# Mentor Feedback — gld-nav-20260625T1600-1782417600-review.md **File:** `03-research/event/gld-nav-20260625T1600-1782417600-review.md` **Date:** 2026-06-26 **Reviewer:** macro-mentor (auto-review) **Type:** New file (full add). Structural review of GLD NAV action Thu Jun 25, 2026. **Status check:** prepopulated, source = yfinance daily closes. ## What's working - **Cross-asset panel is well-constructed.** Spot, ETF (GLD, IAU, SLV), futures (SI=F, HG=F), risk proxies (SPY, TLT, DXY, 10Y) — that's the right instrument set for a GLD review. The silver ETF/futures split is the right thing to surface; most reviewers would have skipped it. - **The 4-day and 7-day trend columns** give the bounce narrative context. Without them, "first up day" reads as noise. - **The "what's unusual" filter** is the strongest section — it tells you which of the 10 cross-asset datapoints actually matter. That's the right discipline. Per rubric §Daily Journal / Flows: "What the market is doing, not what journalists say." - **Rates-backdrop honesty.** "Bond market isn't validating the gold bounce" — this is exactly the kind of cross-check the rubric wants. You don't gold-bid without bond backing without naming it. ## Per-section critique ### Frontmatter - **type: "structural-review"** — not a defined type in `process.md` §6 (published research format) or the rubric. Event files per the rubric are "Event Preview" (pre-event) or this looks like a post-event review. Add `type: event-review` or align with the published-formats list. WHY: file-type drives routing/templating; ad-hoc types fragment the search/audit chain. - **Missing `linked-trades` field.** If you have an open GC/SI thesis in `02-trades/`, this review should reference it. WHY: per rubric §Daily Journal / Positions: "Each position stress-tested against the macro read" — without the link, the review is data only, not decision-grade. - **Missing source timestamp granularity.** `source: yfinance daily closes` — fine, but `10Y yield (^TNX)` row admits "last available" which is sloppy. Either cite the timestamp of the 4.392% print or don't include it. ### Headline - **Reads as data, not a frame.** Per rubric §Daily Journal / Macro (your read): "A frame, not news. Names a regime or transition, not just direction. Identifies the dominant driver." Current headline is three numbers strung together. - **Suggest rewrite:** "First up day in GLD after a 3-day -5.5% rout is positioning exhaustion, not thesis change — copper's +3.2% confirms flow-driven sell, not a fundamental shift. Bonds didn't validate (10Y -1bp), so the bounce has no sponsor yet." WHY: a headline that names the regime/transition forces the rest of the doc to defend it. - **Missing the dominant driver.** What *is* driving the GLD bid right now? Real rates (dormant per §Rates backdrop), USD (flat), risk-off impulse (SPY +0.14% says no). The dominant driver is "short-cover in metals," and the headline should name it explicitly. ### Price action table - **Solid.** The NAV discount row is the right field to compute — it's the only flow signal in the table. Keep it. - **Add the 5-day range band for the discount, not just the endpoint.** You cite "-0.36% to -1.02%" in the prose, but the table only shows -0.92%. WHY: makes the "wide end of range" claim auditable from the table itself. ### Cross-asset table - **Good coverage.** One omission: GLD volume or options skew if available. The SLV ETF/futures divergence is interesting but without volume context, "creation/redemption flow" vs "paper short selling" is a coin-flip. WHY: divergence without volume is anecdote, not signal. - **DXY at 101.439** — fine, but you call out "macro backdrop basically unchanged" elsewhere. Be consistent. If DXY is unchanged, why is it in the table? Either commit to its signal value or drop it. ### What's unusual - **Best section in the file.** Honest cross-asset calls. Two issues: 1. **No position implications stated.** Per rubric §Event Preview: "Position implications (long/short bias, convexity, time horizon)." If this is a *review* format (post-event), then per rubric §Daily Journal / Positions: "Per-position one-liner. Each position stress-tested against the macro read." Neither is here. 2. **"Historically these track within ~10 bp on a daily basis"** — source? You cite no historical anchor for the SLV/SI=F gap claim. Per `process.md` §6 published research format: "Source citations for any non-original data points." This is a non-original claim. WHY: a fact without a source is an opinion; an opinion doesn't grade well in postmortem. - **GLD discount "mild outflow signal"** — this conflates two distinct things. NAV discount can widen from (a) genuine outflow pressure, (b) authorized-participant creation/redemption lag, (c) intraday ETF pricing vs end-of-day NAV calc. Without volume or AP data, you can't distinguish them. WHY: calling it "mild outflow" without that disambiguation is anti-pattern #4 (narrative without pricing) in micro form. ### Rates backdrop - **Honest read.** "The 'real rates up = gold down' mechanism is currently dormant" — good, explicit mechanism check. - **Missing: real yields, not just nominal.** 10Y nominal at 4.392% says nothing without breakevens or TIPS. Per `process.md` §2 coverage universe, "real yields" are in the monitored-not-traded set. WHY: gold is a real-rate asset. A rates section that only quotes nominal is incomplete. ### Implications - **Three problems, all fixable:** 1. **No invalidation fact for the bounce thesis.** The whole section argues "positioning exhaustion, not thesis change" — that's a view. What's the fact that kills it? Per `process.md` §4.1 and `anti-patterns.md` §1, invalidation is a required field and must be a fact, not a price. Suggest: "If 10Y closes above 4.50% AND DXY breaks 102 on the same session, the bounce thesis is dead." 2. **No base/alternative/tail scenarios with magnitudes.** Per rubric §Event Preview: "Base / alternative / tail scenarios with magnitudes." Current section has 4 conditional branches (copper holds / gives back; SLV/SI converges; etc.) but no probability or magnitude attached. 3. **No time horizon.** Per rubric §Trade Idea: "expected holding period." If the bounce view is on, what's the window? 1-3 sessions? Through Fri payrolls? Through next FOMC? - **Watch-list levels are good but underlinked.** $365 line-in-the-sand, $377 over-and-done, $387 next resistance, $360/$355 next downside. These should be the levels in an *active* thesis — but no thesis ID is named. WHY: levels without a thesis are chart-watching, not trading. ## What to change (concrete) 1. **Add `linked-trades: []` to frontmatter.** If a GC or SI thesis is open, name it. If not, that's a signal — write a one-line "no open position; watching $X for entry on Y fact" in §Implications. Per `anti-patterns.md` §5 (confusing activity with progress): documenting the "no trade" decision is a feature, not silence. 2. **Rewrite headline as a frame.** Lead with the regime call, not the price. Per rubric §Daily Journal / Macro. 3. **Cite source for the "10 bp historical SLV/SI=F tracking" claim.** If you don't have one, drop the historical anchor and keep the observation ("1.5% gap is large"). 4. **Add base/alt/tail scenarios to §Implications.** Per rubric §Event Preview format. Magnitudes matter. 5. **Add a real yields line to §Rates backdrop** — TIPS or 10Y breakeven. Otherwise the "real rates up = gold down" framing has no anchor. 6. **Add invalidation fact** to §Implications for the bounce view. Per `process.md` §4.1. ## Process improvements (filed where) - **Add a `_templates/event-review.md` template.** `process.md` §6 doesn't define the event-review format (only event-preview). The rubric is silent. The vault has at least one event-review already (this file). Template the format so future reviews carry frontmatter (linked-trades, source timestamps, real yields row) by default. File under `_templates/`. - **Schema-check frontmatter type field against `process.md` §6 published research formats.** Either expand §6 to include "structural-review" / "event-review", or rename to fit. Otherwise the vault accumulates ad-hoc types. File under `90-archive/process-changelog.md`. - **Coverage-universe verification: GLD/IAU as proxies.** Per `process.md` §2: "Commodities: GC, SI, HG, CL." GLD/IAU are ETF wrappers, not the underlying future. The cross-asset panel is fine for monitoring but if you write a GLD thesis, you trade GC. File a one-line note in `06-universe/traded/GC.md` (or create it if missing) clarifying GLD/IAU as monitor-only proxies for GC. ## Rubric / process references - `process.md` §2 (Coverage universe — GLD/GC distinction), §3 (Tier 1/2 event coverage — Tier 1 includes FOMC/CPI but not daily NAV reviews), §4.1 (thesis required sections — invalidation is a fact), §6 (published research format — no event-review type defined) - `risk-framework.md` §1 (25-75bp band — relevant if/when a bounce trade is sized) - `anti-patterns.md` §1 (no invalidation — applies to the bounce view in §Implications), §4 (narrative without pricing — applies to the SLV/SI=F "flow vs paper shorts" call) - `references/rubric.md` §Daily Journal (macro read = frame, positions = stress-tested one-liner), §Event Preview (position implications, base/alt/tail) - `references/feedback-examples.md` (tone reference for this feedback — direct, WHY on every critique) ## Bottom line Strong first pass on a structural review. Cross-asset panel and the "what's unusual" filter are the right instincts. The doc is data-rich but decision-light: no active-trade link, no invalidation fact, no scenario magnitudes, no regime call. Bring this to grade-B by adding the thesis linkage and the rubric-mandated fields — the scaffolding is there, the trading decision layer isn't.