EXTERNAL REVIEW RESULTS — TRIAGE AND INTEGRATION

Seven Reviewer Threads (Qwen, plus six others), Scored and Resolved

Regenerative Toronto | June 30, 2026

Document ID: MAIN-REVIEW-002 What this is: Seven external review documents were submitted in one batch — one structured PDF report and six freeform text reviews (apparently three separate models, each running multiple "continue" passes). Combined they raised roughly 80 distinct points. This document scores each major one, separates genuine findings from reviewer errors, and records what was actually changed. Methodology: High-stakes claims (potential fabricated citations, claims affecting an active policy recommendation) were independently re-verified against primary sources by Claude before any action was taken — not just trusted at face value. Lower-stakes claims were checked against the library's existing content for internal consistency. A meaningful number of reviewer claims were themselves wrong, including one reviewer thread that visibly garbled its own arithmetic mid-document and another that initially claimed a citation was "fabricated" before correcting itself two messages later. This is noted explicitly where relevant — the review batch should not be treated as uniformly reliable just because it's external.

TIER A — CONFIRMED REAL ISSUES, INDEPENDENTLY VERIFIED, FIXED

A1. Woodhall-Melnik 2025 citation — mischaracterized, specific statistics unverifiable (CONFIRMED, FIXED)

Multiple reviewer threads flagged this as a possible fabrication. Independent verification: the DOI is real and the paper exists — Woodhall-Melnik, Reiser, Monette & LeBlanc Haley (2025), "Housing First for Youth Who Experience Homelessness: A Systematic Review," Urban Affairs Review 61(5), 1381-1415 — confirmed via ResearchGate, CoLab, and the lead author's own institutional publication list. But it is a systematic review of seven other studies, not itself a randomized controlled trial, and its own abstract states findings are preliminary: "HF4Y can improve housing security, but additional research is needed to develop a more robust evidence base." The specific "71.8% vs 35.6%" point statistics this library attributed to it could not be located in any indexed source describing this paper. Verdict: not fabricated, but mischaracterized as an RCT with unverifiable specific figures attached. Removed the specific statistics and the "RCT" framing everywhere it appeared (BRIEF-001, MAIN_REF_002_Sources, DERIV-005, HMK-010, HMK-013) — replaced with an accurate description of the paper as a systematic review with cautious conclusions.

A2. Stergiopoulos 2015 "73% vs 32%" — real and correctly cited, but framed as Toronto-specific when it's pooled (CONFIRMED, FIXED)

One reviewer thread initially called this fabricated, then corrected itself to claim it was a "Winnipeg+Montreal data mashup misattributed to Toronto" with a specific site breakdown. Independent verification: the DOI (10.1001/jama.2015.1163) is real, the paper exists, and CAMH/EENet — a knowledge-mobilization source written by people directly affiliated with the original research team — confirms 73%/32% as the headline result of the study, with no site-by-site breakdown presented that way. The reviewer's specific claimed site numbers could not be independently confirmed. Verdict: the figure is real and correctly sourced — the actual problem is that this library's framing ("the Toronto At Home/Chez Soi RCT") implies Toronto-specific data when the JAMA 2015 paper is a 4-city pooled result (Vancouver, Winnipeg, Toronto, Montreal) that included a Toronto site, not a Toronto-only finding. Fixed the framing in all locations to describe this accurately as the multi-city pooled result.

A3. Modular housing $309,000/unit — real figure, but it's the actual cost of a program that went 63% over budget, not a clean benchmark (CONFIRMED, FIXED — this was the single most consequential finding in the entire review batch)

Independent verification via the City's own June 2023 Auditor General audit ("City of Toronto's Modular Housing Initiative: The Need to Balance Fast Delivery with Stronger Management of Contracts and Costs"): the original Council-approved budget was $190,000/unit (250 units, $47.5M total). The program was revised to 275 units (adjusted budget ~$52.3M) and came in $33 million (63%) over that adjusted budget — consistent with the $309,000/unit actual-cost figure this library cites (≈$85.3M ÷ 275 units). The AG found "insufficient planning," site preparation costs excluded from the original budget, and change orders approved without proper review. This means REC M-08 was citing the over-budget actual cost of a program with a documented City-auditor-confirmed management failure, as if it were a clean cost-saving benchmark — exactly the kind of "true fact, vulnerable framing" issue this library has tried to systematically avoid elsewhere. Fixed: REC M-08 in POLICY-001, the comparison table in HMK-012, and ledger entries CL-225/CL-226 now state plainly that the $309K figure reflects a 63%-over-budget actual cost with documented AG-identified causes, and that the City has since self-assessed 15 of the AG's 20 recommendations as implemented (per a January 2025 status update) — meaning the recommendation is now framed as "modular housing can be cost-effective if the AG's oversight fixes are sustained," not as an already-proven cheaper alternative.

A4. CL-149 overdose call percentage — arithmetic error (CONFIRMED, FIXED)

Three separate reviewer threads independently caught this. The ledger claims "160 calls/month → 350 calls/month; 82% increase." The actual math: (350-160)/160 = 118.75%, not 82%. This is a confirmed, simple arithmetic error, not a sourcing problem — the underlying call-volume figures may well be correct, but the stated percentage doesn't match them. Fixed: corrected to the accurate 119% figure (rounded), flagged for the original CBC source to be re-checked in case the 82% figure was meant to describe a different pair of numbers that got mismatched with these specific figures during drafting.

TIER B — REAL ISSUES, PARTIALLY ADDRESSED OR FLAGGED RATHER THAN FULLY RESOLVED

B1. 1990 Ontario minimum wage rate ($5.40 vs. $5.00) — UNRESOLVED, downgraded honestly

Multiple reviewer threads disputed the $5.40/hr figure used for the 1990 OW-to-minimum-wage ratio (currently sourced to "Feed Ontario 2019," yielding ~70%). I could not independently confirm the exact 1990 Ontario rate — available historical tables kept returning other provinces or incomplete decades, and the official Government of Canada historical minimum wage database requires an interactive form submission I cannot execute. Rather than guess in either direction (the reviewers themselves didn't agree on $5.00 vs $5.40 across different threads, and one reviewer's own recalculation of "74.8%" using $5.00/hr doesn't actually compute to 74.8% by their own stated method — see Tier D), this claim is downgraded from VERIFIED to VERIFY status in the ledger pending direct confirmation against Ontario's official ESA historical rate table. The ~70% figure should be treated as approximate, not precise, until resolved.

B2. $6,311 Housing First cost — needs consistent net/gross labeling (PARTIALLY ADDRESSED)

Multiple threads correctly noted that $6,311/year is the net cost after a 69% offset, not the gross program cost (~$21,000-23,000/year for high-need participants). Checking my own documents: HMK-012 does label this "ACT net annual cost" in its summary table — and the underlying Latimer cost paper is specifically titled "...Housing First With Assertive Community Treatment," so the ACT label itself appears correct, contrary to one reviewer's claim that this should be relabeled ICM. The real gap is that "net cost" doesn't consistently carry an explicit gross-cost comparison wherever it's cited, which matters because a critic could otherwise imply the program is being undersold as nearly free. Added an explicit gross-cost figure (~$21,000-23,000/yr) alongside the net figure in HMK-012's primary table and BRIEF-001.

B3. "78% chronically homeless" — real figure, real definitional ambiguity (FLAGGED, not changed)

One reviewer thread did the work of actually locating the source text: TSSS's 2025 Budget Notes state "approximately 78% of people accessing Toronto's shelter system have stayed in the shelter system long enough to meet the chronically homeless criteria" — confirming the figure is sourced, not invented. But this appears to use an internal/simplified threshold rather than the CMHC/HIFIS national definition (which requires 3 consecutive years of data and produces a much lower 30.2% national figure). This is a genuine definitional-precision issue, not a factual error — the 78% figure is real and TSSS-sourced, but should be presented as "TSSS's own internal chronic-homelessness threshold," not implied to be directly comparable to national CMHC statistics. Added this clarification to HMK-001 and BRIEF-001.

B4. Latimer $53,144 — possible city mislabeling (FLAGGED FOR FOLLOW-UP, not changed yet)

A reviewer thread claimed the Latimer 2017 CMAJ Open paper reports city-specific societal costs (Vancouver $53,144, Toronto $58,972, Winnipeg $45,565, Montreal $56,406, Moncton $29,610), and that this library may be using Vancouver's figure while discussing Toronto. This was not independently verified this session — it's a specific, checkable claim that deserves a dedicated verification pass before either keeping or changing the $53,144 figure. Flagged in MAIN_TRACK_001_Master_Issues.md as a new Tier 1 item rather than acted on without confirmation.

TIER C — REVIEWER FINDINGS THAT WERE ALREADY ADDRESSED IN THE EXISTING LIBRARY (no action needed, reviewers missed existing content)

C1. "21% drop in homeless count" framed as a damaging contradiction

The Qwen PDF report treats the 21% decrease (15,418 → ~12,180) as an unexplained, potentially fabricated statistic that undermines the "escalating crisis" narrative. This is incorrect — the figure is real, correctly cited (City SNA page, April 2026), and DERIV-001 already contains a fully-built-out honest rebuttal to exactly this weaponization risk, explicitly stating "the campaign's response is not to dispute the 21% figure — it is real," then explaining the refugee-claimant-decline driver and citing Ontario's province-wide 7.8% increase in the same period as context. No action needed — the reviewer did not have access to DERIV-001 or did not connect it to BRIEF-001's headline figure.

C2. IHAP funding "$97M" claimed factually incorrect

The Qwen PDF claims this confuses capital and operational funding, citing a $1.5B Canada-Ontario partnership figure. This appears to be the reviewer's error, not the library's — HMK-008's $97M figure is specifically and narrowly sourced to the City's own 2026 TSSS Budget Notes for the Interim Housing Assistance Program specifically (not general housing capital funding), with an explicit federal cost-sharing agreement (95%, Jan 2025-Mar 2027) cited. HMK-008 already contains the exact caveat the reviewer asked for: "the decrease reflects program restructuring, not federal abandonment." The $1.5B figure the reviewer cites appears to be for a broader suite of programs, not IHAP specifically — this needs the reviewer's own sourcing checked, not mine. No action taken; flagged as a discrepancy worth resolving only if a primary IHAP-specific source contradicts the existing TSSS Budget Notes citation.

C3. "No media outreach / journalist targeting plan"

Confirmed: DERIV-002 does not name specific journalists or a release timeline. This is a fair observation, not a reviewer error — but it's a strategic gap (Tier 4, explicitly held), not a factual or framing problem. Logged as a strategic suggestion, not acted on (see Tier E).

TIER D — REVIEWER ERRORS (the reviewer was wrong, not the library)

D1. One reviewer thread's own arithmetic visibly fails mid-document

In the "CLAIM LEDGER REVIEW" document, the reviewer attempts to recalculate the 1990 OW ratio live in its own text and produces an internally contradictory result ("6.23%? That can't be right... I think the intended calculation is... 71.9%... but the ledger says... 69.2%... that uses $5.40/hr... I need to confirm"). The reviewer's own item 1 does not reach a stable conclusion and should not be read as a confirmed correction — it correctly flags that something needs checking (see B1 above) but does not itself establish the right answer.

D2. "No theory of change," "no business case one-pager," "no post-election plan" presented as missing entirely

Several of these (message hierarchy, business case framing, post-election accountability) are reasonable suggestions but were presented by reviewers as complete blind spots without checking whether adjacent material already exists in a different form. This wasn't independently re-verified item-by-item this session given volume — treated as Tier E strategic suggestions rather than corrections, since "we should add X" is a different category of finding than "Y is factually wrong."

TIER E — STRATEGIC SUGGESTIONS (good ideas, not factual corrections, not actioned this round)

The review batch contained a substantial number of legitimate strategic suggestions that are not errors to fix but additions to consider: a federal IRCC processing-time recommendation, a message-testing/polling plan, a "First 100 Days" prioritized subset of the 32 recommendations, a business-case one-pager for fiscally conservative audiences, a "pre-bunking" document anticipating Ford government counter-claims, jurisdictional labeling of each recommendation by level of government, and tiering recommendations into "immediate" vs. "long-term." These are reasonable and several are good ideas — they are logged in MAIN_TRACK_001_Master_Issues.md Tier 4/5 as candidate future work, not implemented in this pass, consistent with this project's standing practice of not silently expanding scope into held/strategic territory without explicit direction.

WHAT THIS TRIAGE DID NOT DO

Given the volume (~80 distinct points across 7 threads), this was not a line-by-line re-verification of every claim. Items not explicitly addressed above received a lighter-touch internal consistency check rather than independent primary-source re-verification. If any specific remaining item matters enough to warrant the same depth of checking given to Tier A/B above, flag it directly rather than assuming silence means it was cleared.

MAIN-REVIEW-002 | Version 1.0 | June 30, 2026 | Regenerative Toronto