Corrections
Changes to a figure, a label or a claim a reader could have relied on, grouped by subject, newest first, from the October 2026 review rounds on. Changes to earlier drafts are not listed. How to report an error.
46 corrections, 2026-10-25 to 2026-11-14, most found in the review rounds run on this prototype. 3 changed a headline conclusion, all about award #338 (marked “Conclusion changed”). Most of the rest are figures, labels and map drawing.
Award #338 and the break-even screen
| Date | Where | It said | It says now | Why |
|---|---|---|---|---|
| 2026-11-14 | Would it happen anyway? and the Triumph break-even card: discount rates | Every company deal shown at 4% real only on /triumph; on /but-for the 2% row read “OMB A-94 Appendix C, 30-year real Treasury” beside the 7% “OMB A-94 base case”, as if both were benefit-cost rates. | /triumph shows every company deal at 4% (the rate on Triumph's own score sheet #205) and at 7% (the federal base case for public investment, OMB Circular A-94 as reinstated on April 8, 2025), with the reason for each: 6 of 14 deals are above the 12.7% median on their agreement's timetable at 4%, 11 of 14 at 7%. The 2% row now says OMB limits Appendix C rates to cost-effectiveness analysis. | OMB memo M-26-09 says Appendix C rates do not apply to benefit-cost analysis of public investment, and a reader weighing the deals should see that the headline count depends on the rate. |
| 2026-10-25 | Home page walkthrough and award #338: the three sites | Treated all 500 jobs as arriving on the new hangar's timetable. | Says the 500 are counted across three sites, including 125 at an existing Lynn Haven building, and gives the bar if those start in the first three years: about 12.8–13.1%, level with the 12.7% median. | The agreement counts jobs at all three sites toward the 500; the Lynn Haven jobs need no new construction. |
| 2026-10-25 | Home page walkthrough and Would it happen anyway? (award #338): the research and the verdict | “Incentives tip between 2 and 25 of every 100 decisions”; at the 12.7% median “the timing decides it, not the research”; other assumptions move the bar “from about 3.4 to 55”. | “Typical incentives probably tip …”, as the paper says; at 12.7% the wages fall a little short on every reading of the agreement's deadlines, and a wage above the floor, wage growth or a lower discount rate could close the gap; the range is 4 to 55, on the same timetable /but-for uses. | The qualifiers matter for a $50,000-a-job grant; every reading of the deadlines sits above 12.7%, so timing is not what decides it; and two pages gave different ranges for the same deal. |
| 2026-10-25 | Award #338 project screen and brief | “Draw schedules are in disbursement requests, which are not posted”, while the home page cited the executed agreement's draw plan; the read-aloud summary named “Spending is our assumption” as the biggest open question. | The Spending row cites Exhibit B ($2.5M in 2025, $17.5M in 2026, $5M in 2027); for every company deal the biggest open question is whether the jobs would have come without the grant. | Two pages disagreed about the same document, and the summary led with a modelling input instead of the question the deal turns on. |
| 2026-10-25 | Home page walkthrough: size of the #338 incentive | The grant set against the research medians as 23% of the project's cost. | $50,000 a job, with the note that the medians are not scaled to incentive size. | Share of project cost is not the yardstick the incentive research uses; cost per job is closer to it. |
| 2026-10-25 | Home page walkthrough, awards pages and job-deal strip (award #338 and other company deals) | Each deal's timetable figure called “the slowest timetable its agreement allows”, with the real bar said to lie between the year-one figure and it; #338's planned bar put at 12.0–12.8% from the score sheet's draw plan. | One reading of each agreement's deadlines. For #338 other readings of the same deadlines give 13.0–15.0%, and the executed agreement's budget (Exhibit B: $2.5M in 2025, $17.5M in 2026, $5M in 2027) gives about 13.4–13.8%. | The screen counts ten years from the first job, so a later start can lower the bar as well as raise it; the timetable figure is not a bound. The executed agreement's Exhibit B replaces the score sheet's earlier draw plan. |
| 2026-10-25 | Home page walkthrough, awards pages and project screen (award #338) Conclusion changed | On the agreement's timetable the bar is 14.0%, so at the all-studies median (12.7%) the wages fall short of the grant; elsewhere, a single 10.3%. | Since replaced: see the deadline-readings entry above, for the 14.0% label and the 12.0–12.8% plan; the 10.3% figure and the coin-toss reading at 12.7% stay. | 14.0% is the slowest pace the agreement permits, not its schedule; the score sheet plans a faster one, and modest changes to assumptions the site itself offers move the 12.7% result either way. A single figure on either side overstated the precision. |
| 2026-10-25 | Home page walkthrough and Would it happen anyway? (award #338) Conclusion changed | The grant must decide 10.3% of deals like it, and at the all-studies median (12.7%) it clears the bar. | Since replaced: see the 10.3–14.0% band above. | The screen counted all 500 jobs from the year the grant was paid; the executed agreement allows four years to build and four more to reach 500 jobs. |
| 2026-10-25 | Would it happen anyway? (award #338) | Said the staff score sheet does not say whether its $19.20 is discounted, and that undiscounted wages times 1.6 give it. | Quotes the sheet: “discounted to present money value”; the stated figure is about 2.0 times the discounted promised wages. | That page built its reconciliation without the sheet's own words; the home page and award page already quoted them. |
| 2026-10-25 | Home page walkthrough | “11 of 18 company-deal score sheets don't ask” whether the plant would have come anyway; “30 of 87 awards have no partial-repayment formula”, set beside #338's safeguard. | 10 of 17 company deals' sheets do not estimate it; 16 of the 17 company deals have a partial-repayment formula, and the awards without one are nearly all training or public-works grants. | #209's amended sheet was counted as a second deal, and the safeguard comparison set a company deal against grants with no job target to prorate. |
| 2026-10-25 | Award #338 | Read the staff's $19.20 as suggesting “no discounting”. | Quotes the sheet's “discounted to present money value”. | The sheet says so; the earlier reading was an inference it contradicts. |
| 2026-10-25 | Home page walkthrough, award #338 Conclusion changed | Showed only the 3.4% research median and said the grant must “pay for itself”. | Since replaced: see the timetable and band entries above; the three medians stay. | Showing only the median least favourable to the deal was one-sided, and gross wages are not a net benefit. |
| 2026-10-25 | Home page walkthrough | “$860M: Florida's share of BP oil-spill economic damages”. | $860M paid to Triumph Gulf Coast so far (2017–2025), with $640M more scheduled through 2033. | $860M is the amount received to date, from the audited statements, not the total share. |
Triumph awards record
| Date | Where | It said | It says now | Why |
|---|---|---|---|---|
| 2026-11-14 | Triumph award #120 (Commercial Aircraft Maintenance Campus): the passed deadline | “Company contracts by 6/30/2025 (no later extension posted)”. | “City construction contracts signed by 6/30/2025 (no later extension posted)”. The scorecard's answer is unchanged: not known, as Triumph's tracker reports construction on track. | The deadline in the amendments is the City's, to sign its construction-manager and design-build contracts, not the company's. A fact-check for the new Before the vote page found it. |
| 2026-11-01 | Triumph scorecards: “Is it on schedule by Triumph's own tracker?” | “No” for 14 awards whenever a deadline in the posted documents had passed with no completion reported, including seven that Triumph's tracker reports on track or ahead (#69, #98, #120, #187, #213, #243, #251). | “No” only where the tracker reports the overdue part delayed (#148, #237); “Partly” where it reports another part delayed (#153, #230, #233, #312); “Not known” otherwise (#69, #98, #120, #187, #190, #213, #243, #251), with the passed deadline and the tracker's report both stated. County on-schedule counts change with it. | The site's own method says that what is not in the posted record never becomes “was not met”; an unposted amendment or report may have moved or met the deadline. The rule broke that. |
| 2026-11-01 | Triumph scorecards: “Does the agreement set a target that can be checked?” | “No: the posted agreement sets no target with a number and a date” for #48, #191, #213, #320, #324 and #333. | “Partly”: their only dated targets are completing the project, delivering reports or spending the money. #186 and #189, hurricane property-tax relief with no target at all, stay “No”. | Each of those six agreements does set dated targets, such as #48's “Substantially Complete by 12/31/2024”; the finding said it did not. |
| 2026-11-01 | Triumph scorecards and award pages: stated returns for #185 and #240 | “The posted sheets state no return.” | #185's sheet states $28.40 and #240's $13.00 in income per Triumph dollar, and the check reads them. #209's original sheet now shows its $7.20 as well; its amended sheet's $27.20 still leads. | The data build read a return only when the sheet gave a single figure; these sheets give several (the 10-year headline first). |
| 2026-11-01 | Triumph awards record: the total, and awards #187 and #233 | The 87 awards as amended came to $801.5M. #187 (Walton Works) was counted at $92,500, its Second Amendment pilot alone, so Walton's share of agreement dollars signed in 2024 or later read 81%. #233 (IHMC) was $20,723,761, “grown through four amendments”, with a total project cost of $110,472,324. The reconciliation with Triumph's $829.1M list called the remaining gap “change from amendments”, largest #187 at $2.9M to $0.1M. | $804.0M. #187 at $2,856,216: the $2,763,716 grant plus the $92,500 pilot; Walton's share signed since 2024 is 76%. #233 at $20,396,286, as the executed Fourth Amendment (effective December 18, 2024) sets it: $529,930 more, total project cost $111,734,640, and the same figure Triumph lists. The reconciliation row now reads “agreement amount differs from the list”: four awards, none of them changed by an amendment. | The data build read #187's amount from a field that had lost its base figure, and took #233's from an unsigned October 2024 draft of the Fourth Amendment ($857,405 more) that is posted beside the signed one. A check of the posted documents found both. |
| 2026-11-01 | Triumph awards #46 and #209: what the agreements say | #46: the Third Amendment “expanded scope to Phases IV and V of the Southwest Crestview Bypass and East-West Connector”, and the metric is “a region-wide military employment count”. #209: Lovell Government Services named as the Company. | #46: Phases IV and V were the project from the start; the Third Amendment adds two collector roads and their signals, the Edney Avenue Collector and the Duggan Avenue Collector. The metric is the Okaloosa County count of employment in EMSI sector 901200 (federal military). #209: Lovell was the named Company in the 2020 agreement; the Second Amendment (August 24, 2023) replaced it with “multiple private sector companies”. | Read again against the executed agreement and amendments. |
| 2026-10-25 | Triumph job deals: the break-even strip | Wages mixed agreement floors with score-sheet and application averages: #365 at $77,000, #331 at $41,900, #258 at $48,026, #344 at $60,000. #331 ranked hardest to justify (15.9% on its timetable); 7 of 14 deals above 12.7%; #338 sixth. | Each deal on its agreement's wage floor where the agreement states one or ties it to a posted figure (12 of 15; #331 and #344 at lower bounds, noted on their pages), otherwise a stated average (#345, #350). #331 is 12.2% on its timetable, #365 11.4%; 6 of 14 are above 12.7%; #338 ranks fifth. Deals with no timetable sit apart at the bottom. | Comparing a floor with an average put deals on different footings; #331's sheet wage was below its own agreement's floor. |
| 2026-10-25 | Triumph awards #350 and #237: wage | #350 at $62,607 and #237 at $50,790: 115% applied to figures that were already 115% of the county average (#350 needs 8.1%, 10.4% on its timetable). | #350 at $54,441 (its application: “$54,441 which is 115% of the 2024 Average Wage Requirements”), needing about 9.4%, 11.9% on its timetable; #237 at its agreement's $44,165 average. | The data build multiplied any “115% of $X” by 1.15; in both records $X was already the 115% figure. The rule is removed. |
| 2026-10-25 | Triumph awards spreadsheet (awards.csv) | A column named wage_floor_usd held the agreement's wage floor for some awards and the score sheet's or application's stated average for others, without saying which. | wage_usd, with a jobs_wage_source column (agreement, score sheet or application) and a guide to every column (awards-columns.csv). | A stated average is not a floor; a reader comparing deals needed to know which they had. |
| 2026-10-25 | Triumph job deals: award #209 (I-10 Park) wage basis | 174 jobs at about $40,630, “115% of the 2019 county average, as the Second Amendment and its score sheet require”: 9.4%, or 14.8% on its timetable. | 174 jobs at $41,770 or more: 9.2%, or 14.4% on its timetable. The Second Amendment sets the wage at 115% of the county average as of the Commencement Date, not 2019; $41,770 is that figure in Santa Rosa's 2020 and 2021 agreements, so it is a lower bound. | Only the amended score sheet names 2019; the agreement itself ties the wage to a later date, which gives a higher wage and a slightly lower bar. |
| 2026-10-25 | Triumph job deals: award #209 (I-10 Park) | The amended grant ($5.4M) set against the original 90 jobs at $52,995: needs 14.0% of deals like it. | Since replaced: see the #209 wage-basis entry above. | The amended sheet states its wage as a percentage, which the data build could not read, so it fell back to the original sheet's terms. |
| 2026-10-25 | Triumph awards table | Five awards whose passed deadline is a completion date tagged “Deadline passed · outcome not posted”. | “Triumph reports construction completed”, as Triumph's tracker says. | The deadline was for completing construction, which the tracker reports done. |
| 2026-10-25 | Triumph per-resident view | Gulf County the largest per resident, at about $2,401. | With award #237, which its grantee said may lapse, left out: Gulf about $1,410, behind Wakulla ($1,815). | The earlier figure counted an award that may not be paid. |
| 2026-10-25 | County floor for Triumph awards | Cited as s. 288.8017, F.S. | s. 288.8013(2)(b), F.S. | Wrong section number. |
| 2026-10-25 | Triumph awards table | Passed deadlines tagged “Deadline passed”. | “Deadline passed · outcome not posted”, with Triumph's own tracker status. | A passed date in the posted documents is not evidence the deadline was missed; an unposted amendment or waiver may have moved it. |
| 2026-10-25 | Triumph award pages | Every stated return shown as “$X per Triumph dollar”. | Each return in its score sheet's own unit, e.g. #191's 5.4%, #69's benefit-cost ratio. | Score sheets use 16 different units; some are not dollars per dollar, and one (#320) is a statewide figure for another program. |
| 2026-10-25 | Triumph awards #49 and #209 | Titled “ed Tonnage Chart March 2018” and “1-10 Park”. | Port Panama City East Terminal Development; I-10 Park. | The first was a document label taken as a title; the second an OCR misreading. |
Other pages
| Date | Where | It said | It says now | Why |
|---|---|---|---|---|
| 2026-11-01 | Services page | Offered a paid “Grant accountability review” with Triumph's awards as the example, project screens asking “can it repay what it borrows”, and a review step in which the commissioner checks the draft. | No paid work on Triumph awards, or for anyone holding or seeking one, or for an applicant reviewing its own project, and none for any government or public body until a funding and conflict-of-interest policy is published. Who may pay is stated. Commissioned work is published unless agreed otherwise at the start; the commissioner checks facts only, and every change is logged. | Epomena's commercial plan rules out selling on Triumph awards, and an analysis of a borrower's ability to repay, done for a paying public body, would be municipal advice that Virtuous Cycle LLC is not registered to give. |
| 2026-10-25 | Site description, county list and prototype notice | Described the break-even screen as “a test of whether a jobs grant changed the decision” and promised to show “whether the jobs would have come anyway”; said figures “may change without notice”. | Says what the screen computes: how often the grant must be the deciding reason for the wages to equal it. The notice points to this log for any change to a figure or claim. | The screen cannot show whether a grant changed a decision, and changes to figures are logged, not made without notice. |
| 2026-10-25 | Health and place screens | Estimates whose intervals overlap called “similar”. | “Not distinguishable at this precision” (“within the margin” in the census table); CDC PLACES figures labelled as modelled. | Overlapping intervals mean the estimates cannot be told apart, not that they are alike. |
| 2026-10-25 | Data registry | Tagged “No AI processing”. | Says some documents were read with AI assistance and spot-checked, and that no AI model computes a number. | Triumph's agreements and score sheets were read from their PDFs with AI assistance. The old tag was wrong. |
| 2026-10-25 | Printable brief | “Enterprise NPV −$6.8M” and “costs only, −$5.2M” for Gulf Breeze undergrounding. | Labelled as the owner's own cash, not public value, and as costs only, not a net loss. | With no benefit valued, a negative figure counts costs alone. It does not show the project lost money for the public. |
Who keeps the tax
| Date | Where | It said | It says now | Why |
|---|---|---|---|---|
| 2026-10-25 | Who keeps the tax (/fiscal): the school levy | “Statewide school funding (offsets state aid)”: the gain “spread across the state's schools and taxpayers”. | “Lower school tax rates statewide”: with the statewide total of required local effort fixed, new value here lowers the required school millage everywhere; Escambia's state aid falls, other districts' rises, and total school funding does not change. True within a year; less certain over several. | Schools gain nothing from the offset; school taxpayers do. Our earlier wording got the incidence wrong. |
| 2026-10-25 | Who keeps the tax (/fiscal) | The school offset labelled “State (school aid it no longer pays)” and described as saving the state money. | Since replaced: see the required-local-effort entry above. | The statute fixes the statewide total (s. 1011.62(4), F.S.), so the gain is spread across districts, not kept as a state budget saving. |
| 2026-10-25 | Who keeps the tax (/fiscal) | The school offset described as 100%, then as 96% with 4% kept by the district; the sources table still said “dollar for dollar”. | About 96% offsets state aid; the 4% kept on paper roughly matches early-payment discounts (up to 4%, s. 197.162, F.S.) and unpaid tax, so the district's net gain is probably near zero. | The formula counts 96% of taxable value, and the sources table had not been updated to match the page. |
Maps, flood record and site hazards
| Date | Where | It said | It says now | Why |
|---|---|---|---|---|
| 2026-10-25 | Maps: county names, city limits and bays | County names placed beside each county's dots, which often put them in a neighbouring county (Escambia in Alabama, Bay in Washington); zoomed in, names of counties out of view pinned to the frame's edge. City limits tinted over open water: Gulf Breeze drew as about 57 km², 78% of it Pensacola Bay and Santa Rosa Sound. Blackwater Bay and Oyster Bay partly drawn as land. | Each county's name sits inside that county, or is left off when none of it is in view. City limits are cut to land, as the counties are. Bay pieces are merged before the 1 km² size test, so Blackwater Bay is 8% land, down from 28%. | A name or a tint in the wrong place tells a reader where something is, and it was wrong. |
| 2026-10-25 | Flood and resilience record: headline | Ranked tracts by all flood-insurance claims. | Ranks by paid claims, gives the number closed without payment, separates repeat-loss properties since mitigated, and says these are counts, not rates. | Claims closed without payment are not losses, and a count with no exposure denominator is not a rate. |
| 2026-10-25 | Maps: county and tract shapes | Rivers and channels cut out as hairline holes, which drew as county lines; Perdido Bay half land on the Alabama side. | Only broad water (bays, lakes, river mouths over about 300 m wide) is cut, on both sides of Perdido Bay; every shape is valid. | The water cut kept every water body over 1 km², and long rivers pass that test. |
| 2026-10-25 | Flood and resilience record: map | Tracts faded by population density. | Full class colours, with each class's share of residents in the bar under the key. | The fading made most high-on-both tracts look like the middle class, so the map hid its own finding. |
| 2026-10-25 | Project screens: site hazards | Quoted s. 163.3178(1) and s. 380.27(2), F.S. when an assumed project line crossed a hazard zone, and called Pace's projects outside the zones from one point. | Statutes are quoted only on a published footprint with at least 5% of the site in the zone; otherwise the overlap is reported as too uncertain to weigh. | Neither project's limits were published, so the overlap was an artefact of an assumed shape. |
Gulf Breeze undergrounding loan
| Date | Where | It said | It says now | Why |
|---|---|---|---|---|
| 2026-10-25 | Gulf Breeze undergrounding: loan coverage | Coverage drawn and coloured as met in every year to FY2041; the rate stress shown as if the rate could change; the answer to “Can it repay?” gave the tightest year only. | Years after the CRA's possible 2039 end are grey, with a 0× line if it is not continued; the rate stress is labelled as applying only to a refinancing; the answer names the 2039 contingency. | The CRA ends by law in 2039 unless continued, two years before the last payment, and the record implies the 2.10% rate is fixed. |
| 2026-10-25 | Gulf Breeze undergrounding: loan structure note | A level payment at 2.10% is $660,709. | $662,947, the level payment on the modelled $10,298,762 balance over 19 years. | The earlier figure did not match the modelled balance. |
| 2026-10-25 | Gulf Breeze undergrounding | “Can it repay? Probably, but not settled”; borrower “not named”; the stakeholder table had the tax increment “paying back” $30.8M. | “On paper; not settled”; the CRA named as borrower; the $30.8M labelled as all the increment over the term; the CRA sunset under s. 163.3755, F.S. cited. | The pledge terms are not published, coverage counts all of the CRA's increment, and the pages disagreed with each other. |