regulation opened up. Generic listing standards (September 2025); Rev. Proc. 2025-31’s staking safe harbour, which the repository already documents.
• A financed vehicle now exists: RSSX (§2.1).

What that does to the argument: less than it sounds.

Every one of those changes makes bitcoin more like a financial asset — more correlated, better arbitraged, lower-vol, more owned by the same institutions that own equity. The repository’s stated reopening trigger is “a correlation to equity back below +0.2 on a window containing a recession.” The 3-year daily correlation is +0.170, which is under the threshold — but the window contains no recession, and the 3-year monthly reading is +0.363. The trigger is not met. Financialisation is the mechanism that removes the diversification case, not the one that establishes it. A short-term correlation of 83.6% to the S&P 500 in early August 2026 is the same point from the other side.

4.4 The fair hearing

What can honestly be said for a small position, and what cannot:

• Cannot: that it is a diversifier. Five of the last five years of evidence, including the ETF era, say it is high-beta equity with idiosyncratic risk on top. It has now underperformed equity over five years while doing that.
• Cannot: that a cash-flow claim has arrived for bitcoin. Ether staking is a real contractual payer and it is not bitcoin — the repository already has this exactly right.
• Can: that the loss from a 1–2% position going to zero is 1–2%; that the investor asked for it; and that a holding one wants is easier to keep than one one resents. That is a preference argument, and the repository already labels it as one.
• Can, and is new: that if it is held, the financed construction (RSSX at 0.67%) has a better hurdle than selling equity to fund it, because the funding-rule gap is real arithmetic independent of any premium. This does not make the sleeve good. It makes the worst version of holding it less bad.

Bottom line: the repository’s verdict — at most 1–2%, from the speculation budget, in taxable, labelled a speculation — survives everything measured here, and one of its supporting figures (1 of 13) should be updated to 2 of 12.

---

Section 5 — Tax and account rules for 2026

Verified against IRS primary sources. The repository’s figures are correct. The one gap is a rule it does not appear to mention at all.

5.1 Retirement limits — IRS Notice 2025-67

Extracted directly from Notice 2025-67 (PDF), read 2026-08-23, and cross-checked against the IRS newsroom release:

The repository’s structural-and-tax-edges.md §1264 and placement.ts carry the deferral, catch-up, IRA and Roth phase-out figures correctly. They do not carry §415(c) = $72,000, which is the number that bounds the mega-backdoor Roth.

5.2 The rule the repository does not mention: mandatory Roth catch-up

SECURE 2.0 §603, effective for taxable years beginning in 2026. Notice 2025-67 line 41–44: the §414(v)(7)(A) wage threshold “used to determine whether catch-up contributions must be designated as Roth contributions, is increased from $145,000 to $150,000.”

Concretely: if your 2025 Social Security (FICA) wages from the employer sponsoring the plan exceeded $150,000, then every catch-up dollar in 2026 — the $8,000 age-50 catch-up or the $11,250 age 60–63 catch-up — must be a designated Roth contribution. No pre-tax option.

Why this matters here specifically. capital-efficiency-and-breadth.md reaches a placement conclusion — a stacked trend wrapper belongs in the pre-tax account, not the Roth and not taxable — and then notes that a 30%-of-portfolio wrapper “consumes roughly nine tenths of a one-third pre-tax account.” This rule makes the pre-tax account smaller for exactly the investor the page is written for: a high earner over 50 loses the ability to add $8,000–$11,250 of pre-tax space per year and gains Roth space instead. The scarce shelter the wrapper needs is now scarcer, and it gets scarcer every year. That is a real, dated, decision-relevant change to a conclusion the repository has already reached.

5.3 Capital gains, NIIT, and the loss cap

From Rev. Proc. 2025-32 (PDF) §3.03 and §3.14, extracted 2026-08-23:

2026 long-term capital gain / qualified dividend thresholds:

2026 standard deduction: MFJ $32,200 · HoH $24,150 · single $16,100 · MFS $16,100 · additional for aged/blind $1,650 ($2,050 if unmarried and not a surviving spouse).

The repository already cites Rev. Proc. 2025-32 §3.03 and the $613,700 figure. Correct.

§1411 NIIT: 3.8%, on modified AGI above an unindexed $250,000 (MFJ) / $200,000 (single). Unchanged; the threshold has never been indexed. The repository states this correctly and correctly observes that any investor in the 18.8% or 23.8% column is at or past the Roth IRA phase-out.

§1211(b): still $3,000, still not indexed. Congressional Research Service analysis and 2026 practitioner guidance confirm that the One Big Beautiful Bill Act made no change to §1211, §1212, or the wash-sale rule in §1091 (CRS RL31562). Indexing from 1978 would put the cap near $13,000. harvesting-and-direct-indexing.md’s entire argument — that the $3,000 cap, not the harvest yield, is the binding constraint — is intact and current.

5.4 Wash sales and crypto

No change. §1091 applies to “stocks or securities,” digital assets are not securities for this purpose, and no 2025–2026 legislation altered that. Proposals to extend §1091 to digital assets have appeared repeatedly in budget documents and have not been enacted. The OBBBA did not do it.

Medium confidence. I confirmed the OBBBA point from CRS and practitioner sources but could not run a fresh check on 2026 crypto market-structure legislation before the search budget ran out. Re-verify before publishing any claim that crypto tax-loss harvesting is unrestricted.

5.5 Backdoor and mega-backdoor Roth

Both remain available. No 2025–2026 legislation eliminated either. The mega-backdoor ceiling is the §415(c) limit, $72,000 for 2026 (less elective deferrals and employer contributions), and requires a plan that permits after-tax contributions plus either in-plan Roth conversion or in-service withdrawal. The backdoor Roth remains subject to the pro-rata rule of §408(d)(2) across all traditional/SEP/SIMPLE IRAs.

5.6 Qualified dividends and the foreign tax credit

• Qualified dividend treatment is unchanged: the LTCG rate schedule above, subject to the §1(h)(11) holding-period requirement.
• Foreign tax credit through funds: §853(a)(1) requires more than 50% of a fund’s total assets at year end to be foreign securities for it to elect to pass foreign taxes through to shareholders. currency-and-the-international-sleeve.md line 507 has this right. The practical consequence the repository already draws — the credit is lost in a tax-deferred account — is unchanged.

---

Section 6 — Costs an investor actually pays

6.1 Two brokers are changing their fee structure, and this is the most decision-relevant

item in the scan after RSIT

Fidelity, effective 2026-06-01: Fidelity issued a notice on 2026-03-16 requiring ETF issuers to pay asset-based platform fees to remain accessible. Issuers that refused were placed on a “service fee eligible” list, and purchases of their ETFs now carry $100 per trade (described as roughly 5% of trade value, capped at $100) (ETF Investments, read 2026-08-23; RIABiz, 2026-05-07).

Schwab: CEO Rick Wurster has confirmed a year-end 2026 rollout of a comparable programme claiming up to 15% of ETF fee revenues, with the same $100-per-trade penalty for non-participating managers. Timing described as flexible.

Vanguard: reportedly reached an undisclosed agreement with Fidelity by May 2026, while maintaining it does not pay distribution fees that incentivise sales.

Scale: combined Schwab and Fidelity assets of $26tn, of which about $3.8tn of ETF assets are liable.

Issuers named as penalised as of June 2026: Roundhill (40+ funds), Kurv, Renaissance, XOVR, IVES, Inspire, Hedgeye, Rareview, Polen, Brandes, WEBs Defined Volatility, Cyber Hornet, Guinness Atkinson, Strategy Shares, Faith Investor Services.

None of this repository’s shelf issuers appears on any published list I could reach. But I could not read Fidelity’s live service-fee list — fidelity.com/trading/etfs/service-fee-etfs returns HTTP 403 to automated fetches. This is an unresolved verification, and it is the one I would resolve first, because:

• A $100 flat fee on a $10,000 purchase is 100 bp — larger than the entire expense ratio of every fund on the shelf except the trend wrappers, and roughly four times AVDV’s 36 bp annual fee, paid on day one.
• The exposed issuers are structurally the small ones: Simplify ($39m–$1.6bn per fund), Tidal/Return Stacked ($52m–$523m per fund), Brookmont ($88m), Man ($40m), iMGP. These are precisely the boutiques with the least fee revenue to hand over. Avantis (American Century), Dimensional, iShares, Invesco, Vanguard, SSGA, WisdomTree, Schwab and JPMorgan are not plausibly at risk; Simplify, Tidal, Brookmont, Man Group and KraneShares plausibly are.
• A conclusion that a wrapper is buyable is a conclusion about a broker, not only a fund. capital-efficiency-and-breadth.md compares CTAP, RSST and MATE at 0.81%–0.99%; a $100 entry fee at the investor’s actual broker would reorder that comparison, and none of the three is large enough to be safe from the list.

Commissions themselves are unchanged: Fidelity, Schwab and Vanguard all remain $0 on online stock and ETF trades. Options differ — Schwab $0.65/contract; Vanguard restructured its schedule on 2026-07-10 into tiers based on “Qualifying Assets,” replacing its previous $1.00/contract flat rate.

6.2 Bid-ask spreads — the number the shelf’s cost model omits

Every ETF must publish its 30-day median bid-ask spread under Rule 6c-11(c)(1)(v). Read from issuer pages, 2026-08-22/23:

Sources: Simplify CTAP, Simplify SDMF, Simplify CTA, Return Stacked RSST, RSIT, RSSX, iShares product pages.

The CTAP finding is material and it changes a conclusion. capital-efficiency-and-breadth.md runs a careful, correct argument that CTAP’s fee table understates its true cost — the swap on affiliated CTA hides a 0.75% expense ratio, taking all-in from a headline 0.10% to about 0.81%, and to about 0.99% once the waiver lapses on 2026-12-04. That argument is right and it stops one step early. CTAP also has a 33 bp median bid-ask spread. A round trip costs about 66 bp, and even a one-way purchase held for a year adds 33 bp to a first-year all-in cost of roughly 0.81%. The page’s conclusion — “neither the funding rule nor the fee table separates the candidates” — is drawn from delta and the fee table. The spread does separate them: 33 bp against RSST’s 9 bp is a 24 bp/yr difference on a one-year hold, on a shelf where the entire fee dispersion among the three wrappers is 18 bp. For any investor who rebalances, CTAP is the most expensive of the three, not the cheapest. src/content/shelf.ts now carries a spread field, populated for the funds whose Rule 6c-11 disclosure was read here and absent for the rest.

The general point for the site rewrite: the shelf models fee − securities lending and calls it net cost. For a fund with a 1 bp spread that is nearly complete. For a fund with a 33 bp spread and an annual rebalance it is missing the largest term.

---

CORRECTIONS

Specific things in this repository that are now stale, wrong, or incomplete, with the correct current value. Ordered by how much they change a decision.

C-1 — docs/research/capital-efficiency-and-breadth.md: CTAP’s cost is understated by its bid-ask spread, and the “no candidate is separated” conclusion does not survive it. The page compares CTAP (~0.81% all-in, rising to ~0.99% on 2026-12-04), RSST (0.99%) and MATE (0.97%) and concludes the fee table does not separate them. CTAP’s 30-day median bid-ask spread is 0.33%; RSST’s is 0.09% (Simplify, Return Stacked, read 2026-08-22). A one-way purchase adds 33 bp to CTAP’s first-year cost; a round trip adds 66 bp. The candidates are separated, and CTAP is the dearest of the three, not the cheapest. SDMF’s spread is 0.24% against a 0.35% fee. Add a spread field to the shelf.

C-2 — src/content/shelf.ts: the Return Stacked shelf is 2 of 8, and the missing one that matters is RSIT. RSIT — Return Stacked International Stocks & Managed Futures ETF, inception 2026-05-06, 0.98%, $68.53m, median spread 0.15%, equity leg roughly 75% SPDW + 25% MSCI EAFE futures (so a base leg near 1.00 and delta ≈ 0.00). It is the international twin of RSST and it lands directly on the intersection of two live repository recommendations (add AVDV; trend is the best-supported financed diversifier). Also missing: RSSX (US Stocks & Gold/Bitcoin, 2025-05-29, 0.67%, $70.59m, spread 0.28%), RSSY, RSBT, RSBY, RSBA. All six are now on the shelf, at the fees their own Forms 497K dated 2026-04-27 (RSIT 2026-05-05) print, and none carries a measured structure. Source: returnstackedetfs.com, read 2026-08-22.

C-3 — Section 6 of everything: two brokers are introducing a $100-per-trade fee and the repository’s cost model has no broker in it. Fidelity began charging $100 per purchase on ETFs from non-participating issuers on 2026-06-01; Schwab has confirmed a comparable programme for year-end 2026 claiming up to 15% of issuer fee revenue. On a $10,000 purchase that is 100 bp. The issuers most exposed are the small ones — Simplify, Tidal (Return Stacked), Brookmont, Man, iMGP — which is most of the diversifier shelf. I could not read Fidelity’s live list (HTTP 403); verify before publishing. Sources: RIABiz 2026-05-07, ETF Investments.

C-4 — docs/research/current-regime-and-pricing.md §1.3: “No term-premium estimate is supported by this cache” is wrong. Two estimates exist and agree. Kim–Wright 10-year term premium = +0.8393% at 2026-08-14, from FRED THREEFYTP10 — 9,555 daily rows from 1990-01-02, fetched 2026-08-23, in the same FRED contract the page already uses. NY Fed ACM ≈ +0.80% at 2026-08-13. Both are materially positive for the first time in about five years, which means the page’s +0.82 pp 10y−3m slope is nearly all term premium — a conclusion it currently declines to draw. Change the label from unresolved to measured.

C-5 — docs/research/structural-and-tax-edges.md and src/content/placement.ts: the mandatory Roth catch-up rule for 2026 is missing, and it shrinks the account the trend wrapper is supposed to live in. SECURE 2.0 §603 is effective for taxable years beginning in 2026: per IRS Notice 2025-67, the §414(v)(7)(A) threshold rose from $145,000 to $150,000, and a participant whose 2025 FICA wages from the sponsoring employer exceeded $150,000 must make every 2026 catch-up dollar as designated Roth. That removes $8,000–$11,250/yr of pre-tax capacity for a high earner over 50. capital-efficiency-and-breadth.md places the stacked wrapper in the pre-tax account and already warns that a 30% wrapper consumes nine tenths of a one-third pre-tax account; this rule makes that account grow more slowly, every year, for exactly this investor.

C-6 — docs/research/structural-and-tax-edges.md: the §415(c) limit is not recorded. $72,000 for 2026, up from $70,000 (Notice 2025-67, line 21–22). It is the ceiling on the mega-backdoor Roth and the repository discusses shelter capacity extensively without carrying it.

C-7 — docs/research/alternative-sleeves-audit.md §3: “positive in 1 of 13 worst-decile equity months” should be 2 of 12 on data through 2026-08. On the 120-month CBBTCUSD/SP500 panel I computed 2026-08-23, bitcoin was positive in 2 of the 12 worst equity months (2019-05 at +62.2%, and 2026-03 at +2.2% against the S&P 500’s −5.1%), with a mean of −5.0%, not −7.51%. The 2026-03 observation is new and it is a point in bitcoin’s favour. The verdict does not change — but a page that measures this precisely should carry the updated count. Related: the 3-year daily correlation is now +0.170, the lowest on the panel, which is below the page’s own +0.2 reopening threshold; the trigger is still not met because the window contains no recession and the 3-year monthly reading is +0.363.

C-8 — docs/research/alternative-sleeves-audit.md §5: the cat bond ETF is now $88.44m, and that is the only figure in that section this scan changes. AUM went ~$12m (Aug 2025) → $25.3m (2025-11-26) → $88.44m (2026-08-22), against the audit’s $88.2m at 2026-08-20 — two readings two days apart, both current for their date. Average daily volume was up ~672% Sept–Nov 2025. The ticker is ILS, which is what the audit already uses; ROAR was pre-launch coverage only and is worth recording just so a reader who meets it can place it. The fee is the audit’s and is not restated: 2.65% gross, 1.58% net under a cap to 2027-04-30, and 2.00% actually paid. The price objection is untouched and the reopening condition is still a spread level, but “retail history is short” is nine months shorter. Sources: ETFGI, Artemis, StockAnalysis.

C-9 — src/content/shelf.ts carries no AUM field and no inception date field. For a shelf whose own live-managed-futures.md measures 52% attrition of the 2019 managed-futures cohort inside 6.5 years, size and age are the two variables most predictive of the failure mode the repository has actually documented. Four shelf funds are under $50m: MATE $39.76m, SDMF $39.16m, JPFP $32.75m, and (off-shelf but discussed) ILS $88.44m. A verdict that a fund is a candidate should print its size.

C-10 — IVLU and EFV: the shelf’s 31 bp is right for both, and an aggregator reading is not. Read from the funds’ own summary prospectuses, both dated 2025-11-28, on 2026-08-24: IVLU files 0.30% management, 0.01% other expenses, 0.31% total (Form 497K); EFV files 0.31% management, 0.00% other expenses, 0.31% total (Form 497K). Neither has a waiver line. A 0.30% quote for IVLU is its management fee with the other-expenses line dropped, and 0.33% is not a number in EFV’s fee table at all. Do not amend the shelf on either. The general lesson is worth more than the two numbers: an aggregator’s “expense ratio” and a fee table’s total annual fund operating expenses are different fields, and they diverge by exactly the other-expenses line. Every other fee among the 59 verified exactly.

C-11 — src/content/shelf.ts: three funds carried null where a public fee exists. All three were read from the funds’ own summary prospectuses on 2026-08-24 and are now on the shelf. GWX = 0.40% total, no waiver, per its 497K dated 2026-01-31 (filing) — the registrant now styles it State Street SPDR S&P International Small Cap ETF. DFEV = 0.46% gross less a 0.03% waiver = 0.43% net, the waiver running only to 2027-02-28, per its 497K dated 2026-02-28 (filing); a bare 0.43% hides an expiry date. TIP = 0.18% total, per its 497K dated 2026-02-27 (filing). One arithmetic loose end travels with the last of these: TIP’s netCostBp is 17.92 against an 18 bp fee, which implies 0.08 bp of securities lending nobody has read. The gap is the size of the unread term, not a measurement of it, and securitiesLendingBp stays null rather than carrying a number obtained by subtraction.

C-12 — docs/research/current-regime-and-pricing.md §1.7: gold is at the 98.5th percentile of its real price and 17% below its own January high. Gold futures set a record $5,542.40 on 2026-01-29; spot was ~$4,587/oz on 2026-08-21 (Fortune). The page’s percentile is correct and its price is correct; the drawdown is missing, and “expensive and falling” is a different sentence from “expensive.”

C-13 — Nothing in the repository carries a dollar level, and a tenth of 2025’s international outperformance was the dollar. Broad trade-weighted dollar index (FRED DTWEXBGS) = 118.9028 at 2026-08-14, −1.44% over one year, −0.97% over three, at the 57th percentile of its ten-year range (min 106.49, max 130.04), computed 2026-08-23. MSCI EAFE beat the S&P 500 by ~13 pp in 2025 and roughly 10 pp of that was currency. currency-and-the-international-sleeve.md should carry the level and the percentile, because the 2025 result is unreadable without them.

C-14 — docs/research/valuation-and-the-allocation.md should carry the fact that forward-looking models have already collected the international gap. Vanguard’s VCMM at 2026-06-30 puts US equities at 4.2–6.2%, developed ex-US at 4.5–6.5%, and emerging markets at 2.0–4.0% — EM below the US, cut by 1.6 pp in one quarter after EM returned +33.6% in 2025 (Vanguard VEMO, read 2026-08-23). A 1.70× valuation gap and a 0.3 pp expected-return gap are both true, and the site rewrite must not present the first as though it were the second. This supports the page’s existing conclusion; it is missing evidence, not a contradiction.

C-15 — docs/research/factor-products.md and the managed-futures shelf: FFUT is missing and it is now among the largest funds in the category. Fidelity Managed Futures ETF, inception 2025-06-05, $346.23m (etf.com, StockAnalysis). Its current fee table, read from the summary prospectus dated 2026-05-30 on 2026-08-24, is 0.80% management plus 0.02% other expenses = 0.82% gross, less a 0.02% reimbursement, for 0.80% net (Form 497K). The 0.80% is an expense cap running to 2027-05-31, not a plain waiver, and FDS may recoup within the fiscal year anything it reimbursed. Two net-asset figures for this fund are in the repository and both are current for their date: $255.9m at 2026Q2 (trend-marginal-value.md, second census table) and $346.23m at 2026-08-22. Category context the shelf also lacks: managed-futures ETFs total roughly $6bn across ~13 funds as of March 2026, of which DBMF alone is $4.0bn — so the rest of the shelf’s trend funds are competing for about $2bn, in a category whose documented failure mode is closure.

C-16 — the crypto table’s fees did not move, and the aggregator readings that said they had are wrong. StockAnalysis showed HODL at 0.25% and EZBC at 0.29% on 2026-08-22. Both trusts’ own Q2-2026 Forms 10-Q, read on 2026-08-24, say otherwise. HODL’s Sponsor Fee is 0.20%, and the filing states the waiver of that fee on the first $2.5bn ran “from November 25, 2024 through July 31, 2026” and that “[a]fter July 31, 2026, the Sponsor Fee will be 0.20%” (10-Q filed 2026-08-13) — so the waiver expiry the audit already flagged takes the fee from an effective zero to 0.20%, not to 0.25%. EZBC’s Sponsor fee accrues “at an annualized rate equal to 0.19%” (10-Q filed 2026-08-14). The audit’s table stands unamended. What is genuinely new is IBIT at $58.78bn against the $43.4bn the audit records at 2026-06-30 — a different date, not a different fact, and mostly mark-to-market. The rule this entry earns: a 1933-Act trust’s fee is in its own 10-Q, and a sponsor-fee waiver is the field an aggregator is most likely to get wrong.

C-17 — src/content/shelf.ts name fields now embed a benchmark-provider brand that changed once and could change again. The “Vanguard Morningstar …” names are correct — Morningstar completed its $375m acquisition of CRSP in February 2026, Vanguard announced the renaming of 13 US equity index funds on 2026-04-29, and it took effect 2026-07-29 with no change to methodology, holdings, fee or ticker (Vanguard, Morningstar). This is not an error to fix; it is a note to add, because a reader will assume it is one.

C-18 — Structural change the repository has not registered: Dimensional can now issue ETF share classes of its mutual funds (SEC exemptive order, 2025-11-17; first active manager, second ever). Substantially identical applications from other managers will follow. If a DFA mutual fund gains an ETF share class, its capital-gain distributions fall toward the ETF’s, and the ETF-versus-mutual-fund placement arithmetic in structural-and-tax-edges.md and placement.ts becomes an empirical question rather than a structural one. No shelf fund is affected yet. Sources: Dimensional, 40 Act Blog.

C-19 — Open item with a date attached: JPFP’s first Form N-PORT is still not filed as of 2026-08-23. EDGAR shows J.P. Morgan Exchange-Traded Fund Trust (CIK 0001485894) filing N-PORT-P batches on 2026-06-26 and 2026-07-28, with filings through 2026-08-18 and no JPFP series report. capital-efficiency-and-breadth.md’s stated due dates of 2026-08-29 or 2026-09-29 stand, and the earlier is six days away. delta remains uncomputable. This is the cheapest open item in the corpus.

---

What could not be verified

Stated plainly so nothing here is read as more complete than it is.

1. Fidelity’s live service-fee ETF list — fidelity.com/trading/etfs/service-fee-etfs returns HTTP 403 to automated fetches. Whether any shelf issuer is on it is unknown, and it is the highest-value open question in this scan.
2. A fresh literature sweep on valuation timing (Goyal–Welch–Zafirov II; Siegel; Straehl–Ibbotson). The session’s web-search budget was exhausted. The repository’s own §2.3 and §2.4 cover this ground and nothing found here contradicts them.
3. GMO and Research Affiliates current CMA vintages — GMO’s public forecast page returned 404. Only Vanguard’s June 2026 VCMM running was obtained.
4. An exhaustive census of buffered / defined-outcome / derivative-income launches since mid-2025. Not attempted at depth. The absence of a finding in §2.4 is a coverage gap, not a null result.
5. NY Fed ACM term premium from the NY Fed’s own file — read from aggregators citing the release, not from the primary CSV. Medium confidence. Kim–Wright is primary.
6. 2026 crypto market-structure legislation and any wash-sale extension to digital assets. Confirmed the OBBBA made no change; did not confirm nothing else did.
7. Bid-ask spreads for KMLM, MATE, JPFP, NTSX, GDE, ILS, and the Avantis and Dimensional funds — their issuer pages did not yield the Rule 6c-11 field to automated parsing. Every one of these publishes it and it is worth collecting properly.
8. The six Return Stacked funds’ and FFUT’s holdings. Their fees and sizes are filed or issuer-published and are now on the shelf; not one of them has had a Form N-PORT read here, so none has a computed delta, and RSIT’s “75% SPDW + 25% MSCI EAFE futures” is an issuer-page description rather than a filed measurement. No structure claim about any of the seven is verified.