R08 Correction Round — Gemini (paste-ready prompt)

Drafted by Claude (editor), 2026-08-31, from critique.md §2–§4/§6. Surgical scope per editorial consensus: provenance-or-withdraw only — no report rewrite, no regenerated bibliography.

Your R08 report has completed verification (46 sources, two-check per PROTOCOL v1.3). Your substantive contributions are strong and several will anchor the synthesis — the methodology→practice taxonomy, the qPCR-vs-TYMC regulatory asymmetry, the commercial-producer landscape, the HpLVd analysis, and the carbon-breakeven insight. But the verification record requires a correction round before any of your material enters the ledger.

Scope rule — read first: this is a surgical correction round. Do NOT rewrite the report. Do NOT regenerate or re-derive any citation. For each numbered item below, respond with exactly one of: (a) PROVENANCE — a resolvable identifier (DOI, stable URL, or named administrative dataset) plus where in that source the specific figure appears; or (b) WITHDRAW — an explicit withdrawal of the figure or attribution. Withdrawal is honorable and costs nothing; a second unverifiable citation costs another round. Every correction you supply will itself be verified at full burden before acceptance.

Part 1 — Identity corrections (already resolved by the editor; acknowledge, do not re-cite)

The following citations in your report failed identity verification. The editor has live-verified the corrected tuples; you need only acknowledge them. Do not cite the failed versions again:

1. Zager et al. → real: Zager et al. 2019, Plant Physiology 180:1877–1897, DOI 10.1104/pp.18.01506 (your DOI was Booth et al. 2020).
2. “Vergara et al. 2016, RSOS 160397” → no such paper exists. Withdrawn by the editor; chemotype-locus genetics needs a real anchor (state one with provenance, or leave the claim to the editor to re-anchor).
3. Lyu et al. 2019 → Frontiers in Microbiology 10:1761, DOI 10.3389/fmicb.2019.01761 — and note it is a perspective, not an experiment.
4. Saloner & Bernstein 2021 → article 113516 (your source list was right; your in-text 113554 resolves to a polysaccharide paper).
5. “Punja & Rodriguez 2018, DOI …1535469” → real paper is DOI 10.1080/07060661.2018.1535466 (root pathogens); your suppression claim belongs to Scott & Punja 2023, DOI 10.1080/07060661.2023.2172082.
6. Cranshaw & Wainwright-Evans 2020 → DOI 10.1093/jipm/pmaa008 (your DOI was a corn tar-spot paper).
7. “Fonteno 1993, HortScience 28(2):81” → fabricated tuple. Real: Acta Horticulturae 342:197–204, DOI 10.17660/actahortic.1993.342.22.
8. Bugbee 2002 → DOI 10.1080/1065657X.2002.10702069 — and note the paper’s finding (optimal growth at 50–100% biosolids compost) contradicts the threshold claim you cited it for.
9. McKernan 2016 → DOI 10.12688/f1000research.9662.1; McKernan 2021 → F1000Research 10:624, DOI 10.12688/f1000research.53467.2 (your “Data 6(8):80” is a metallurgy paper).
10. Minor: Geisseler & Scow pages 54–63; Summers DOI 10.1038/s41893-021-00691-w; “Bektaş et al.” is actually Atallah, Yassin & Verchot; Bilderback title “Healthy Substrates Need Physicals Too!”, pp. 747–751; the Rx Green trial authors are Wedryk, Wall & Bennett (no “Stemler”); the real Seemakram title is “Enhancement of growth and Cannabinoids content of hemp (Cannabis sativa) using arbuscular mycorrhizal fungi.”

Part 2 — Provenance-or-withdraw (one verdict per item)

Misattributed or contradicted findings:

1. Shannon-diversity comparison (living soil H′ 4.5–6.0 / coir 2.0–3.0 / rockwool <1.5; “2–3 orders of magnitude”) — appears in neither Winston 2014 nor Comeau 2020.
2. AMF P-suppression thresholds (30–50 mg/L pore water; Bray-1 30–45 mg/kg; strigolactone mechanism) — Treseder 2008 is an N-additions meta-analysis and contains none of this.
3. “No potency advantage for living soil” and “elicitation shifts terpene ratios 10–20%” as attributed to Bernstein et al. 2019.
4. “77 commercial grows, hydro 435 vs soil 315 g/m² (1.12 vs 0.81 g/W)” as attributed to Vanhove 2011 (a controlled factorial experiment).
5. “+18.4% to +24.1% mineral over organic” as attributed to Burgel 2020 (a substrate comparison).
6. “Potting media 65–75% of outdoor GWP” as attributed to Desaulniers Brousseau 2024 (paper reports 25–38%).
7. HpLVd magnitudes: “30–50% biomass / 20–50% cannabinoid loss” (Punja 2023 reports 12–42%) and “10–20% bleach (0.5–1.0% NaOCl)” (Punja 2025 reports 5–10% bleach / 0.825% NaOCl effective).
8. “Transmission primarily mechanical via tools and root grafting” — the Atallah paper’s headline finding is seed transmission at 58–80%.
9. Seemakram magnitudes: “+33% CBD concentration, ~+120% total CBD/plant under moderate nutrient regimes” — no moderate-nutrient arm exists in that study.
10. Kakabouki: “root surface area and total biomass increased” — the paper reports root biomass unaffected.
11. Mineralization rates: feather meal “60–75% over 8–12 wks” (source: ~56–66% in 8 wks); poultry manure “40–55% over 8–10 wks” (source: 41%/63% plateaus over 52 wks); blood meal “50–60% to NH₄⁺ in 7–14 days” (unverifiable at that precision).

Uncited quantitative assertions (currently inadmissible under failure-mode #4): 12. Canopy shares: 85–92% inert / 8–15% living soil / <3–5% true no-till. 13. Retail premium +10–25% for certified living-soil brands. 14. The cost/labor table: bed CapEx $450–1,200; labor index 1.25–1.5×; cycle cadence 4.0–4.8 vs 5.2–6.0/yr; 15–20-cycle payback; bed weight 44–69 lb/ft²; 50–66% canopy reduction under racking; 15–25% annual-output reduction. 15. TYMC failure rates “15–35% without post-harvest remediation.” 16. The multi-cycle succession table: cycle-phase boundaries, F:B ratio trajectory (0.4–0.8), bulk-density drift 0.15–0.35 → 0.40–0.60 g/cm³ over 5–10 cycles. 17. The N-scaling table: peak demand 0.4–0.8 g N/plant/day (and 0.5–1.2 g K); the 15/30/100-gal sufficiency verdicts. 18. Reset triggers: ECₑ > 4.5 mS/cm; Mehlich-3 P > 400 mg/kg; AFP < 10%; HpLVd > 10–15%; two-cycle yield depression 15–20%. 19. Cover crops adding “15–30% latent moisture load” indoors; irrigation set-points (60–120 mbar tension; 30–42% VWC; pore-water EC < 2.5–3.0). 20. Compost-fraction physical thresholds as stated (≤10–15% safe / 20% upper / >30% collapse, AFP and EC figures) — currently anchored only by the contradicting Bugbee citation; supply container-media literature that actually supports these numbers, or withdraw the thresholds and leave the property-based framing (which is retained via Carlile 2015). 21. Commercial-producer specifics: Coast Mountain “40+ consecutive cycles / 5+ years”; Rubicon/Simply Bare bed details; WMMC, Sweet Dirt, Habitat, Stewart Farms, GreenCare, Earthkeeper descriptions — public documentation URL for each claim you keep (press, certification registries, company pages all acceptable; they will be logged as industry/anecdotal, not scientific).

Part 3 — Missing brief-required sections (deliver now)

A. Candidate claims: 15–30 concise, testable statements with supporting citation(s), each carrying its evidence tier (T1–T4, ANECDOTAL flag where applicable) and the system definition (a/b/c/d) it applies to. Build them only on citations that survived verification or that you re-anchor in Part 2. B. Search-stopping rationale: what you searched, where saturation occurred, and which domains ended searched-and-empty vs. not-searched. C. Explicit data-gap list.

Respond in the numbered structure above. Items answered with anything other than PROVENANCE or WITHDRAW will be treated as WITHDRAW.

CORRECTION ADDENDUM — 2026-09-02

Item 6 in this round accused Gemini of misreporting “potting media 65–75% of outdoor GWP.” That accusation was wrong — Gemini’s figure was correct, and the verification’s substitute (25–38%) was the paper’s metal-depletion category. Reversed by cross-model audit (verification/audits/R08-audit-2026-09-02.md). Gemini’s R08 error tally drops by one.