The corrections log

Every correction, newest first

• Sep 15 · The CDOT weekly capture has a gap: no snapshot exists for 25 August. Not a figure moving — a disclosure of a limit in how this site collects data. This site preserves CDOT's payee-level expenditure feed weekly, because the publisher serves it on a rolling window with no archive of its own: a week not captured here is a week gone for good. Captures exist for 4, 11 and 18 August, then jump to 1 and 8 September. Fourteen days passed between the 18 August and 1 September captures instead of the usual seven, so one week's snapshot does not exist. No published figure on this site is computed across that missing week. Every CDOT figure this site publishes is read from the newest snapshot's own monthly profile, and a month is refused as PENDING whenever its recorded row count falls below three quarters of the archive's median month — a bar a week this thin would already fail. Going forward, any published window that would span the missing week shows PENDING with this gap named as the reason, rather than a figure computed across it. The gap itself is named in the CDOT archive report's own capture-log table, on the row for 1 September, rather than reading like an ordinary week.disclosed
• Sep 5 · Four project rows said no document states their figure. Two of those documents were already in our own archive. On 18 August we logged that seven project rows cited a document that does not state their figure, and that four of the seven “could not be fixed, so they were re-graded instead”. All four are fixed today. The documents that state them existed the whole time. Capitol Complex Renovation, ≈ $216.7M. The Joint Budget Committee’s FY 2026-27 capital figure-setting document states it on page 15: “The total cost of the project is currently estimated at $216.7 million, with an unfunded $31.7 million General Fund cost.” CSU Clark Building, ≈ $136.6M. Page 8 of the same document states all three parts together — $55.8M of earlier Capital Construction Fund appropriations, $55.0M of the university’s own institutional funds, and the $25,798,175 final phase the Long Bill enacted. The row now prints that addition. MyUI+, $86.2M, and DRIVES, $93.3M. Both come from the Legislative Council Staff memorandum on major IT capital construction projects, written for the Joint Technology Committee in December 2022. It says on page 4 that the Department of Revenue “was appropriated a total of $93.3 million for the project”, and MyUI+’s $57.8M and $28.4M sit in two consecutive sentences on page 8. No figure on this site changes. All four amounts are what they were. What changes is that each one now names a document that says so, with the page, and the two we add up print the addition. The four rows move from confidence C back to B: B here means an official record that may lag, or that restates a figure it did not originate, which is what a 2022 memorandum and a staff estimate are. The part worth publishing is what “no source states it” really meant. One of the two documents was already in our own archive, cited by thirty other rows on the same page, and nobody had read past its tables. The other answered on the first request, at the address this site had recorded for it since July. A check now carries each of the four sentences with its page, and fails if the published amount, the citation or the quotation drifts off the other two.corrected
• Sep 5 · The Sep 3 entry said the page no longer states the road safety split in prose, and three other places still did. The Sep 3 correction two rows down was about one paragraph on What you pay to register a car and closed by saying “the page no longer states a split in prose at all: it derives the sentence from the statutory schedule.” That was true of that one paragraph. The fee table’s own destination cell, an earlier plain-English paragraph higher on the same page, the receipt’s client-side calculator, and the road safety surcharge’s own definition on Ask WB all still read “60% state, 22% counties, 18% towns” as a flat, present-tense fact — three days after the entry above said the sentence was fixed. No figure changed and no figure was wrong; this is the same shape as the Sep 4 entry above it, found the same way: a fix that reached one surface and stopped. The destination string is now built from the same statutory schedule the corrected paragraph already used — C.R.S. 43-4-205(6)(b) at 60/22/18 before 1 September 2025, C.R.S. 43-4-205(6.3)(b)(I) (Senate Bill 25-258) at 56/24/20 from that day — so all four surfaces read one derivation instead of four typed copies, and a selftest now asserts the three shares sum to 100 and match the schedule in force. Found by a page-by-page audit of the site’s own public copy.corrected
• Sep 4 · Three pages divided one budget by three different populations, and the entry below said we had already fixed it. On 5 August we logged the correction two rows down: the $100 receipt had been dividing by an unsourced “5.99M residents (est.)”, and we wrote that “every per-resident figure is now on that single Census estimate”. That was true of one page. The fix reached the dashboard and stopped there. Until today Follow Your Dollar published ≈ $8,275 per resident and the worked example on the methodology page published ≈ $8,280, both still on the withdrawn population, while the dashboard published ≈ $8,243. One quantity, three answers, on three pages a reader can open side by side. Follow Your Dollar also multiplies that figure by a county population to draw its population-share benchmark, so every one of the 64 county benchmarks was about 0.4 percent high. The figure is ≈ $8,243. That is $49,560,763,693 of FY 2026-27 appropriations over 6,012,561 Colorado residents — the U.S. Census Bureau’s Vintage 2025 estimate for 1 July 2025, and the same count the incarceration rate on this site divides by. Neither the budget nor the population moved. What changed is that all three pages now read them instead of remembering them. The part worth publishing is why nothing caught it. Our checks proved the dashboard’s figure against the pipeline on every run, and never asked whether any other page agreed with it — so a correction we had published, and believed, was false on two of the three pages it covered. Every surface that prints this figure or its denominator is now written from one derivation, and the check fails if any of those pages prints either withdrawn number anywhere: inside a figure, or in the sentence beside it.corrected
• Sep 3 · The registration page said the road safety surcharge splits 60/22/18. For ten of the twelve months printed beside it, it did not. On What you pay to register a car we wrote that statute splits the Highway Users Tax Fund “60 percent to state highways, 22 percent to counties and 18 percent to towns and cities”, and printed the FY2026 figures underneath it. Senate Bill 25-258 changed that split on 1 September 2025 — to 56 percent, 24 percent and 20 percent, for any registration period beginning on or after that day and before 1 September 2027. Ten of the twelve months behind the FY2026 total are on the new split and two are on the old one, so the sentence above them was true of neither year as printed. No figure changed and no figure was wrong. The Treasurer’s workbook reconciles to the cent, and it still does. What was wrong is what we told a reader those figures were a share of: the FY2026 county share of $50,971,605.00 is $3,491,735.38 more than 22 percent of the collection beside it, and the state highway share is $6,983,471.33 less than 60 percent of it. Nothing here could have caught it, and that is the part worth keeping. Our check on this workbook asserted that the three distribution rows add up to the collection. They do — on either split, because moving money between the parts leaves the whole alone. A sum is not a proportion, and only the second one can see a redistribution. The Treasury’s own rows are still labeled 60% State Highway, 22% County and 18% City while carrying the new percentages, so the labels agreed with our sentence and disagreed with their own figures. Every distribution row is now checked every run against the share the statute in force actually puts it on, month by month, and a year that straddles the change says so beside its own total. The page no longer states a split in prose at all: it derives the sentence from the statutory schedule, so it will follow the law back when this window closes in September 2027. Found while writing the weekly brief.corrected
• Sep 3 · Reader input — this site previously said that ratings, comments, follows and the county sign-up were demo only and that nothing was collected. From today they are real. What you send goes to a store this project controls: a rating or comment is kept with no name, address or identifier and is published only as totals; a sign-up or follow keeps your address and what you follow, sends a confirmation link before anything else, and unsubscribes in one click. We do not record IP addresses, and we do not track whether an email was opened. The handling note is in the methodologydisclosed
• Sep 2 · Three shares on the registration page, two of them called “the bill”, and the two bills were $29 apart. On What you pay to register a car the ownership tax was “63.8% of what you pay”, the two FASTER surcharges — road safety and bridge safety — “53.6% of every fee on the bill”, and the Keep Colorado Wild park pass “13.1% of the bill”. Every one recomputes exactly, and they sit on three different denominators: $191.98, the $69.58 fee subtotal, and $220.98. The page shows two totals on purpose, because the $29 pass is added unless you decline it — but “every fee on the bill” reads as though the pass is inside a subtotal it is not in, and two sentences said “the bill” and meant totals $29 apart. No figure changed. Each now names the quantity it divides by, in the words the receipt table above it already uses. Found by the audit of every derived share on this site.clarified
• Sep 2 · A share on the Energy episode counted homes that no fuel on its chart heats. Under the fuel chart on Energy Costs we wrote: “The six fuels above reach the other 95%.” That 95% was 100 less the two fuels we had just excluded — propane at 4.5% and heating oil at 0.19% — so it quietly handed every home heated by wood, coal, solar or something else, and every home reporting no heating fuel at all, to the fuels on the chart. On the Census table we cite, the two charted fuels that actually heat homes are utility gas and electricity, at 66.6% and 25.5%: about 92%, not 95%. The sentence also called six rows six fuels, where diesel and gasoline each appear twice and neither heats anything. The chart did not change and no fuel price moved — the sentence beside it did. A denominator’s complement is not the same thing as its parts’ reach, and taking one for the other is how a part-to-whole absorbs a category nobody meant to include. Found by the audit of every derived share on this site; the 2 September entry below says that audit had found no such denominator, and this is the one it found.clarified
• Sep 2 · The fund-source shares wore the wrong label, and were typed in three places. The four percentages under “Who pays for it” — General Fund 35.0%, federal 31.0%, cash funds 28.2%, reappropriated 5.8% — carried a FACT chip. The dollars behind them are FACT: the Long Bill prints each one. The percentages are those dollars divided by the act’s $49.6B total, which is arithmetic we did. This site’s own word for that is CALC, and a CALC shows its formula, so now it does. No figure changed. 35.0% is still 35.0%. The second thing was worse and no reader could have seen it. The same split was written into the site three times — the bar on this page, a caption on Follow Your Dollar, and a sentence inside that page’s script. Only the bar was checked against the budget data. The other two were typed. They happened to be right; nothing would have told us when they stopped. The Long Bill is re-read as it is amended, so the checked copy would have moved and the typed copies would have stood still. There is one copy now, and all three places are written from it. Found by auditing every share on this site for a denominator that leaves a category out; that audit found no such denominator, and found this instead.corrected
• Aug 27 · A chip on the Education page used a label the Honesty Framework does not have. The infant care price on Education carried a chip reading CCAoA. Every other chip on this site answers one question: how do we know this? FACT when a publisher states it, CALC when we did arithmetic and show it, PENDING when we will not guess. That one answered a different question — who published it. The figure is unchanged at $20,978, and so is its source. It comes from Child Care Aware of America, a national nonprofit rather than a government body, and the tile has always said so in words underneath. What was wrong is where that belonged. A reader learns one small piece of grammar on this site, and a chip that answers a different question from every other chip takes it away from them. It was also a label defined nowhere. Not on the Honesty Framework (opens in a new tab) page, not in the methodology register — though our own rule is that any new chip is defined on both, in the same ship as the chip. The chip now reads FACT, which is what our dataset has labeled this observation all along, and the nonprofit’s name stays where it was.corrected
• Aug 24 · Four blank TABOR years are filled, and a fifth figure was on the wrong basis. — Four fiscal years on Cost of Living showed a dashed empty box where a refund belonged: FY 2018-19, FY 2020-21, FY 2022-23 and FY 2023-24. The page said why — the year’s own document had not been read. It has now. The State Auditor publishes a Schedule of TABOR Revenue for every fiscal year, and all four are read: $428,335,506, $453,572,197, $3,568,639,070 and $1,371,507,859. Reading them turned up a figure we had already published on the wrong basis. These schedules print two numbers that look alike and are not. One is the refund arising from that year’s surplus. The other is a running total of everything still owed from earlier years as well. Our chart says it shows the first — “TABOR refund by surplus fiscal year” — and for FY 2021-22 it was showing the second. That bar read $3,868,404,442. Of that, $133,769,928 was money left over from the 2015, 2018, 2019 and 2021 surpluses. The FY 2022 surplus itself was $3,734,634,514, and that is what the bar reads now. It is still the largest refund in the series and still went out as the S.B. 22-233 checks of $750 per single filer and $1,500 per joint return; what changed is that the figure now answers the question the chart asks. FY 2024-25 keeps its $293,303,186. The subtraction there is a different thing — prior-year over-refunds set against this surplus under state law, not another year’s money. Why we had it wrong. Two years were written out by hand, each choosing which of the two numbers to take. One place now makes that choice for every year, and the build refuses to ship a year whose two figures do not add up between them.corrected
• Aug 24 · We now count which broad region you are reading from — Until today the only place we recorded was the country. We now also record a broad region: at its finest a group of counties, more often a whole state, and outside the country nothing but that. How it works, because this one is worth stating plainly. When a page loads, your browser asks our own server one small question. The answer is that region and nothing else. The closer location your network reports is read where your request first reaches us, and thrown away in the same breath. It never reaches the page, it is never written down, and nothing about it is kept between one page and the next. No cookie, no storage, and nothing you typed. This is the only request this site makes on your behalf that you did not ask for. We would rather say so than have you find it. The methodology page carries the full wording.disclosed
• Aug 23 · Our comparison bars were drawn too short — Each bar takes its length from its own percentage. This is the chart comparing Colorado’s spending mix with the 50-state average. The number beside each bar sat in the same box and could not shrink. So the bar gave up room to the number. The longer the bar, the worse it got. Colorado’s Medicaid bar should fill 85% of its track. On a wide screen it filled 75.7%. On a phone it collapsed to a stub. No percentage on this page changed, and none was wrong. What was wrong is that the picture disagreed with the figures beside it. A reader comparing two bars by eye was comparing the wrong lengths. Every bar now draws the length its own figure sets, at every screen width. The same repair fixed the chart on a phone. There the value boxes had overlapped each other, and the “All states” column ran off the side of the screen.corrected
• Aug 22 · We now count a few interactions, not just page views — On Aug 22 we published the rules for this and said it was not yet running. It is running now, in this deploy, which is what those rules said would happen. What is counted: when a citation is followed out to its publisher, we record the publisher’s web address; when the Honesty Framework is opened, which page it was opened from; when a data table is opened, which table; and on the one prototype page that plays questions one at a time, which question was shown and which controls were used. What is not: nothing you type, no cookie, no identifier, and nothing that says which reader did any of it — these are counts, not records of people. Two things we designed and did not ship: counting clicks on a refusal, and on a corrections entry. Neither is a button, so neither could ever have fired, and a number stuck at zero would have read as “nobody does this” when the truth was “nothing was listening”. Details in the methodologydisclosed
• Aug 22 · Privacy charter, extended and pre-registered — Nothing new is collected today. Two changes to what we say, not to what we measure. First, the full traffic disclosure now appears on every locale, in the same words. It had been written out in detail on two of them and in a single line on the rest — the same policy, unevenly stated. Second, we have written down the rules for something we plan to add: counting interactions as well as page views, so we can see which sources get followed and which controls get used. Three rules are settled before any of it ships. Every event has a name from a fixed list. Nothing you type is ever recorded — not a ZIP code, not a search, not the words of a comment. Location stays coarse, a region or a state, never an address. We are publishing those rules now so they can be read before the counting starts rather than after, and a check in our build refuses to let the page claim otherwise in either direction. Details in the methodologydisclosed
• Aug 22 · The Honesty Framework did not allow for a forecast, and Cost of Living publishes three. Nothing published changed, and no figure moved. What changed is the page that exists to explain the labels. The framework said a blank labeled PENDING “beats an estimate every time” and, of NOT PUBLISHED, that “we never fill it with an estimate.” Both rules are about our estimates and neither bends. But the TABOR section on Cost of Living shows three fiscal years — FY 2025-26, FY 2026-27 and FY 2027-28 — as outlined bars marked “estimate”, because Legislative Council Staff forecast the refund and publish the figures. Quoting a publisher’s own forecast is not the same act as inventing a number, and the site had been doing the first while its framework page described only the second. A reader clicking from one of those bars to the page meant to explain it met a page that read as forbidding it. The framework now defines the state, at The forecasts: a publisher’s forecast may be published, with their name on it, marked as a forecast everywhere it appears, never added to a certified figure, and superseded through this log when the certified figure arrives. The methodology page says the same. Also corrected here: this log said it held 24 entries and held 25. It now says 26, and a check fails the build if the two ever disagree again.corrected
• Aug 21 · The registration calculator was a fee short for electric vehicles. The “work out your own bill” calculator on What you pay to register a car charged electric and plug-in hybrid vehicles one extra fee: the road usage fee. The law charges two. The second is the Electric Vehicle Fee — $50 when it was created in 2013, indexed to inflation since, and $63.05 this year on the DMV’s own fee page. The law’s words about the road usage fee are “in addition to ... the fee imposed by subsection (25)(a)”, and the law defines “electric motor vehicle” to cover plug-in hybrids too. So every total the calculator produced for an electric or plug-in hybrid vehicle was $63.05 short, from the day the page shipped on 8 August until today. The worked receipt on the page is a gasoline SUV, so the printed example was never wrong, and no figure in our dataset changed. The calculator now charges both fees, to both fuel types. Why our own check missed it. The build refuses to ship unless every row of the DMV’s fee list is accounted for — that check exists precisely to catch an omitted line. The row “Electric Vehicle Fees” was accounted for. It was mapped to the one fee we carried, while the section behind it listed two. A row that names a section rather than a single fee can hide a second charge inside a check that reads clean. The mapping now names both, and the build fails if either goes missing. A calculator that understates a bill is exactly the kind of error this log exists forcorrected
• Aug 19 · The school funding figure cited a document written before the bill passed. This page published Total Program funding for FY 2025-26 as $10,035,615,918, tagged FACT, credited to HB25-1320, the School Finance Act. That figure is real. It is not the enacted one. It is Legislative Council Staff’s estimate from the initial fiscal note of 3 April 2025. That note says of itself that it “reflects the introduced bill”. The Governor signed the act seven weeks later. We were citing a projection of a bill on a line that says enacted. The figure now reads $10,031,606,090. That is the statewide total in the staff’s own district run↗, dated 9 June 2025. That is their district-by-district table of school funding, as appropriated in HB25-1320 and SB25-315. The enacted act states two floors, and only one of them applies. Section 16 sets the same $10,031,606,090. Section 15 sets $10,036,070,748, and it never came into force. The final fiscal note says so in its own words. Senate Bill 25-315 took effect on 23 May 2025. So Section 16 took effect, and Section 15 did not. The arithmetic on this page now works. Funding per pupil beside it reads $11,858. The new total divided by the 845,943 funded pupils that same document counts gives $11,858. The old total over its own pupil count gives $11,852. We had a total and a per-pupil figure from two stages of one bill, and neither divided into the other. FY 2024-25 moved too, and it got stronger. It read $9,777,300,000 and was a calculation: a memo’s rounded “about $10.03 billion” minus its stated $252.7M increase. Both fiscal notes print the actual figure, $9,778,950,899. It is now a sourced fact rather than a subtraction, $1.65M higher. What you see on the tile moves by a hundredth of a billion. It read $10.04B and now reads $10.03B. FY 2024-25 still rounds to $9.78B, and the change against last year is still +2.6%. The visible repair is small. The reason for it is not. A citation is the promise that you can check us, and this one sent you to a document that does not hold the number. Every document behind these figures is now kept in our own archive, so this stays checkable if one is taken downcorrected
• Aug 18 · Seven projects cited a document that does not state their figure. No project figure changes. Every row in the project ledger names the source its number comes from. Seven named the wrong one. All seven pointed at HB26-1410, the enacted Long Bill — and that act does not contain those figures. Two of them are impossible on their face. The Capitol Complex row published ≈$216.7M against an act whose entire capital construction section is $160,355,815. The MyUI+ row published $86.2M against an act whose entire information-technology section is $32,325,503. A reader who followed either citation to check the number would have landed on a document a third to a quarter of its size. Three are fixed outright. The CMHIP, CO School for the Deaf & Blind and Arkansas Valley rows do have a source that states their figure. It is the Joint Budget Committee’s capital figure-setting document, already listed on each row as a second source. It is now the first one, and the Long Bill citation is gone. Four could not be fixed, so they were re-graded instead. Capitol Complex, CSU Clark Building, MyUI+ and DRIVES move from confidence A and B to C. On this site A means strongly sourced and C means weak, and a figure whose source we cannot name is not strongly sourced. The numbers themselves may well be right. The Clark Building row even breaks its own total down into $55.8M of Capital Construction Fund, $55.0M institutional and a $25.8M final phase, which adds up exactly. But we cannot yet point you at the document that says so. Confidence grades are display only here; no total on this page is calculated from them, so nothing moved. We are logging this because a citation is a promise that you can check our work. Seven of ours could not be checked, and three of those pointed at a number that was not merely absent from the cited document but different from the one it prints. A check now compares every ledger row against the source it names, and prints the ones that still cannot be matchedcorrected
• Aug 18 · “Budget gap closed — $1.2B”, and five of the six figures beside it. This page carried a tile reading $1.2B, under the words “shortfall balanced in the FY27 package”. Beside it sat a card of six FY27 pressure points. All seven were tagged FACT. None of them named a source. So we went to the document they were said to come from. It is the Joint Budget Committee’s own Budget Package & Long Bill Narrative↗, dated April 7, 2026. Five of the seven claims are not in it. We also read the FY 2026-27 Summary Tables and the FY 2026-27 Appropriations Report. None of the three states a $1.2 billion budget gap, in any wording. The narrative never names a headline shortfall at all. What it describes is different, and is now what the tile says. House Bill 26-1363 cut the General Fund reserve requirement from 15.0% to 13.0% for FY 2025-26 and FY 2026-27. The budget package balances to that reduced requirement. Two of the old figures survived exactly and are still here. One is the reserve ending $340.1M below the old 15% rule. The other is the $213.3M estimated to come from the Kids Matter Account. One was close to a real number but was not it. We published Medicaid provider rate reductions at −$270M. The narrative’s provider-rate line is −$288,969,529. The 2.0% across-the-board cut inside it is $222,022,950. Neither is $270M. Three do not appear at all. Those are “cuts across smaller departments −$150M” and “employee & contractor compensation savings −$120M”. The third is “Medicaid cost growth ~9% vs 3.2%/yr”. The card now carries six figures the narrative does state. Each one is read from the document rather than typed. Three are reconciled against a total the JBC prints itself. No other figure on this page changed. The budget, the General Fund, the receipt and the project ledger are untouched. We are logging this at length because the failure was not arithmetic. It was a source that nobody had checked, on figures that looked exactly like the checked ones beside themcorrected
• Aug 17 · One project in the ledger was two years of money added together. The CU Boulder Guggenheim Geography renovation was published at ≈ $24.2M. That is not what the state was asked for in this budget year. The Joint Budget Committee’s figure-setting document puts the FY 2026-27 request at $5,022,196, and a further $19,200,888 in a separate column for FY 2027-28. Those two add to $24,223,084, which is the figure we had. Every other requested project on this page is one year of money, so this row was on a different basis from the twenty-nine beside it and nothing said so. It now reads $5.0M, the later year is stated on the project’s own page instead of being folded in, and the row’s grade moves from B to A because the figure now comes from the primary document rather than a summary of it. The requested-money total below the ledger moves with it, from ≈ $0.8B to ≈ $0.7B. No funded figure changes: requested money has never entered the headline, which is the whole reason the two are kept apart. We found it while building a check that reads that document row by row — the other twenty-nine matched it to the dollar, and this one did notcorrected
• Aug 17 · Nine bills on Before It’s Law had no figure where the note printed one. A fiscal note is Legislative Council Staff’s estimate of what a bill would cost. We read that estimate off the note’s State Expenditures row and stopped there. Colorado’s older notes often leave that row empty. They print the money one line below it, on the fund it comes out of. HB16-1022 is the plainest case. Its note shows nothing on the State Expenditures line and $243.2 million on the General Fund line beneath it. We published “Row printed, no figure”. We now read that line too. Nine bills gain a figure: HB16-1022, HB16-1075, HB16-1105, HB16-1117, HB16-1422, HB17-1361, SB16-023, SB16-064 and SB17-029. No figure changed to a different figure. Twelve of the sixteen notes are confirmed to the dollar by the note’s own “Appropriation Required” sentence, written elsewhere in the same document. The counts above the table move with them. Bills whose estimate can be compared go from 3,584 to 3,592, and “down beats up” goes from about 2.1 to 1 to 2.0 to 1. What we will not do is add up a block we cannot read honestly. Some notes move money between funds rather than spend it, so the lines cancel to zero. Publishing that zero as the cost of a bill would be inventing a number. Those blocks stay blank, and each one now records whycorrected
• Aug 14 · The citation under the Corrections budget tile — no figure changes. The tile's $1.19B FY 2025-26 total funds figure linked to the JBC's Corrections staff briefing PDF. That is not where the figure comes from. It comes from the FY 2025-26 Appropriations Report, exactly as this page's own detail table beneath it already states. This page also explains that the briefing and the Appropriations Report can print different totals for the same year — the FY 2024-25 restatement below is one example, off by $1.7M. So a reader clicking the link to check $1.19B landed on a document that could show a different number from the one they were checking. The link now points at the Appropriations Report, and its label reads “JBC Appropriations Report” rather than “JBC briefing”, so the label matches what it links to. $1.19B, the +19.2% change, and the ≈2.5% share are all unchangedclarified
• Aug 11 · Four 2024 offense counts on this page moved, because the Bureau of Investigation revised them. Murder 266 → 267, non-consensual sex offenses 6,250 → 6,274, assault 21,277 → 21,292, robbery 3,794 → 3,823. The gap column against the CDPS extract moved with them, by a tenth of a point in each case. The divergence those rows exist to show is unchanged. Nobody got anything wrong. CBI revises its crime statistics nightly. Our policy is that the latest vintage wins. These numbers are what it published when we last read it. It is logged because a published figure changed and the page did not otherwise say these figures move. A number that changes quietly is what this log is for, whether or not anyone erredrefreshed
• Aug 10 · The assessment rate on Follow Your Dollar was a year out of step with the mill levies beside it. Part C multiplied tax year 2025 mill levies by the tax year 2026 assessment rate. Colorado assesses a home twice, and the two rates behaved differently. The school rate is 7.05% in both years, so half the calculation was right. The local government rate is 6.8% in 2026 and 6.25% in 2025. So the local half of every estimate ran 8.8% high. That is about $89 on the county line of a $600,000 home in Jefferson County. Now 6.25%, paired with the 2025 levies it belongs to, and the whole section is labeled tax year 2025. Why it took sourcing rather than a one-line fix. House Bill 24B-1001 did not set a number. It set a rule: 6.25% if statewide value growth was at or below 5%, and 6.15% if above. Both of the Division of Property Taxation’s own rate pages return a server error where the answer should be. The determination is printed in the Division’s 2025 Annual Report: the State Board certified growth of 4.27%. Our check now re-applies that rule rather than trusting the transcription, so the wrong branch cannot be typed in unnoticed. No mill levy changed and no county figure changedcorrected
• Aug 8 · The T-REX delivery record, on Still Promised and Floyd Hill. Both pages said the project “finished 3.2% under budget and 22 months ahead of schedule in December 2006”. Three things were wrong with that sentence and none of them was the arithmetic. The date: RTD's own record says the line opened 17 November 2006. The denominator. The 3.2% is $39 million against the $1.23 billion design-build contract price. It is not against the $1.67 billion program budget printed in the same sentence, where the same $39M is 2.3%. A percentage sitting beside a budget it was not measured against reads the obvious way. A reader is entitled to read it that way. The tense. Our only source was an FHWA Public Roads article dated September/October 2001, five years before the project finished. It states what the winning team agreed to when it bid, and it cannot state what was delivered. Both pages now say so. The direction is properly sourced for the first time: RTD publishes that the line opened ahead of schedule and under budget. Nobody publishes the delivered size, so we no longer imply one. Found while re-probing our own sources: the FHWA link had also stopped serving the article entirelyrevised
• Aug 8 · How the $13.3–16.5 billion transit estimate is graded — no figure changes. The 2014 Advanced Guideway System capital range was published as CDOT's own and graded A. CDOT publishes only a summary of that study, and the report itself is available by records request. That summary does not print that range. It prints a different pair instead, for a bigger system. Those are $16.5B from DIA to Eagle and $13.6B from Fort Collins to Pueblo, inside a $30 billion statewide total. So one of our two numbers appears on CDOT's page attached to a different corridor. That is how a figure gets quietly mis-cited. The range itself is right, and it covers the phase we name: I-70/C-470 near Golden to Eagle County Regional Airport. But it reaches us through reporting published at the time of the study's release, which is grade B. Regraded, with the source named. The $1–3 billion private-financing ceiling is CDOT's own. So is the “not financially feasible” finding. Both stay grade A. The two multiples against Floyd Hill's cost inherit the weaker grade. Every calculation here inherits from its worst inputregraded
• Aug 7 · The built share of the B Line, on Still Promised — we published 15% of the balloted corridor as built. It now reads 15.1%, and the share never built is stated beside it as 84.9%. The arithmetic never changed: 6.2 miles of 41 is 15.1%. What changed is the rounding, and the reason we gave for it. Our own note said rounding down to 15% “understates how little was built”. That is backwards. Rounding the built share down makes the shortfall look larger, which flatters the argument the page is making. The house rule is to round against ourselves. It also left the built and unbuilt shares adding to 100 only because two roundings happened to go opposite ways. Both are now stated to one decimal, so they sum to exactly 100.0 and neither favors us. No mileage figure changed and nothing about the corridor changedrevised
• Aug 5 · The source line under the headline — no figure changes. It cited the JBC FY 2026-27 Summary Tables beside the Long Bill. Those tables state their own basis in their opening paragraph. They reflect the bill “as introduced in the House”, which is a proposal. After the two corrections below, no figure on this page comes from them. Citing a proposal on a page of enacted figures is a real error. A reader has every right to be annoyed about it. It is gone. The line now names the enacted act and its chapter law. The refresh date was also stale by a monthclarified
• Aug 5 · The change against last year, on both budget tiles. The headline tile said the budget was up +3.9% and the General Fund up +1.5%. Both came from the Joint Budget Committee's summary tables. Those tables say of themselves that they reflect the Long Bill “as introduced in the House”. Those tables compare a proposed $49,598,197,293 against an adjusted $47,727,166,661. Neither of those is the enacted $49,560,763,693 the tile displays. The percentages were arithmetically correct and described a different pair of numbers from the one shown above them. They now read +6.4% and +2.7%, comparing this enacted Long Bill against the last enacted Long Bill. That is one act against the same act a year earlier. Both are parsed from the enacted text and reconciled against the totals the acts print for themselves. The General Fund comparison needed care. The FY 2025-26 act prints General Fund Exempt as a column of its own, and this one folds it into General Fund. Comparing the printed columns would have shown the General Fund falling 24.6% in a year it roserevised
• Aug 5 · Shown, not controlled — the “What lawmakers actually control” split read ≈ $31.8B (64.2%) appropriated against ≈ $17.7B (35.8%) informational. Read off the footnotes the bill prints beneath its own statewide total. On that reading the informational “(I)” money is $17,988,473,956 and the controlled share $31,572,289,737 — 63.7% / 36.3%. A difference of about $246M, and every figure on the bar now comes from the act's own arithmetic. We also added the caveat this figure most needs: the share is not comparable year to year. The identical footnotes in the FY 2025-26 Long Bill give 12.8%. This act marks nearly all federal funds “(I)”, where its predecessor marked about a quarter of themrevised
• Aug 5 · “17 agencies” in the $100 receipt — the grouped row read “All other departments · 17 agencies”. Colorado has 23 departments and the receipt names seven, so the group is 16. No money moved: the grouped total of $7.10B was right, and so was every share. Only the count of departments behind it was wrongfixed
• Aug 5 · Per-resident figures — the receipt divided by “5.99M residents (est.)”. That population appeared in no source we publish. It also disagreed with the 6,012,561 the incarceration rate on this same page divides by. Every per-resident figure is now on that single Census estimate, named on the page, so the two cannot drift apart. The total moved from $8,275 to $8,243 and each department row moved by a few dollars. The budget itself is unchangedrevised
• Aug 4 · The “Get weekly updates” button — no figure changes. The button in the site header read “Get weekly updates” on every page. No weekly brief has been published on any WorksBetter site yet. So the button named a schedule that does not exist. It did so in the navigation, the one piece of copy that appears on every page and that nobody re-reads. It now reads “Get updates”, and will not name a schedule until a brief has actually shipped. Nothing about the data changed. It is logged because a promise about how often we will publish is a claim like any other. And this one was on every page.clarified
• Aug 4 · How the imprisonment-rate series was described — no figure changes. The note beneath this chart said each BJS release “prints two years side by side”. It went on: “consecutive releases overlap by one year and every overlapping year is read twice from two separate publications.” That reads as though all ten years were checked against a second document. They were not. The three biennial PDFs partition their years: 2014-15, then 2016-17, then 2018-19, with nothing bridging them. So only 2021 and 2022 are printed by two releases. The other eight years have exactly one source. The table beneath the note had it right all along, naming a single release for eight rows. The page contradicted itself, and the note was the part that was wrong. The wording now states which years were cross-checked and which were not. The table column is headed Read from rather than Confirmed by. Every rate on this chart is unchanged. We are logging it because overstating how thoroughly a figure was checked is a claim about our own work. That is what this log exists forclarified
• Aug 2 · Corrections General Fund, FY 2023-24 and FY 2024-25. The General Fund column showed $1,005,212,589 for FY 2023-24 and nothing at all for FY 2024-25. It now shows $510,212,589 and $750,267,875. Total funds are unchanged in both years. The state publishes this budget two ways. The version we had been using reports those two years before the HB24-1466 refinance. That refinance swapped roughly $495M of state General Fund onto federal pandemic-recovery money. It later switched back. Both versions state the same total; only the split differs. We now take the whole series from the one document that shows the swap. Every row's fund lines then add up to the total printed beside them in that same document. A flat General Fund line across those years was misleading. It made the state look like it kept paying for prisons out of its own pocket. For two years it largely did notrevised
• Aug 2 · Corrections appropriation, FY 2024-25 — we published $1,174,720,793. The correct figure is $1,173,002,813, a revision of $1,717,980 (0.15%). The state publishes this appropriation in two separate JBC series. They agree to the dollar on the other six years we show. We had spotted the disagreement and were showing the older figure, because the newer briefing contradicted its own arithmetic. A third document resolves it. Two consecutive Appropriations Reports state this year, both add up, and the later one restates it downward. The briefing had updated its total and left its fund lines stale. Our rule was right and our reading of it was too narrow. We compared vintages within one series instead of across both. The rule now runs across both. That changed this one figure and no other. The General Fund figure for that year is withdrawn rather than restated. The two series present the HB24-1466 refinance differently, so it is not comparable with the years either side of it. We would rather show nothing than a number that cannot be read against its neighborsrevised
• Aug 1 · Privacy charter — this site previously said “no analytics.” We added aggregate, cookieless, first-party page-view counting to watch for anomalies and learn what to improve. No cookies, no identifiers, and nothing you type is ever included. Disclosed here before the first count was collected; details in the methodologydisclosed
• Jul 13 · I-70 Vail Pass — CDOT's project page says $164.2M; CDOT press materials say $200–250M. Total updated to ≈ $200M while the two official figures are reconciledrevised
• Jul 12 · Central Maintenance Level 1 — request-stage figure replaced with the enacted appropriation of $103.2M once we verified it against HB26-1410 as passedrevised
• Jul 12 · HRWorks — reclassified from Underway to Halted after program termination was confirmed in legislative staff recordsrestated
• Jul 11 · I-25 South Gap — a paid-to-date figure briefly appeared on the wrong project's vendor table. Caught in internal review and fixed the same dayfixed