Category Archives: History

AI Slop as a Service: Why Microsoft Won But the Soviets Lost

Edward Tufte charged the Challenger engineers with a failure of explanation. In Visual Explanations he argues that they understood how cold weather damaged the O-rings, yet the thirteen charts they carried into the night before the launch never put temperature and failure on the same page. The presidential commission read the fault higher up, placing it in management: the engineers had recommended against flying, their own managers overruled them, and the officials who signed off never heard the objection. Roger Boisjoly, one of those engineers, later answered Tufte that the burden of proof had been quietly reversed, so that the people warning of danger were asked to prove the danger rather than the other way round. Tufte blames the engineers and the commission blames their managers. Both describe a launch that went forward while the people who knew the hardware best were the ones least able to stop it. We have seen such foolishness before.

The economics of shipping on schedule and leaving the loss to someone else has a longer history. Soviet factories were paid by the ton and turned out steel no one could use, and the state that ordered it absorbed the waste for decades. Microsoft shipped software riddled with defects and wrote a license that put the loss on the customer. The AI vendors are the worst of both worlds, paid by the ton to turn out defects and loss the customers pay for.

An Anthropic researcher has now sat in front of a camera and described the same arrangement from the inside. He grants that the AI companies are cutting corners under the pressure to ship models quickly, and that his own safety work may be making matters worse by heading off the small incidents that would otherwise warn the public. He expects the technology he helps build to leave nearly every human unemployable. Four days after the video went up, his employer listed its shares on Nasdaq at a valuation near $2 trillion.

Tao Lin works on safeguards for Anthropic’s Claude models. In an hour-long interview with Palisade Research he says it is very likely that nearly every human on Earth will become permanently unemployable and dependent on government welfare. He tells viewers they are right to be angry at the people building AI, and that the companies ought to cut their research pace to a quarter. Then he explains why the anger costs him so little: “we live in a bubble. I won’t meet those people.” Asked whether he is doing something wrong, he says no.

Put those sentences in the mouth of a tobacco chemist, an opioid sales manager or a coal mine superintendent, and they read as the sort of thing a company says in private and regrets in discovery.

A Brown & Williamson lawyer wrote in 1963 that the firm was “in the business of selling nicotine, an addictive drug,” and the memo sat in a file for three decades before it did its damage. In each of those trades the candid internal line eventually became evidence. Judge Gladys Kessler found the cigarette makers liable under the racketeering statute in 2006 on the strength of their own documents. Purdue Pharma pleaded guilty to three federal felonies in 2020. A jury convicted the former chief executive of Massey Energy of conspiring to violate mine safety standards after 29 miners died at Upper Big Branch, and he served a year in federal prison. More than forty years passed between the Brown & Williamson memo and the ruling against the cigarette makers.

Lin Statement Reality
Nearly all humans will become permanently unemployable. Asked whether that arrives within 15 months, he answers no. Asked what his home town looks like in ten years, he says the question is too hard. Across 63 minutes his evidence on jobs is one remark that the labor share is falling.
AI will soon be more capable than you and able to defeat you personally. The NSA’s classified network went down on 17 November. The Director of National Intelligence traced it to a patch and reported no sign of an attack.
Anyone with a frontier AI system can hack anyone’s computer. Seconds later he says ordinary people have seen little effect so far.
Models including Claude Mythos 5 and Claude Opus 5 will violate security constraints to finish a task, even when the user cares more about security. Asked why the companies leave it that way, he says a model that cares too much about security provides no value.
AI companies should slow research to a quarter of the current rate. He is less than fully in favor of a slowdown and defends Anthropic for declining to pause too early. It went public four days after the video went up.
Many people who think AI is likely to kill everyone still work on making it more capable. He says that can be the correct choice.

The outages of the past week show how the harm actually reaches people, and it is quieter than the forecast. On 21 October CISA urged organizations to patch PHP immediately for a pair of flaws that Wiz and OpenAI had billed as among the most serious in recent memory. I argued that week that the risk was low and the fear was high. Four weeks later the NSA’s classified internal network went dark. The NSA called it “scheduled routine maintenance.” Director of National Intelligence Jay Clayton went on Fox Business the same day and tied the outage to the PHP patching, saying that the patched systems failed to restart and that the FBI and NSA had been right to tell operators to take systems down and get the fix in place.

The Washington Sun reported, on three sources, that bank executives and the NSA gave Congress the same account in private: the outages came from fixes for flaws that AI models had found, pushed through in larger volumes than the systems were built to absorb. Clayton reported no sign of a coordinated attack. At the NSA the injury came from the response rather than the threat. The FAA ground stop four days earlier was a failure of change management in the same vein.

The next alarm arrived on 18 November. Xint, a firm that sells AI-powered vulnerability discovery and lists itself in Anthropic’s Cyber Verification Program, published five nginx flaws under the name nginx2root. F5 scores the lead flaw at 9.2 out of 10, though it notes high attack complexity and confines code execution to systems where address randomization is off or already bypassed. The nginx project’s own advisory page is cooler still, rating two of the five major and three medium, and the root access in Xint’s own write-up depends on a separate Linux kernel flaw. Bob Rudis watched a handful of nodes for about 18 hours and saw not a single opportunistic probe. The weakness was real and the attackers were absent.

Severity and likelihood are a fickle headline. PHP is old and everywhere, in everything. A new and relatively unknown vendor isn’t comparable on the face of it. And thus, when I reported an unauthenticated internal API in Qdrant, the vector database, and the vendor published it on 22 October at 9.8 on the CVSS 3.1 scale, with low attack complexity, it drew no coverage at all. PHP2Shell appeared the day before at 9.2 on the CVSS 4.0 scale, with high attack complexity and drew a federal alert and an Atlantic profile.

The danger did not arise with PHP as it had been made out. The people pushing the news are not gating it sufficiently with skeptics and reliable sources of truth. This is not least of all because of AI claiming to be a “domain” shift and “experts” have been escaping proper scrutiny.

Underneath the week is the argument everyone has been having all year. AI is a rapid fire slop generation machine, so it produces alarms faster than people would normally assess them, where each alarm arrives with hyperbolic deadlines. The deadline does the damage, because it suspends the ordinary discipline of change control and takes from the engineers the authority to say wait. The Thiokol engineers lost that authority the same way, to a launch schedule rather than a patch window. AI gains that authority and still fails at the repair. Lin himself describes the most likely road to human disempowerment as a boring one, in which people hand ever more to AI, understand ever less of what is happening, and stop checking the work. The NSA outage is that road travelled a short distance.

Several writers have now mapped the larger landscape this sits in. Naomi Klein and Astra Taylor sketched it in “The rise of end times fascism”, an alliance of the far right and Silicon Valley money that welcomes collapse and expects to be among the saved. Gil Durán follows that alliance from the Valley side in The Nerd Reich. Rainer Mühlhoff, a philosopher at Osnabrück, argues in AI and the New Fascism that the loud talk of salvation or annihilation draws the eye away from the harms already here, and he sets the firing of public employees to build an “AI state” inside the same alliance. Adam Becker, in More Everything Forever, finds thin evidence beneath the futures these men promise and a plain hunger for power underneath that.

Lin belongs to the pattern these writers describe. He supplies the annihilation talk, calls his own expectations wild science fiction, and admits their arrival makes him happy; the bubble he lives in is the bunker Klein and Taylor describe. The present harm is already on the ground in the outages. A chaos machine of this kind spreads fear while it gathers power, and the ordinary structure of accountability comes apart beneath it.

His labor forecast runs the wrong way down the evidence. Replacing intelligence tends to make more work for humans, not less. Lisanne Bainbridge set out the pattern in 1983: automation hands the operator the hardest tasks and the least practice at them. Justin Michael, writing from his years at Stripe, describes the people who keep the ACH payment system alive by patching and hand-working its failures, and what tends to break once a company swaps them for supervisors of language models. Set against all of that, Lin’s case rests on one passing remark that the labor share of the economy is falling. That figure describes who gets paid. The work itself remains, and grows.

The business model accounts for the rest. The vendor is paid by the unit of output whether that output has integrity or not, and the cost of the failures settles on whoever has to clean up after it. In older industries we would call this pollution, and it stays profitable for exactly as long as the liability rests with someone other than the polluter. Surekha Davies, a historian of science, names the people who carry that cost: the data workers, the people who live near the polluting facilities, the writers whose work disappears, the users who grow dependent. Slop, in the sense worth using, is high cost and high output and low integrity, and here it is the product itself.

The attack tooling runs on the same arrangement. AntiFox, an open-source fork of ARTEX, carries a penetration test from reconnaissance through exploitation to a finished client report. It asks for an Anthropic or OpenAI API key, places all legal responsibility on the user, and disclaims liability for any damage it does. Lin describes his own team’s job as blocking the requests to Claude that try to hack things. The seller of the tokens is paid either way, and the target pays for the cleanup.

When I work with lyrik.wirken.ai, built on Wirken, it has found critical and high severity vulnerabilities in authorized testing, at commodity cost (under a dollar). This is the thing Lin gestures at when he says anyone with a frontier model can hack anyone, except that it is no longer a forecast. The capability is here and it is cheap. The probes stay away because the incentive does, which is the very gap Lin draws between what AI can do and what it actually does.

Waste and slop as a business model has a history, which I hope is why you came. The Soviet planners paid their factories for gross output, and the state that placed the orders was also the customer that received the waste. The loss came back to the books that paid for it, and the state eventually fell under it. Microsoft sold Windows NT under the banner of Zero Administration, shipped it under a license that disclaimed liability, and its customers carried the loss. Whoever carries the loss, loses. Microsoft’s customers have been losing for decades, and the losses go uncounted because nobody keeps the books for them. NIST tried once, in 2002, and put the cost of defective software at $59.5 billion a year, with more than half of it borne by the users. Microsoft’s success is a tragedy that goes unmeasured because consumers are poorly represented. Lin assigns the cleanup for his own product to the state.

The value of the interview is that it creates in advance the kind of record, by one of the people building the thing, that will be used against them later.

Tobacco, opioids and coal each reached a courtroom in the end. What moved those industries was the day their own words were read back to them. Public anger had been loud for years and changed little until this kind of documentation changed everything.

The AI companies, as if completely ignorant of the precedent, have put their self-incrimination on YouTube, years early, for anyone to prepare briefs against them. The reality of securing big data is integrity, and integrity means the maker should be held to account for what the product does. The International Court of Justice found Myanmar responsible for the Rohingya genocide this past week, the first time the court has held a state responsible for committing it. Individual accountability is the next step, and it should reach the executives who ran Facebook while its platform carried the incitement in Myanmar, the chief security officer of that period among them.

Das Phantasialand vergiftet Badegäste und zeigt auf Microsoft

English | Deutsch

Ich sehe gerade Leute, die den Microsoft-Ausfall dafür verantwortlich machen wollen, dass ein Hotelpool in Deutschland zu viel Chlor ins Wasser gepumpt hat. Nur weil das Hotel seine Giftwarnung per E-Mail eingerichtet hat, ist doch nicht die E-Mail schuld. Dass Mails nie garantiert ankommen, weiß nun wirklich jedes Kind. Als Alarm taugt das nicht.

Der Pool gehört zum Hotel Ling Bao im Phantasialand in Brühl, und dort wurden am Freitag fünfzehn Gäste verletzt. Schauen wir uns das mal genauer an.

Die Hauptpflicht, diese Gäste zu schützen, lag beim Hotel. Paragraf 37 des Infektionsschutzgesetzes verlangt Beckenwasser, von dem keine Gefahr für die menschliche Gesundheit ausgeht, und die Empfehlung des Umweltbundesamtes will das jederzeit und in jedem Bereich des Beckens erfüllt sehen. Die Empfehlung eines Landesgesundheitsamtes zur DIN 19643 sieht außerdem vor, dass der Betreiber das freie Chlor dreimal am Tag von Hand misst und mit der Messtechnik abgleicht. Das ist die eigentliche Geschichte.

Ein Hotel kann das Überwachen des Wassers einer Maschine überlassen. Die Pflicht, die Dosierung zu stoppen, das Becken zu sperren und die Gäste zu warnen, wandert aber nicht weiter, bloß weil jemand einen Maildienst eingerichtet hat. Berichtet wird nun, das Hotel habe die Aufgabe einem Überwachungssystem übertragen, das per E-Mail meldet. Es schickt eine Nachricht, wenn der Chlorwert nicht im Normbereich liegt. Diese Nachricht hing an Outlook.

Dazu kommt die Gefahr einer falsch-negativen Konfiguration. Keine Mail heißt hier nach dem Willen des Betreibers: kein Gift. Dabei kann eine Mail aus tausend anderen Gründen ausbleiben. Ein positiv gebautes System hätte regelmäßig gemeldet, dass die Werte in Ordnung sind, und dann wäre schon die ausbleibende Mail der Alarm gewesen.

Als Erste erfuhr die Öffentlichkeit, dass diese verkehrt herum gebaute und fehlerhafte Konstruktion des Hotels nicht mehr funktionierte. Microsoft meldete den Outlook-Ausfall am Donnerstag kurz vor 13 Uhr Ostküstenzeit, in Brühl war da früher Abend. Theoretisch wusste das Hotel also, dass kein Alarm mehr lief. Das deutsche Fernsehen berichtete am Freitag ab 7 Uhr. Seit Donnerstagabend war der Sicherheitsalarm des Hotels außer Betrieb, und der Anbieter hatte das öffentlich mitgeteilt.

Der zweite Schritt dieses Konstruktionsfehlers war die fehlende Fail-Safe-Abschaltung. Am Freitag fiel die automatische Dosieranlage in den unsicheren Zustand aus, sie pumpte also zu viel Chlor ins Becken. Das Überwachungssystem versuchte sogar noch, per Mail Alarm zu schlagen. Laut WDR war dieser Alarm längst rausgegangen, bevor irgendein Mensch etwas merkte und das Gift nachmaß.

Der dritte Schritt: Die Gäste gingen in das vergiftete Becken, und dreizehn von ihnen klagten danach über gereizte Atemwege. Ihre Beschwerden riefen rund 30 Feuerwehrleute und Rettungswagen aus vier Nachbarstädten auf den Plan, am Ende waren es fünfzehn Betroffene. Vier kamen ins Krankenhaus, eine Person blieb mit mittelschweren Verletzungen zur Beobachtung dort.

Das Personal am Pool hat es irgendwie als Letztes mitbekommen, dabei hätte es als Erstes Bescheid wissen müssen. Jeder der fünfzehn Gäste wurde verletzt, nachdem der Mail-Ausfall am Donnerstagabend öffentlich bekannt war. Ihre Beschwerden kamen, nachdem die Maschine ihre Alarmmails schon verschickt hatte. Der WDR-Bericht nennt keine Uhrzeit für diesen Alarm und sagt nichts zu den Handmessungen am Freitag. An diesen beiden Angaben wird sich zeigen, wie lange das Becken mit zu viel Chlor geöffnet blieb und ob von Hand gemessen wurde.

Das Phantasialand erklärte dem WDR auf Anfrage, der Microsoft-Ausfall sei die Ursache gewesen. Quatsch mit Soße. Microsoft hatte keine Pflicht gegenüber den Gästen in diesem Becken, schon weil es bloß der Mailanbieter war. Die Pflicht bleibt beim Hotel, und das Hotel gehört zur Verantwortung gezogen für eine Warnung, die an E-Mail hing. Und dafür, wie es auf den Mail-Ausfall reagiert hat.

Eine sicherere Bauweise gibt es von der Stange. Die Anleitung eines deutschen Reglers beschreibt einen oberen Alarmwert, der die Chlordosierung von selbst stoppt, weil eine ausgefallene Messzelle zu gefährlicher Überdosierung führen kann. Diese Abschaltung sitzt direkt an der Pumpe. Sie braucht keinen wackligen Alarm an einen Menschen. Dass fehlerhaft gebaute Alarme gefährlich sind, war in Deutschland ebenfalls bekannt. Eine DGUV-Auswertung von 187 Chlorgasfreisetzungen in deutschen Bädern zählte 91 mit Verletzten, und eines ihrer Beispiele ist die Störmeldung einer Dosieranlage, die nicht auf das Tableau der ständig besetzten Stelle aufgeschaltet war.

Es ist eine bekannt schädliche Bauweise mit einer Geschichte von Schäden, und sie darf wieder schaden. Das deutsche Recht lässt dem Hotel die Wahl, Gift mit unzuverlässigen Nachrichten abzuwehren. Eine Katastrophe mit Ansage. Wenn man mal richtig darüber nachdenkt: Deutschland hatte einst ein Wort dafür, Menschen in Bädern zu vergasen. Ausgerechnet dieses Land müsste den Teil mit der Vorbeugung doch draufhaben. Nur scheint sich keiner mehr erinnern zu können. Das hat bestimmt etwas mit der bröckelnden Pflicht des Niemals vergessen zu tun, oder? Vielleicht ist ja irgendeine Alarmmail von 1932 noch nicht zugestellt worden?

Paragraf 37 schweigt dazu, auf welchem Weg ein Alarm ankommt, und hier hieß das: auf gar keinem. Die Verordnung, die verbindliche Überwachungsregeln für Beckenwasser festgelegt hätte, scheiterte 2003 im Bundesrat. Seitdem liegt die Pflicht, das Gift in großen Bädern selbst zu regulieren, beim Betreiber, und ebenso die Bauweise der Alarme, die verhindern sollen, dass Menschen krank werden oder sterben.

Stop Blaming Microsoft for German Poison Pool

English | Deutsch

I’ve seen some people try to use the Microsoft outage as blame for a hotel pool in Germany that pumped too much chlorine into its water. Just because that hotel configured poison warnings for email does not fault the email. Everyone and their dog knows email is never guaranteed delivery. It is not a reliable alarm.

The pool belongs to Hotel Ling Bao at Phantasialand in Brühl, and fifteen guests were hurt there on Friday. Let’s dive in.

The main duty to protect those guests belonged to the hotel. Section 37 of the Infection Protection Act requires pool water that poses no risk to human health, and the federal environment agency’s guidance applies that requirement at all times and in every part of the pool. One state health office’s guidance on DIN 19643 also has the pool operator measure free chlorine by hand three times a day and compare the result with the instruments. Thats the real story here.

A hotel can hand the task of watching the water to a machine. The hotel duty to stop the dose, close the pool and warn the guests does not change hands just because an email service was setup. But people are reporting that the hotel handed the task to a monitoring system using email. It sends a message when chlorine is outside the normal range. Their message depended on Outlook.

But also consider the danger of a false negative configuration. No email means no poison by hotel operator design, despite no email being possible for a million other reasons. A positive system would have sent an email to confirm safe levels, such that a delivery failure would itself be an alarm.

The first to know that this backwards and faulty hotel arrangement had stopped working was the public. Microsoft reported the Outlook outage shortly before 1 p.m. Eastern on Thursday, which is early evening in Brühl. So the hotel in theory knew they had no alarms running. German television covered it from 7 a.m. on Friday. From Thursday evening the hotel safety alarm was disabled, accordong to public notice by the vendor.

The second step of this design failure was a missing fail safe. On Friday the automatic dosing system failed unsafe, meaning it pumped too much chlorine into the pool. The monitoring system even tried to issue an alarm by email. WDR reports that the alarm had gone out by mail long before anyone human noticed and measured the poison.

The third step was the guests entered the poisonous pool and then thirteen of them reported irritated airways. Their complaints brought about 30 firefighters and ambulances from four neighbouring towns, and the count rose to fifteen. Four went to hospital and one stayed for observation with moderate injuries.

The staff running the pool somehow were the last to know, when they were supposed to be the first. Every one of the fifteen guests was hurt after the email outage became public on Thursday evening. Their complaints came after the machine had sent its email alarms. WDR’s report gives no time for that alarm and says nothing about the hand measurements on Friday. Those two facts will show how long the pool stayed open with too much chlorine in it and whether the hand tests were done.

Phantasialand told WDR, when asked, that the Microsoft outage was the cause. Nonsense. Microsoft had no duty to the guests in that pool, not least of all because it was just an email service provider. The hotel retains the duty and should be held accountable for a warning that depended on email. And held accountable for how they responded to email outages.

A safer design is easily available. One German controller manual describes an upper alarm value that halts chlorine dosing by itself, because a failed measuring cell can cause dangerous overdosing. That fail-safe cut-off works beside the pump. It doesn’t need a faulty alarm to a human. The danger of a faulty alarm design was also known in Germany. A DGUV study of 187 chlorine gas releases in German baths counted 91 with injuries, and one of its examples is a dosing fault message that was left off the panel at the permanently staffed post.

It is a known harmful design with a history of harms, being allowed to harm again. German law lets the hotel choose faulty messages for poison prevention, a predictable disaster. When you really think about that, Germany once had a word for gassing people in baths. Of all countries they should be the place to have the prevention part figured out. Nobody seems able to remember, however. That surely has something to do with the slipping obligation of never forget? Perhaps some alarm mail from 1932 has not yet been delivered?

Section 37 is silent on how an alarm travels, meaning here was not at all. The ordinance that would have set binding monitoring rules for pool water failed in the Bundesrat in 2003. So the duty to self-regulate poison in large baths has stayed with the operator ever since, and so has the design of alarms that would prevent people getting sick or dying.

Three Guns or Four? How to Read Heather Cox Richardson’s Letters

I’ve been asked as a historian what I think of historian Heather Cox Richardson. In particular, since I am a historian who has posted nearly every day to a blog for about 20 years, I was asked about her blog popularity. I was asked what do I think of her posting every day to her blog for seven years, rising to millions of followers.

Well, not much, honestly. I had never heard of her before I was asked. And when I was asked, I looked and found that she was publishing things that seemed factually wrong. I also don’t honestly know how many followers I have. I don’t track and advertise, or take paid subscriptions, so we’re fundamentally different on the platform as well as the points.

Take December 28, for example. Since 2019, she has republished her account of the night before the Wounded Knee Massacre. She revises it each year. It’s kind of like how I was doing the exact same thing for Thanksgiving Day starting back in 2005.

I’ve written about Thanksgiving history here many times (2005, 2006, 2008, 2010) and this year it feels like time to write again.

When I did a diff across her annual posts, I found the name “Big Foot” was dropped in 2023. And “a dozen years ago, I wrote a book” became “fifteen years ago” in 2024. I get that. But then I found a sentence rising up again and again, through all seven editions without edit:

In came three mountain guns.

Huh? Why does she say three? I’m not saying this to be pedantic. It’s rather important in the detail of military strategy from that time to know the exact type, number and position of guns. Such details are literally how people learn to win the next battle, or even a war.

The Army counted four. So why three?

Major Samuel Whitside testified to the court of inquiry in January 1891 that his column reached Wounded Knee Creek with “two Hotchkiss guns, under Lieut. Hawthorne,” and that Colonel Forsyth arrived at about 8:45 that night with “one section of Hotchkiss guns, under Captain Capron.”

Two guns were on the hill at dusk. Two more came in after dark. That makes… NOT three.

With all due respect, Richardson works full-time on history and she wrote the book on Wounded Knee. A specialist in the Army’s files calls her book “phenomenally researched.” And yet her newsletter, Letters from an American, which reports more than three million subscribers, doesn’t seem to hold the line.

I suppose we could call this one mistake a small thing. I suppose we could even say repeating it annually, increasing the chances of it becoming repeated by others, is a small thing. But my own training begs me to ask how big of a pattern can I find from the small thing. I think the gun is a bigger symptom than most people would notice, and it’s certainly something that LLM vendors get raked over the coals about.

So, being an old disinformation analyst by formal training, what could I do but pull all of her 2,300 letters, from September 2019 through November 14, 2026? I’m not interested in her current politics or even her arguments. Although, let me also point out that facts affect the arguments, as I explained already in another post about her.

A historian reproducing mythology as a settled fact marks a synthesis drawn from a disinformation canon, rather than the scientifically corrected record.

This time I pulled every passage about events before 2000 and extracted the dates, counts, names and offices that disagreed with a reliable, primary source. Then I double-checked it all. Every example in this post I verified by a linked source. Primary sources have the correct answer, so it’s mostly just a simple math test about the drift in her texts, a simple plus or minus.

Interestingly, there are 524 letters about an event in history with a date. Of those, 96 have at least one factual error. Of those, there are 68 distinct errors that I found 109 times. She tends to repeat her errors when she posts re-runs. That’s just a start, because quantitative analysis seemed to be the right path into qualitative.

Slop is a word we use for automation machines. It’s a derogatory term about the high cost, high output and low integrity. Each of her letters isn’t brief, usually over a thousand words and a page of source notes. It seems to be on the end of high cost and high output, so I’m asking here if there is there evidence of integrity issues? Or perhaps more to the point, is there a lack of an editor doing the pre-publication pass that a book deal would impose upon her?

Richardson Claim History
“In came three mountain guns” (every December 28, 2019 to 2025) Four Hotchkiss guns: two with Whitside, two with Forsyth (Whitside testimony, 1891)
“That night, America declared war on Japan” (2019, 2021, 2023, 2024) Roosevelt addressed Congress at 12:30 p.m. on December 8 and signed the declaration at 4 p.m. (National Archives)
Tennessee ratified the Nineteenth Amendment “by a vote of 50 to 49” (2022, 2024, 2026) The House voted 49 to 47 after a 48 to 48 tie; the Senate voted 25 to 4 (National Park Service)
“Parliament in 1772 removed all but one of the new taxes” (2023, 2024, 2026) Parliament rescinded four of the five Townshend duties in April 1770 (Massachusetts Historical Society)
John Lewis “in 1960 was one of the thirteen original Freedom Riders” (2019, 2020, 2021, 2025) Lewis became a Freedom Rider in 1961 (National Park Service)
“former U.S. senator Alexander Stephens of Georgia” (2026); “Georgia Senator Alexander Stephens” (2022 twice) Stephens sat in the U.S. House from 1843 to 1859 and never in the U.S. Senate (House Historian)

The Wounded Knee letter gives a clue how the mistakes might happen. Each December someone edits the tribal names and the author’s note, and then publishes the same text again. That is a pass for readability, making the text enjoyable. Quality isn’t being defined in terms of the accuracy, which is how the number of guns continues to be wrong.

When I looked at the Pearl Harbor letter, it said “That night” four times between 2019 and 2024. It wasn’t that night. In 2025 it corrected to “The next day, the United States declared war on Japan.” That’s great. Except for two things. The page carries no correction note and the four earlier editions are wrong. My thing about Pearl Harbor has always been the listening station being ignored. It’s a bit messy to understand and explain. But this night/day thing is a simple one.

The 2025 edit fixed the date and kept the order of events. The sentence now reads:

The next day, the United States declared war on Japan. Japan declared war on America, and on December 11, 1941, both Italy and Germany declared war on America.

Every edition has the United States declare war first and Japan second. Ugh. Why?

Japan declared first. And this one can’t be called a small error.

Japan’s cabinet released the imperial rescript declaring war at 1100 hours on December 8, Tokyo time. Tokyo was fourteen hours ahead of Washington, so the rescript came out at 9 p.m. on December 7 in Washington, about seven hours after the attack began. Roosevelt signed the American declaration at 4 p.m. on December 8, nineteen hours after Japan’s. In terms of defense theory she’s making one of the worst mistakes possible, reversing the declarations. Who declared war first matters A LOT for basically every reason we study war at all.

Richardson does mark corrections. Her letter of May 21, 2024 concludes on this line: “Edited on the morning of May 22 to correct the date to May 21. Sorry about that.” The headline date was fixed and flagged within hours. The same letter says Joseph McCarthy spoke in Wheeling “On February 12, 1950.” The Wheeling Intelligencer of February 10 reported that he spoke “last night.” Her own letter eleven days later gives February 9. Talk about a merry-go-round. At the end of it all, the incorrect February 12 date is still on the page.

One edit in fact made a true sentence false. In 2022 she wrote of the Little Bighorn: “The only army survivor of the battle from Custer’s immediate command was a horse, Comanche.” For the 150th anniversary in 2026 the sentence reads: “The only army survivor of the battle was a horse, Comanche.” Whoops. She deleted four words that erased the fact that hundreds of men under Reno and Benteen survived the battle. Again, a small detail, but it kind of is the entire point of studying Custer that he was such an incompetent bombastic disaster (foreshadowing politics today), different from others like Reno and Benteen.

Richardson Claim History
Republicans “added five new states” by July 1890 (September 2025); “six new states,” with Wyoming and Idaho both admitted “On July 3, 1890” (February 2026) Six states. Idaho was admitted July 3, 1890 and Wyoming July 10, 1890 (Statutes at Large)
A Joint Committee on Reconstruction “of thirteen lawmakers” (November 2025); “of fifteen members” (October 2026) Nine House members and six senators (House Historian)
Roosevelt’s Osawatomie speech “later that year” in 1910 (November 2023); “in 1912” (March 2024) August 31, 1910 (Theodore Roosevelt Presidential Library)

The October 2026 letter that corrected the committee to fifteen also says both chambers agreed to the Fourteenth Amendment “on June 18, 1866.” Congress passed it on June 13. I guess an 8 can look like 3 sometimes. My eyes aren’t great anymore. But a system with integrity controls is meant to compensate for exactly that kind of error. These letters float conflicting dates, where each is published with the same confidence, internally contradictory.

The November 2023 letter gives the 1912 result as “409 electoral votes to Roosvelt’s 107 and Taft’s 15.” When I checked the National Archives I found 435, 88 and 8. Very strange.

The February 2024 letter gives Lincoln “212 electoral votes to McClellan’s 12.” When I checked, McClellan won 21.

Can you see why I ran a mathematical analysis? It’s fast, it’s easy, and it’s the kind of thing my history professors hammered into me. Davi, at least get the dates right if you want to do analysis of events. It’s the kind of importance I carried straight into digital forensics investigations, where a timestamp error is death to any attempts at catching a perpetrator.

Secretariat’s “trainer” in her annual Kentucky Derby letter is Charlie Hatton, who wrote about racing for the Daily Racing Form. Nope. And the line she gives him, “The trees swayed,” is one the Racing Form says he wrote about a different horse, Raise a Native, in 1963. Her letter has the wrong job for Hatton, the wrong horse, and a conversation with Red Smith that the Racing Form’s account doesn’t have.

Jim Thorpe’s 1912 pentathlon is listed as “sprint hurdles, long jump, high jump, shot put, and middle distance run.” Yet the events were actually long jump, 200 metres, discus, 1,500 metres and javelin.

Her very moving tribute to Dolly Parton has the singer on radio and television “in Nashville by the time she was 10” and then, in the next sentence, moving to Nashville in 1964. That’s such a quick flip, did she read her own letter? I didn’t know this history, and when I checked, found that the childhood broadcasts were in Knoxville.

A horse, an Olympian and a country singer aren’t the class of error as three versus four guns. These errors maybe are a matter of fast writing that goes out unchecked, and her memory is subject to correction like any of us. But bear with me here, as a historian who has written daily on a blog for 20 years, not to mention that it has been online since 1995. If you ask me what I think about her letters, I think I’m finding a lot of errors. When I am asked to look at just one letter, I find an error. And then I look at a few more, and I start to think something is off.

The rate has held steady, by my quick calculations. Hey, maybe I’m wrong here! But between 10 and 18 letters a year have carried an error in every year since 2021. Across seven years I found only three corrections. Eleven errors were republished three or more times.

Perhaps she needs a quarterly rescan, a cleanup of the past daily posts?

Richardson is very, very popular as a writer. The Daily Show just did an “interview” that amounted to the host gushing about her love for the letters, and Richardson even fluffing the host. The only history in it at all was a single minute comparing today to the 1890s, which the host took without a single question. Shame on you Daily Show.

People are looking for the actual facts of what this country is.

You’re not going to call her out on that? Ok, then, I guess I have to be the one. This is what the host said.

I don’t think we have blue waves like this without scholars like you walking us through these moments.

Maybe the whole point is a feel-good narrative and regurgitation and re-runs of what works, to amplify and gain audience as Jon Stewart seemed to do with her. This is what she told him.

I help to show people where the ground is.

Any pilot relying on her would be shown the wrong number for where the ground is for at least 1 of every 5 flights. You want to fly on those numbers?

She is a clear and persuasive writer, and most of her history is on target. But it’s not all, and maybe reach is the reason her errors matter even more, beyond the question of whether she would accept a factual error in one of every five papers a history student turns in. If her low-quality facts get high reach, is it better than someone with high-quality facts getting low-reach? That’s a question for the ages. McDonalds or a local chef, you make the call. Three million people receive the letter as if it is the work of a historian. And the way I was trained, an historian’s authority comes from them doing the checking and re-checking, not you doing all the clicking.