Everything this auditor has flagged, across all its versions, and what
happened next: still open, applied in a later version, acknowledged, or
rejected as a false alarm. A finding raised again by a later run is
superseded by the newer one, so each problem is counted once.
91 findings: 64 open, 26 applied, 1 acknowledged. 3 repeats were superseded by a later finding.
Still present in version 4: Skeptic's check "Strongest objections answered" found it again (audit of 13 October 2026). See feedback 2516104304975577084-6627e0e44f48002d6489c7755053bbf2.
Still present in version 4: Skeptic's check "Confidence matches the evidence" found it again (audit of 13 October 2026). See feedback 2516104304977799079-9ae8f1385f7a6b295b448970089565d1.
Still present in version 4: Skeptic's check "Rates the real claim" found it again (audit of 13 October 2026). See feedback 2516104304980894134-9d876c865e1865e64620d05a4e6ecedb.
Version 2 no longer says the high-school and Olympic margins match. It adds objections on the depth of girls' fields (high-school jump margins exceed Olympic ones, and the 4A pole vault figure rests on one outlier), on teenage girls' dropout from sport, and on comparing champions with average athletes.
Version 4 adds a 2006 randomized, double-masked trial (Owsley et al., IOVS) in which 30 days of high-dose retinol modestly sped rod-mediated dark adaptation in adults over 50, with no change in thresholds. It is added as an evidence item, a source and a new objection on speed of adaptation and older adults. The conclusion, key points and summary no longer say no evidence suggests a benefit beyond deficiency; they explain the finding as a local retinal shortage, note the high dose and that carrots were not tested, and keep the Misleading rating at Moderate confidence.
Version 4's response to the first objection now says the 1942 study is a weak test: many participants showed signs of deficiency and the authors doubted their equipment was sensitive enough, so its null result is close to uninformative about well-fed people, and the well-nourished case rests mainly on physiology rather than trials.
v3 folds the National Eye Institute assessment into the mechanism evidence, whose limitation now says its three U.S. sources are one line of evidence. The old e4 became a distinct item on who is deficient, which discloses its reuse of the NIH fact sheet, and method step 2 notes shared sources.
v3 adds an objection, a key point, conclusion text and an evidence item on people with malabsorption (cystic fibrosis, liver disease, bariatric surgery) or marginal status, citing Norsa et al. 2019 (Nutrients). It notes no study tests carrots in marginal status, and that the rating is unchanged.
A further search found no modern trial in well-nourished people, so v3 lowers confidence to Moderate, explains why in the conclusion and method, calls the 1942 study a weak test, and softens the title and answer to "no boost shown otherwise". The Misleading rating stands.
Version 3 rewrites the summary, and the context and a new conclusion paragraph say that the count, rent and other-region evidence all rest on the one Point-in-Time series, while the bed inventory and shelter-performance data are separate collections that show system strain, not the number of people homeless. The rents and other-regions limitations say the same.
(a) Version 3 gives King County's unsheltered-rise rank under each adjustment: 4th of 67 as counted and with the central estimate, 21st at the high estimate. It says the total rise at the high estimate (14%) is below the peer median and qualifies "did worse than most of its peers" in the conclusion. (b) Counting transitional housing too (HUD measure 1a, ES+SH+TH), the median stay still rose from 50 to 145 days (60th of 67), so moving beds from transitional housing into shelter does not explain the rise. This is now in the shelter-stays evidence and the conclusion.
Version 3 drops "the rise holds under every method" and "roughly doubled". It names the 2015, 2017 and 2022 method changes and shows that the count did not fall within any span counted the same way: 8,899 to 10,730 (2012-2016), 11,643 to 11,751 (2017-2020) and 13,368 to 18,365 (2022-2026). The conclusion keeps High confidence for the verdict and adds that confidence is lower for the size of the rise.
Version 4 no longer calls NASBO independent. Its evidence title, conclusion, method, source roles and source description now describe it as a national compilation of state-reported figures on a different (fiscal-year expenditure) basis. The budget-doubled limitation notes that evidence 3-5 analyse the same appropriations series, so they test the size of the increase rather than independently confirm it. The hand-written count of "five independent record-keepers" is removed.
Version 4 rates the claim True. Every measure within the claimed decade more than doubled: final appropriations 114.6%, enacted budgets 107.6%, all funds 106.9%, NASBO 121.5%. The budget also outgrew personal income over the decade (114.6% vs 90.3%). The 2005-07 income comparison and the post-recession-trough argument stay in the conclusion and objections as context for what the doubling means, not as a reason to lower the rating.
Version 4's context and method explain the scope: the Near General Fund is measured because it is the part of the budget paid from general state taxes and the figure both sides cite. Because a reader may take "state spending" to mean the whole budget, version 4 also reports the all-funds operating budget, which rose 106.9% ($68.8 billion to $142.4 billion), so either reading supports the claim.
