Roughly 10 months after it was released, the Humane AI Pin, a terribly executed and terrible product is now officially a $700 brick of e-waste after the company sold its software to HP and told its dozens of customers that they are now out of luck. 

Every Humane AI Pin ever created will stop functioning at the end of the month. Well, that is not exactly correct. As Engadget has pointed out, Humane told customers that nearly every function of the AI pin will stop working on February 28, but that true diehards can continue to access “offline” features, which primarily seems to be checking whether the battery is charged or not: “After February 28, 2025, AI Pin will still allow for offline features like battery level, etc., but will not include any function that requires cloud connectivity like voice interactions, AI responses, and Center access.” Humane went on to say that “We encourage you to recycle your AI Pin through an e-waste recycling program.”

We Can, and We Must, Clown on the Humane AI Pin Forever

The Humane Ai Pin joins a rich tradition of terrible tech products that includes the Juicero, Coolest Cooler, and Magic Leap that we must remember forever.

404 MediaJason Koebler

If you are not familiar, the AI Pin is a $700 piece of junk that was supposed to be an “AI assistant” but instead barely worked, was perhaps a fire hazard, and whose main functionality was triggering fragile venture capitalists on Twitter who self-immolated when the reviews were understandably very bad.

There is very little to say about the Humane AI pin right now other than they are very lucky that the vast majority of tech journalists in the United States are too busy writing about the Elon Musk-led ransacking of the federal government to dunk on this company in the way it truly deserves (we are also doing this but need a break for five minutes). 

The company and the tech was wildly hyped, wasted gazillions of dollars (it raised $240 million in funding), made something terrible, existed for less than a year, and are now hazardous e-waste that is a huge pain in the ass to safely dispose of. The saving grace of all of this is that Humane sold so few devices (roughly 10,000) that the number of consumers who are affected is relatively low as these things go and therefore, there are fewer of them that need to be recycled. 

The Humane AI Pin is the latest in a long line of internet of things devices that cost a lot and then became e-waste when the company decided to stop supporting it or went out of business.

On recycling: I have been to electronics recycling centers, and small wearables like this are labor intensive to recycle because they have small, difficult-to-remove batteries. An iFixit teardown wondered whether Humane pin was one of the “worst devices ever,” and stated that both the Humane AI pin and the Rabbit R1, another AI wearable, “have batteries that are a pain to remove, hidden behind thoroughly glued-down panels,” and that “making the battery so difficult to reach is perplexing at best.”

Anyways, we must never forget the Humane AI Pin. Good job everyone.

A worker at the General Services Administration told colleagues in a Slack message Tuesday that they have resigned in protest after Elon Musk ally Thomas Shedd requested “admin/root access to all components of the Notify.gov system,” which is a government system used to send mass text messages to the public that contains information the worker said is highly sensitive and would give Shedd unilateral, private access to the personal data of members of the public.

Shedd is a former Tesla engineer who now runs Technology Transformation Services (TTS), a group of coders and software engineers within the GSA, who is closely allied with Elon Musk and DOGE. Notify.gov contains not just the phone numbers of everyday people but also information about whether they participate in government programs such as Medicaid, which is based on a person's financial situation. In recent days, Musk has become obsessed with the idea of "fraud" in Medicaid, Medicare, and Social Security, and in identifying those he suspects are committing fraud.

“The TTS commissioner, Thomas Shedd, has required us to provide admin/root access to all components of the Notify.gov system,” the Slack message, seen by 404 Media starts. It then says this would allow Shedd to “view all personally identifiable information (PII) moving through the Notify system, including phone numbers and variable data for members of the public.” It says Shedd “would be able to download and store this data without anybody else receiving a notification.”

The doge.gov website that was spun up to track Elon Musk’s cuts to the federal government is insecure and pulls from a database that can be edited by anyone, according to two separate people who found the vulnerability and shared it with 404 Media. One coder added at least two database entries that are visible on the live site and say “this is a joke of a .gov site” and “THESE ‘EXPERTS’ LEFT THEIR DATABASE OPEN -roro.” 

Doge.gov was hastily deployed after Elon Musk told reporters Tuesday that his Department of Government Efficiency is “trying to be as transparent as possible. In fact, our actions—we post our actions to the DOGE handle on X, and to the DOGE website.” At the time, DOGE was an essentially blank webpage. It was built out further Wednesday and Thursday, and now shows a mirror of the @DOGE X account posts, as well as various stats about the U.S. government’s federal workforce. 

Two different web development experts who asked to remain anonymous because they were probing a federal website told 404 Media that doge.gov is seemingly built on a Cloudflare Pages site that is not currently hosted on government servers. The database it is pulling from can be and has been written to by third parties, and will show up on the live website. 

Both sources told 404 Media that they noticed Doge.gov is pulling from a Cloudflare Pages website, where the code that runs it is actually deployed.

A government website created by the Trump administration to “track government waste” has been left unupdated with a default WordPress sample page that includes language about an imaginary architecture firm. 

“Waste.gov: Tracking government waste,” the tagline for the website, archived here, says. The rest of the webpage, however, is about an imaginary architecture firm called Études, pulled from a sample webpage for a default WordPress theme called Twenty Twenty-Four. 

This remains the case a day after Elon Musk, in the Oval Office, told reporters that all of DOGE’s supposed waste-cutting actions are transparent and are available on government websites. Musk is currently in charge of finding and eliminating "waste."

“We actually are trying to be as transparent as possible. In fact, our actions—we post our actions to the DOGE handle on X, and to the DOGE website. So all of our actions, which are maximally transparent,” Musk said. “In fact, I don’t think there’s been — I don’t know the case that where [sic] an organization has been more transparent than the DOGE organization.” 

The DOGE website contains the line “An official website of the United States government,” a single image of a dollar sign, the words “Department of Government Efficiency. The people voted for major reform,” and nothing else. The website for the US Digital Service, which has been renamed the US DOGE Service, has not been meaningfully updated since Trump was inaugurated. 

Waste.gov, meanwhile, says “Études is a pioneering firm that seamlessly merges creativity and functionality to redefine architectural excellence,” and various default images and text from the “Twenty Twenty-Four” WordPress theme, which is also billed as “a flexible default theme.” The theme is “suitable for everyone, from casual bloggers to creative photographers or small businesses,” the theme’s page advertises. Seemingly, the inclusive nature of this theme extends its utility to those seeking to gut the federal government. Specifically, Waste.gov is an exact mirror of the “Entrepreneur demo” for that theme.

Notably, Waste.gov does not comply with various executive orders issued by Donald Trump because it contains the word “diverse” in the line “Our comprehensive suite of professional services caters to a diverse clientele, ranging from homeowners to commercial developers.” The imaginary architecture firm also offers “a commitment to innovation and sustainability.” Research on climate change and sustainability has been purged by this administration. 

The White House registered both waste.gov and DEI.gov—which redirects to waste.gov—last week, Reuters reported.

Update: After this article was published, Waste.gov was put behind a password wall.

This week we discuss a new Microsoft study that finds using generative AI is "atrophying" people's cognition and critical thinking skills, the right's war on Wikipedia, and, in the subscriber's section, the idea of posting against fascism.

Articles discussed:

Microsoft Study Finds AI Makes Human Cognition “Atrophied and Unprepared”Wikipedia Prepares for 'Increase in Threats' to US Editors From Musk and His AlliesYou Can’t Post Your Way Out of Fascism

Subscribers-only video and embed below:

The Wikimedia Foundation is building new tools that it hopes will help Wikipedia editors stay anonymous in part to avoid harassment and legal threats as Elon Musk and the Heritage Foundation ramp up their attacks on people who edit Wikipedia. Some of the tactics have been pioneered by Wikimedia in countries with authoritarian governments and where editing Wikipedia is illegal or extremely dangerous.

Last month, Forward obtained a document created by the Heritage Foundation called “Wikipedia Editor Targeting,” which set a goal to “identify and target Wikipedia editors abusing their position by analyzing text patterns, usernames, and technical data through data breach analysis, fingerprinting, HUMINT (human intelligence), and technical targeting.” 

The document discusses creating sock puppet accounts to “reveal patterns and provoke reactions,” discusses trying to track users’ geolocation, searching through hacked datasets for username reuse, and using Pimeyes, a facial recognition software, to learn the real identities of Wikipedia editors. Molly White of Citation Needed has an extensive rundown on Elon Musk’s crusade against Wikipedia, and both Slate and The Atlantic have written about the right’s war on Wikipedia in recent days. 

In a series of calls and letters to the Wikimedia community over the last two weeks, Wikimedia executives have told editors that they are trying to figure out how to keep their users safe in an increasingly hostile political environment. “I’m keeping an eye on the rising noise of criticism from Elon Musk and others and I think that’s something we need to grapple with,” Wikimedia founder Jimmy Wales said in a meeting on January 30.

“We’re seeing an increase in threats, both regulation and litigation across the world,” Wikimedia Foundation CEO Maryana Iskander told community members during the same January 30 meeting. “We’re all just trying to understand what is happening not only in the United States [but across the world], so the best we can do is monitor, check-in on staff, and try to understand what’s needed … that’s the most honest answer I can give you to an impossible set of questions we’re all grappling with on a daily basis.”

Wikimedia lawyers told the community that the project is trying to change how editing Wikipedia for logged-out accounts works. Currently, if a user edits an article while not logged in, their IP address will show publicly, which can provide information to someone looking to file a defamation or libel lawsuit. Wikimedia is launching a “temporary accounts program” which will give editors who are not logged in a temporary username rather than showing an IP address. “It’s a way of ensuring that for logged-out users, their IP address isn’t visible to everyone asunder but rather available only to people who are really engaged in anti-vandalism,” Phil Bradley-Schmieg, a Wikimedia lawyer, said. 

Bradley-Schmieg also suggested that Wikimedia’s human rights team, which is focused on “helping users stay safe, particularly in countries where freedom of speech and expression is under attack on a regular basis,” may need to play a larger role across the entire project. 

Jacob Rogers, another Wikimedia lawyer, said during a separate meeting on January 30 that some Wikimedia projects in non-English languages have a feature where users are allowed to create and register a sock-puppet account (a dummy username, basically) to edit controversial articles and to register that account with administrators.

“A number of the different language projects have the option to make legitimate sock puppet accounts if you’re going to work on something you know is going to be controversial, you can make a sock puppet and register it with admins on that project so it’s more obscure, kept separate from the rest of your life,” Rogers said. 

Both Rogers and Bradley-Schmieg said that Wikimedia has worked to limit the amount of data that the foundation has on any given user. IP addresses associated with edits are deleted or anonymized after 90 days, for example.

“The foundation has very little data about most users, so if somebody is stepping up their harassment and coming to the foundation, we generally don’t really know anything about users in most cases and there’s not a lot they can get from us,” Rogers said. In the first six months of 2024, the last period for which data is available, Wikimedia received 26 formal requests for information on users; it provided info in two cases. Six of those requests came from the United States, the most of any jurisdiction.

Wikimedia has also created a legal defense program that will in some cases fund the defense of Wikipedia editors who are attacked through the legal system as long as that editor or staffer was contributing to a Wikimedia project in good faith, Rogers said. Wikimedia has recently fought cases in both India and Germany.

While Musk’s and the Heritage Foundation’s attacks on Wikipedia have escalated in recent days, these general trends are not new, and they were outlined as a threat in the foundation’s 2024 annual plan, which states the following: 

“Human rights threats are growing. Physical and legal threats against volunteers and staff who fight disinformation continue to grow. Accusations of bias and inaction by those whose preferred narratives do not prevail on Wikipedia may be encouraged and amplified by purveyors of disinformation,” the foundation wrote in an update to users. “Law is weaponized in important jurisdictions. Bad-faith lawsuits, by people who don’t like the verified information appearing on Wikipedia pages, are succeeding in some European countries. Some incumbent leaders are abusing their powers to silence and intimidate political opponents.”

Iskander said in the meeting that the foundation is going to consider the safety of Wikipedians for its in-person events, such as Wikimania, an annual conference and party.

“We’re paying very close attention trying to understand what the impacts might be and ensure those might be considered in any decisions we make. I will remind folks part of our processes in any event related to community gatherings is to do a risk assessment for community conferences for Wikimania,” she said. “It’s an imperfect and imprecise exercise but there’s a real intentionality around being thoughtful about the places that we’re selecting to ask people to gather and manage within our control.”

It is not clear whether any of these steps will be sufficient, or whether any of them are going to make Wikipedia more resilient to right-wing attacks. What makes Wikipedia so strong is the fact that it has a distributed global base of dedicated volunteer editors, and a governance structure that is not very easy to infiltrate. Wikimedia’s decentralized power base makes it resistant to but not invulnerable from takeover attempts.

During one of the meetings, Rogers was asked if Wikimedia would consider moving its headquarters out of the United States because of the political situation here. Rogers said moving “would probably not do very much because the projects would remain accessible in the United States and many things would still be subject to US law even if the foundation moved its headquarters to a different jurisdiction.”

“I think a move would be extremely expensive and cost something in the tens to hundreds of millions of dollars,” he said. “I see that as one of the most significant, expensive, and extreme possible options. You would only do that if it was like, the only solution to a major problem where doing that would make sense.”

The Wikimedia Foundation did not respond to a request for comment. The Heritage Foundation did not respond to a request for comment.

Donald Trump fired National Archives director Colleen Shogan Friday night, she said in a LinkedIn post and confirmed to 404 Media.

"It was an honor and privilege to serve as the 11th Archivist of the United States," Shogan told 404 Media. "I did so with integrity every day."

Earlier on LinkedIn, Shogan wrote: “This evening, President Trump fired me,” Shogan wrote. “No cause or reason was cited. It has been an honor serving as the 11th Archivist of the United States. I have zero regrets - I absolutely did my best every day for the National Archives and the American people.”

Earlier this week, ABC News reported that Secretary of State Marco Rubio has been serving as the acting Archivist of the United States “since shortly after President Trump’s inauguration.” Archives employees told me Thursday that there was no indication Rubio was involved in the Archives in any way, and Shogan gave an all-hands meeting at the National Archives Tuesday in which she said she had been working with the administration. At the time, National Archives and Records Administration officials told workers that they had not been approached by Elon Musk’s Department of Government Efficiency, which has been gutting federal agencies. 

It remains unclear who is going to run the agency, which is in charge of record keeping for the entire federal government. This is a particularly important role considering that large parts of the federal government are currently being purged or shut down, and there is uncertainty about what will happen to their records.

"At the direction of @realDonaldTrump the Archivist of the United States has been dismissed tonight," Sergio Gor, the director of the Presidential Personnel Office, tweeted. "We thank Colleen Shogan for her service."

A spokesperson for NARA declined to comment. 

While responding to the most damaging wildfires in the history of California, FEMA employees received an order “for immediate compliance” this week that states they must immediately change their vocabulary to comply with the Trump Administration’s preferred terminology on gender and immigration. 

For example, FEMA employees are no longer allowed to call undocumented immigrants “migrants” or “undocumented individuals,” they must instead call them “undocumented aliens or illegal aliens.” FEMA can no longer refer to the idea of “integration,” it must begin to say “assimilation.” 

The subject line of the email was "For Immediate Compliance."

"While the following chart presents examples of terminology that should be replaced, it should not be considered to be comprehensive, particularly in the immigration space. Please consult your program counsel for additional language if you are unsure," the email says. It then includes this table:

The “terminology changes” were sent to FEMA employees in the recovery division of Region 9 of FEMA, which includes California, Arizona, Nevada, Hawaii, and U.S. Pacific Islands. The language changes mirror those required of ICE agents and Department of Homeland Security employees (FEMA is a part of DHS). Similar notices have gone out to huge parts of the federal government, including NASA. While the most devastating wildfires in the Pacific Palisades and Altadena neighborhoods of Los Angeles are no longer burning, FEMA is busy with recovery efforts as the city tries to begin planning how to rebuild. California governor Gavin Newsom has called the fires the “worst natural disaster in US history.”

But the email obtained by 404 Media indicates that Trump’s war on words extends to the federal workers tasked with helping, for example, people in Los Angeles recover from devastating wildfires. The mission of the recovery division is “to provide assistance to individuals and communities overwhelmed by all hazards, including acts of terrorism, natural disasters, or other emergencies.”

“This is the most 1984 email I’ve ever seen,” a FEMA employee told 404 Media. “For a group that hates political correctness—this contains politically correct terminology.” 

Wednesday night, ABC News reported that Secretary of State Marco Rubio “has been the acting archivist” at the National Archives and Records Administration “since shortly after President Trump’s inauguration.” 

The idea that Rubio might be their new boss—and has been their boss, apparently, for weeks—is news to National Archives employees and apparently to its current director Colleen Shogan. Shogan held an all-hands meeting with archives employees on Tuesday in which nothing was said about Rubio and in which Shogan still seemed to be in charge. 

“The suggestion that he’s been the archivist since the transition is a lie or misunderstanding, we just had a staff meeting with Shogan Tuesday,” one NARA employee told 404 Media. “Everyone is very confused. My coworkers seem to mostly assume it’s a bad news source.”

404 Media agreed to provide anonymity to sources in this article to discuss sensitive issues. 

404 Media obtained notes from that all-hands meeting, in which Shogan was introduced as “the archivist of the United States” and in which she spoke at length about working with the Trump administration in recent weeks. Shogan is still listed as the Archivist of the United States on NARA’s website. 

“We are required by law to follow the direction of the President, and we must implement the administration’s guidance,” Shogan said. “I understand that many of you are experiencing uncertainty and pressure from these developments, and I acknowledge the times of change are not easy, but I also believe that change can be an opportunity for growth, innovation, and strategic thinking.”

A NARA employee told 404 Media that Shogan has been working with the White House in recent days and that they are not aware of any changes to her role. A theme of Tuesday’s all hands was the idea that much of the federal government is being surprised with a variety of executive orders and actions and that everyone is doing the best they can with limited information. 

Employees working for the agency now known as DOGE have been ordered to stop using Slack while government lawyers attempt to transition the agency to one that is not subject to the Freedom of Information Act, 404 Media has learned.

“Good morning, everyone! As a reminder, please refrain from using Slack at the moment while our various general counsels figure out the best way to handle the records migration to our new EOP [Executive Office of the President] component,” a message seen by 404 Media reads. “Will update as soon as we have more information!”

Another message seen by 404 Media provides an update and asserts that the US Digital Service (which is now DOGE) will “split” from the Office of Management and Budget (OMB). 

“I spoke to the DOGE team about Slack. Because of the USDS split from OMB, OMB is asking us to stop generating new slack messages starting now,” the message says. “We expect this to be a temporary pause, and we expect to continue having access to historical Slack material. We may have intermittent access as we go through this system transition so continue to use good data hygiene and backup any critical material. We will keep you updated.”

The messages indicate that, under Elon Musk’s leadership, DOGE is actively taking steps to make sure its communications and records are not subject to the Freedom of Information Act, a records transparency law commonly used by journalists and lawyers to hold government accountable. Instead, DOGE is asserting that rather than reporting up through the Office of Management and Budget as the United States Digital Service did for years, it is reporting through the Executive Office of the President and to White House Chief of Staff Susie Wiles. Under OMB, it was generally subject to FOIA. Under the White House Chief of Staff, records it creates are generally not subject to FOIA.

Thomas Shedd, a Musk-associate and now head of the General Services Administration’s Technology Transformation Services (TTS), told government tech workers in a meeting this week that the administration plans to widely deploy AI throughout the government. Shedd also said the administration would need help altering login.gov, a government login system, to further integrate with sensitive systems like social security “to further identify individuals and detect and prevent fraud,” which employees identified on the meeting as “an illegal task.”

Shedd, who is a former Tesla engineer, said the government should “try to get consent,” regarding login.gov changes but that “we should still push forward and see what we can do.”

WIRED and the New York Times previously reported on aspects of the meeting. 404 Media has now obtained audio of the full meeting and quotes it extensively below. Shedd told TTS workers that the administration would need help making radical changes to various government systems: “Things are going to get intense,” he said.

These potential changes, he said, would include things like creating “AI coding agents” that would write government software for many different agencies and would be trained in part on existing government contracts, larger scale automations of government, and, critically, changes to Login.gov. 

“Just like a fun one that we've been thinking through with Login, specifically in TTS is, as most of you know, Login can't access government information on individuals. And so there's no connection that Login has with social security or any other government system, even though we're part of the government,” Shedd said. “And so part of one of the things to work through is how do we make it so that those agencies that has that information of very secure APIs that can be leveraged by login to further identify individuals and detect and prevent fraud?” 

“I'm not saying that this is an easy task, but it is a task that's worth trying to pursue and one that only we can do as an internal team, right? We can't bring a third party in, hire them and have them work on a project like this. It has to be an internal technology team that works on this,” Shedd continued.

During a question-and-answer session, a government employee told Shedd that The Privacy Act forbids agencies sharing personal information without consent. 

A 404 Media reader has reformatted the Simple Sabotage Field Manual to be more legible as a PDF and shared it with us, so we’re going to share it with you.

Last week, I wrote about the “Simple Sabotage Field Manual,” a World War II era guide to resisting fascism for normal people. At the time, it was the fifth most popular book on Project Gutenberg, a website that hosts public domain books. It had been downloaded 60,000 times over the last month.

Currently, the book is by far the most popular on the site and has been downloaded more than 230,000 times in the last 30 days. Project Gutenberg has copies of the book as plaintext, HTML, Kindle, and EPUB formats, but doesn’t host a PDF version of the book. The original version the CIA declassified is available as a PDF, but parts of it are blurry and some of the text is difficult to read.

Over the weekend, a 404 Media reader made a new layout of the guide as a searchable PDF and sent it to me, which I thought was very cool! They asked that I not identify them, but you can find here:

“May it run wild and free,” they said. 

Code updates to a government database that helps track whether a federal program to get children ready for school at age five is actually working show software engineers are purging it of references to "forbidden words" related to DEI.

The updates, shown in Github commits, are to a database for the Department of Health and Human Services’ Head Start program. They show a project called “Remove-DEI,” which reveal some of the back-and-forth that is happening behind the scenes to align federal agencies with Donald Trump’s executive orders that forbid almost anything having to do with race or gender within federal agencies. The Github pages show software engineers discussing amongst themselves how to best remove all instances of “forbidden words” from a specific database, and the code updates they used to do it. The changes also show that, while thousands of government datasets are disappearing from the internet, even ones that remain are having parts of their utility deprecated or broken in a way that may not be visible to those outside the government.

The Office of Head Start is a government agency that spends roughly $12 billion per year to get families and children between birth and age five ready to succeed in schools, with a special focus on providing and administering grants to groups that provide assistance for “America’s most vulnerable young children.” Head Start centers were briefly impacted by Trump’s spending freeze, leading centers to worry about making payroll. 

The changes show that the U.S. government or people working on its behalf are not just manually deleting references to diversity, equity, inclusion, and accessibility (DEIA) but are also writing and tweaking code to remove references to DEI in a more blunt-force way. The HHS change is emblematic of hundreds that 404 Media has reviewed in recent days. At HHS, a recent GitHub commit details a project called “Remove-DEI” which removes the ability to search or filter in this HeadStart for information on how well programs that target “families affected by systemic discrimination/bias/exclusion” are actually working.

The changes—which are among at least hundreds across the federal government—come to a database operated by the HHS’ Office of Head Start’s Training and Technical Assistance Centers. 

This specific database is behind a government login wall, but allows government employees to search for information about grants and programs that had a focus on “Equity” and had a target population of “Children/Families affected by systematic discrimination/bias/exclusion.” 

Code in the database was tweaked to remove the ability to search or filter according to these terms. A description of the change explained on Github reads “Review the option for equity: Removal of the equity topic from the topic drop down, removal of the equity topic from all filters, Removal of the DEIA standard goal, ‘Families affected by systemic discrimination/bias/exclusion’ removes as a target population.” 

The coder also explains that they tweaked how topics are filtered in the database as a way of “making sure that when we mark a topic as deleted, it is removed from all the relevant places.” 

The coder asked their colleagues to “confirm equity has been removed from the places above. I ask also that you scan the website for other places where we need to remove the forbidden words.” The code was written by employees at a company called Ad Hoc LLC, a government contractor that works with HHS on the database. Ad Hoc is being paid $7.2 million to manage the database, according to federal records.

Ad Hoc was created in the aftermath of the HealthCare.gov launch debacle, and describes itself as “a digital services company that helps the federal government better serve people.” Ad Hoc declined to comment.

HHS told 404 Media that it is not allowed to comment: "HHS has issued a pause on mass communications and public appearances that are not directly related to emergencies or critical to preserving health. This is a short pause to allow the new team to set up a process for review and prioritization. There are exceptions for announcements that HHS divisions believe are mission critical, but they will be made on a case-by-case basis.”

Archivists Work to Identify and Save the Thousands of Datasets Disappearing From Data.gov

More than 2,000 datasets have disappeared from data.gov since Trump was inaugurated. But analyzing exactly what happened and where it went is going to take some time.

404 MediaJason Koebler

The tweak is one of hundreds that have been revealed across government via Github’s commit tracking, which shows version changes to code, websites, and other projects managed on the site. It also gives insight into how the hundreds of websites and datasets being deleted are actually being purged. WIRED reported earlier Monday that the federal government is now using scripts to forcibly remove gender pronouns from federal employee email signatures.

Elon Musk’s DOGE, the newly formed government agency aiming for drastic cuts across the U.S. government, has brought in an HR employment attorney and consultant who has spent the last few years teaching companies her “refreshing approach to diversity and inclusion” which include attempting to “redefine” DEI in a manner that she says is more consistent with Christianity and offers a “non-woke” version of HR practices, 404 Media has learned.

Stephanie Holmes is in charge of HR at DOGE, two people familiar told 404 Media. Holmes is one of many new faces at the agency, which has been rebranded from the United States Digital Service to “United States DOGE Service.” DOGE, which stands for the “Department of Government Efficiency,” has also brought in a series of employees from Musk’s other companies and asked government tech workers to show Musk’s aides their code. DOGE higher ups re-interviewed every existing employee of the US Digital Service immediately following Musk’s takeover.

A 404 Media review of Holmes’ previous speaking engagements, which touch on her perception of diversity and maintaining company culture, provide insight into what might be in store for DOGE and the federal government at large. Holmes’ association with DOGE has not been previously reported.

Holmes is the founder of an HR consulting firm called BrighterSideHR and the author of a document called the “True Diversity Toolkit,” published through the conservative Philanthropy Roundtable think tank that recommends employers define DEI as “diversity of thought” or “diversity of viewpoint” rather than through a lens of “critical race theory.” Holmes has spoken about her approach to diversity at the Federalist Society, Catholic University, and the conservative Catholic organization Napa Institute’s “Principled Entrepreneurship” conference, which has become a hotspot of conservative political organizing power.

Holmes told an audience at a Federalist Society event that she started BrighterSideHR to counter “progressive ideology” in corporate America.

“Working in the HR space and seeing the DEI efforts and progressive ideology that HR was pushing into corporate America was particularly concerning to me, and I didn’t see any other alternatives for employers in the HR space. I care a lot about these issues and saw a problem I wanted to help fix,” she said. “I left my job and started BrighterSideHR, an HR consulting company to offer an alternative kind of more values aligned space for employers.”

“I do workplace training, discrimination, harassment training, how to do workplace investigations,” she added. “It’s simply just a non-woke version, offering employers an alternative approach to diversity and inclusion.”

The BrighterSideHR website shut down in recent weeks, and now says it is “no longer active.” An archived version of the site says “We focus on employee conduct at the workplace as opposed to imposing a particular ideological viewpoint.”

At the Napa Institute’s conference panel on “Practical Steps for Dealing with DEI,” Holmes sat on a panel with former Trump administration official and current Heritage Foundation fellow Roger Severino. A moderator introduced the panel by saying “we’re here to discuss diversity, equity, and inclusion, or DEI, or as I like to put it, DIE. Many of us are quite aware of diversity, equity, and inclusion and how it has its roots really going back to Marxism.”

Holmes said on the panel that the “mainstream kind of leftist approach to DEI presents us with a lot to push back against.”

“It is really inconsistent with our faith and I also think that this presents us with an opportunity to not only say why we’re against this, why we’re opposed to mainstream DEI initiatives, but it’s important for us to be part of the conservation and to use it to say what we are for and why we have a positive vision and positive solution of DEI in a way that is consistent with our values,” she said.

She said she advises employers to “move away from defining diversity exclusively focused on employees’ race, sex, or other protected category,” and to instead focus on “bringing together employees with diverse backgrounds, viewpoints, perspectives, and beliefs to achieve common workplace goals.” She said employers need to also be “reframing the term inclusion to incorporate that in a way that’s more aligned with our faith.”

When asked whether any of the panelists “knew of a DEI program that incorporates Catholic values,” the Heritage Foundation’s Severino said “don’t use that word DEI ever again in a positive light. That phrase should be deemed toxic now.” Holmes said, however, that she has taught companies that they probably need to continue using the term because employees have asked for there to be a consideration of diversity at work. 

Larger companies must “balance how to kind of umm, play the game essentially. So I oftentimes use the term ‘diversity’ and ‘inclusion.’ I do use the term ‘equality’ instead of ‘equity’ because I think that’s particularly problematic, but I also understand that sometimes it’s just not politically feasible within a company to fully implement different terms,” she said. “I sometimes use those terms generally speaking because it’s just too politically, too much of a political hot potato to do otherwise.”

The United States Digital Service did not respond to a request for comment.

At an all hands meeting inside Meta Thursday, Mark Zuckerberg did not address Meta’s $25 million settlement with Donald Trump that will see the company paying $22 million for the eventual establishment of the Trump Presidential Library. But Zuckerberg did say that he had to be increasingly careful about what he says internally at Meta.

“Everything I say leaks. And it sucks, right?,” Zuckerberg said.

Meta made changes to the question-and-answer section of the company all hands meeting because of the leaks, Zuckerberg said, according to meeting audio obtained by 404 Media. 

“I want to be able to be able to talk about stuff openly, but I am also trying to like, well, we’re trying to build stuff and create value in the world, not destroy value by talking about stuff that inevitably leaks,” he said. So rather than take direct questions, the company used a “poll” system, where questions asked beforehand were voted on so that “main themes” of questions were addressed.

“There are a bunch of things that I think are value-destroying for me to talk about, so I’m not going to talk about those. But I think it’ll be good. You all can give us feedback later,” he added. “Maybe it’s just the nature of running a company at scale, but it’s a little bit of a bummer.”

In the hour-long meeting, Zuckerberg repeated many things he has said publicly, such as the possibility of replacing software engineers with AI, the fact that he thinks open source AI will soon overtake closed-source AI, and the fact that he believes the company can now work more easily with the Trump administration that he has changed his platforms to align with. 

Datasets aggregated on data.gov, the largest repository of U.S. government open data on the internet, are being deleted, according to the website’s own information. Since Donald Trump was inaugurated as president, more than 2,000 datasets have disappeared from the database.

As people in the Data Hoarding and archiving communities have pointed out, on January 21, there were 307,854 datasets on data.gov. As of Thursday, there are 305,564 datasets. Many of the deletions happened immediately after Trump was inaugurated, according to snapshots of the website saved on the Internet Archive’s Wayback Machine. Harvard University researcher Jack Cushman has been taking snapshots of Data.gov’s datasets both before and after the inauguration, and has worked to create a full archive of the data.

Because data.gov is an aggregator that doesn’t always host the data itself, this doesn’t always mean that the data itself has been deleted, that it doesn’t exist elsewhere on federal government websites, or that it won’t be re-hosted elsewhere. Further research will be necessary to determine what has happened to any given dataset, or to see if it turns up elsewhere on a government website. For example, 404 Media found some datasets in Cushman’s analysis that are no longer accessible on data.gov but can still be found on individual agency websites; we also found some datasets that seem to still exist because data.gov links to working websites but give a file-not-found error message when trying to download the file itself. 

Disproportionately, the datasets that are no longer accessible through the portal come from the Department of Energy, the National Oceanic and Atmospheric Administration, the Department of the Interior, NASA, and the Environmental Protection Agency. But determining what is actually gone and what has simply moved or is backed up elsewhere by the government is a manual task, and it's too early to say for sure what is gone and what may have been renamed or updated with a newer version. 

This is because data.gov doesn’t always host the data that it is indexing. Sometimes the data is hosted directly on data.gov, but other times it links to an individual agency’s website, where the data is actually hosted. This means archiving and analyzing data.gov is not straightforward.

“Some of [the entries link to] actual data,” Cushman told 404 Media. “And some of them link to a landing page [where the data is hosted]. And the question is—when things are disappearing, is it the data it points to that is gone? Or is it just the index to it that’s gone?”

For example, “National Coral Reef Monitoring Program: Water Temperature Data from Subsurface Temperature Recorders (STRs) deployed at coral reef sites in the Hawaiian Archipelago from 2005 to 2019,” a NOAA dataset, can no longer be found on data.gov but can be found on one of NOAA’s websites by Googling the title. 

“Stetson Flower Garden Banks Benthic_Covage Monitoring 1993-2018 - OBIS Event,” another NOAA dataset, can no longer be found on data.gov and also appears to have been deleted from the internet. “Three Dimensional Thermal Model of Newberry Volcano, Oregon,” a Department of Energy resource, is no longer available via the Department of Energy but can be found backed up on third-party websites. 

Determining what is gone, why it’s gone, and where it went seems like it would be straightforward, and it would seem like you could attribute all of it to malice on the part of an administration that has declared war on climate change and government equity efforts. But archivists who have been working on analyzing the deletions and archiving the data it held say that while some of the deletions are surely malicious information scrubbing, some are likely routine artifacts of an administration change, and they are working to determine which is which. For example, in the days after Joe Biden was inaugurated, data.gov showed about 1,000 datasets being deleted as compared to a day before his inauguration, according to the Wayback Machine.  

Because of the overall large number of datasets as well as the way that data.gov works, it is still too early to say what, specifically, has been deleted, though archivists and academics like Cushman are working on triaging the situation. It can reasonably be surmised that climate and environmental research and data, as well as research about marginalized communities and minorities are among the datasets that have been purged. This is in part because the Trump administration deleted huge swaths of climate data during his first term, and because Trump issued an executive order asking all federal agencies to delete anything related to diversity, equity and inclusion. 

Data.gov serves as an aggregator of datasets and research across the entire government, meaning it isn’t a single database. This makes it slightly harder to archive than any individual database, according to Mark Phillips, a University of Northern Texas researcher who works on the End of Term Web Archive, a project that archives as much as possible from government websites before a new administration takes over. 

“Some of this falls into the ‘We don’t know what we don’t know,’” Phillips told 404 Media. “It is very challenging to know exactly what, where, how often it changes, and what is new, gone, or going to move. Saving content from an aggregator like data.gov is a bit more challenging for the End of Term work because often the data is only identified and registered as a metadata record with data.gov but the actual data could live on another website, a state .gov, a university website, cloud provider like Amazon or Microsoft or any other location. This makes the crawling even more difficult.”

Phillips said that, for this round of archiving (which the team does every administration change), the project has been crawling government websites since January 2024, and that they have been doing “large-scale crawls with help from our partners at the Internet Archive, Common Crawl, and the University of North Texas. We’ve worked to collect 100s of terabytes of web content, which includes datasets from domains like data.gov.” 

The Environmental Data & Governance Institute (EDGI) published a report in 2019 detailing “How the Trump administration has undermined federal web infrastructures for climate information,” which included not just deleting datasets but also, in some cases, not deleting datasets but deleting the links to them, changing descriptions of them, or making them much harder to find. For example, during Trump’s first term, the Department of Transportation’s information on climate change was deleted, republished in a different form elsewhere, then deleted again from that new place, the report found. 

James Jacobs, a Stanford Libraries researcher who also works with a group called Free Government Information,” told 404 Media in an email that data.gov “has always been kind of a government data junk drawer (I call it that lovingly ;-)). That is, it was a really great effort to get the vast federal apparatus to start to think about collecting and preserving data. But there are no specific regulations that tell agencies that they *have to* use data.gov. Some agencies use it heavily, some put up a few excel spreadsheets and called it a day.” 

“I assume some of those datasets in data.gov have bad urls to old agency pages that no longer exist (it’s really problematic when an agency decides to redesign its site and its base domain changes and all the links to important information and data are broken),” Jacobs added. “Some of it is probably link rot and content drift and some of it is no doubt Trump admin policy driven (e.g. anything having to do with DEI).”

Harvard’s Cushman said that, because this is the internet, there are always things that are being added, breaking, changing, or vanishing, and that some of this happens on purpose and some of it happens on accident. So determining what is being purged, when there are so many data points, is not always trivial. “If you want to answer why any given thing is gone, it becomes an individual research question.” Cushman said he is working on compiling this info now and will publish it soon. 

All of this is to say that even under the best circumstances, government datasets and research can get lost or deleted, and archiving it is not always easy. When an administration specifically makes a point of deleting research, this already fragile ecosystem is stressed even further. All of these suddenly disappeared datasets must be taken in with the context that we know the Trump administration has ordered agencies to delete and edit specific webpages, and 404 Media’s own reporting has shown targeted deletions of pages relating to diversity, equity, and inclusion as well as climate change. 

In a post from this week on Free Government Information, Jacobs explained that “the government information crisis is bigger than you think.”

“There is a difference between the government changing a policy and the government erasing information, but the line between those two has blurred in the digital age,” Jacobs wrote. He explained that before the internet, government documents were printed and were archived by being distributed among many different libraries as part of the “Federal Depository Library Program.” The internet has made a lot of government information more accessible, but it has also made it a lot more fragile. 

“In the print era, libraries did a good (but not perfect) job of preservation through inertia (ie collect and catalog a document, put it on a shelf, and leave it there until a patron wanted it),” Jacobs told 404 Media in an email. “In the digital era, that system of distribution/preservation/access has broken down because digital publications are no longer ‘distributed’ to libraries, and government entities a) publish a LOT more on the internet; but b) have no clear regulations or policies regarding preservation.”

It is absolutely true that the Trump administration is deleting government data and research and is making it harder to access. But determining what is gone, where it went, whether it’s been preserved somewhere, and why it was taken down is a process that is time intensive and going to take a while.

“One thing that is clear to me about datasets coming down from data.gov is that when we rely on one place for collecting, hosting, and making available these datasets, we will always have an issue with data disappearing,” Phillips said. “Historically the federal government would distribute information to libraries across the country to provide greater access and also a safeguard against loss. That isn't done in the same way for this government data.”

In 2015, a federal worker named Katherine Spivey gave colleagues a presentation about how to “write plainly,” so that the general public can more easily understand content on government websites. One of her pieces of advice, among many, was to “use pronouns” such as the word “you” to describe the reader rather than jargon like “beneficiary” or “purchaser.”

“There’s already a great barrier between citizens and the government,” Spivey said. “Remember, your reader is a person, not an entity … use pronouns to speak directly to your readers. It requires a lot less work and it requires a lot less words.” 

Spivey’s presentation had nothing to do with gender identity, gender pronouns, diversity, equity, or inclusion. It was about the broad concept of “pronouns,” the part of speech we (a pronoun!) use constantly. And yet, after Donald Trump was inaugurated, the government webpage archiving a video of Spivey’s presentation was first edited to remove a timestamp link that went to the section of the video about “pronouns.” Later, the page archiving the video was deleted entirely (a copy of the video is still available on YouTube and on the Internet Archive).

The tweak is one of hundreds that have been revealed across government via Github’s commit tracking, which shows version changes to code, websites, and other projects managed on the site. Github is also revealing a widespread, scattershot effort to not only change government policies on DEI but also to wholesale nuke language that actually has nothing to do with it and are retroactively changing descriptions of research and events that happened in the past to remove any reference to DEI. The Github pages reveal not only the imprecision with which these changes are being made but also a willingness to literally rewrite and delete history.

Many of the deletions catalogued on Github demonstrate the pettiness and lengths to which the Trump administration is going to seek and destroy anything  that it could possibly conceive as being related to DEI. They also show that the government has hundreds of employees and contractors who have been tasked with being the anti-DEI police across the entire government. Many of the changes are frivolous, but many of them are not, and represent the destruction of critical institutions, research, and public data.

There are far more alarming deletions than Spivey’s video, of course.

The Federal Committee on Statistical Methodology, an office of the government that determines how the federal government should carry out statistical research to, for example, determine if a federal program is working, has nuked its page about best practices for researching “sexual orientation, gender identity, and sex characteristics.” This page had years of research about how to best do basic government research about the American people for the Census, the National Institutes of Health, and other government agencies to “allow for better understanding of how sexual and gender minority populations [are faring] relative to the general or other population groups, including economic, housing, health, and other differences. These insights can lead to potential resources and interventions needed to better serve the community. These data meet critical needs to understand trends within larger population groups.” 

Similarly, the National Institutes of Health deleted a page about the Sexual & Gender Minority Research Office, which has done critical research about the health and wellbeing of LGBTQ+ people. 

It is impossible to catalog everything that has been deleted, tweaked, or scrubbed. But here are some more: 

• The U.S. Web Design System has deleted its pages on inclusive web design, which many web designers referred to when thinking about how to make their websites more accessible. The Github shows that much of the research and underlying principles that went into it were also deleted.
• According to Github, a page about behavioral guidelines for government employees is slated to get rid of a bullet point that says “don’t make derogatory comments on race, color, sex, sexual orientation, gender identity,” though the change has not appeared on the live site yet. 
• GSA research about how much money the federal government spends at businesses owned by women, veterans, and other groups as been deleted from the internet
• The description of a government panel at a conference from 2022 about neurodiversity has been edited to remove the term “DEIA” from its description. Similarly, the word “inclusion” has been deleted from a government training about why the Americans with Disabilities Act is important for accessibility at sports venues. 

A declassified World War II-era government guide to “simple sabotage” is currently one of the most popular open source books on the internet. The book, called “Simple Sabotage Field Manual,” was declassified in 2008 by the CIA and “describes ways to train normal people to be purposefully annoying telephone operators, dysfunctional train conductors, befuddling middle managers, blundering factory workers, unruly movie theater patrons, and so on. In other words, teaching people to do their jobs badly.” 

Over the last week, the guide has surged to become the 5th-most-accessed book on Project Gutenberg, an open source repository of free and public domain ebooks. It is also the fifth most popular ebook on the site over the last 30 days, having been accessed nearly 60,000 times over the last month (just behind Romeo and Juliet). 

“Sabotage varies from highly technical coup de main acts that require detailed planning and the use of specially-trained operatives, to innumerable simple acts which the ordinary individual citizen-saboteur can perform,” the guide begins. “Simple sabotage does not require specially prepared tools or equipment; it is executed by an ordinary citizen who may or may not act individually and without the necessity for active connection with an organized group; and it is carried out in such a way as to involve a minimum danger of injury, detection, and reprisal.”

The guide’s intro was written by William “Wild Bill” Donovan, who was the head of the Office of Strategic Services during World War II, which later inspired the creation of the CIA. The motivating factor for writing the guide, according to a passage within it, is that citizen saboteurs were highly effective at resisting the Nazis during World War II, and the Office of Strategic Services wanted to detail other ways sabotage could be done: “Acts of simple sabotage are occurring throughout Europe. An effort should be made to add to their efficiency, lessen their detectability, and increase their number,” the guide states. “Widespread practice of simple sabotage will harass and demoralize enemy administrators and police,” the guide states, adding that citizens often undertake acts of sabotage not for their own immediate personal gain, but to resist “particularly obnoxious decrees.”

Because it was written during active wartime, the book includes various suggestions for causing physical violence and destruction, such as starting fires, flooding warehouses, breaking tools, etc. But it also includes many suggestions for how to just generally be annoying within a bureaucracy or office setting. Simple sabotage ideas include:

• “Insist on doing everything through ‘channels.’ Never permit short-cuts to be taken in order to expedite decisions.”
• “Make ‘speeches.’ Talk as frequently as possible and at great length. Illustrate your ‘points’ by long anecdotes and accounts of personal experiences. Never hesitate to make a few appropriate ‘patriotic’ comments.”
• “Bring up irrelevant issues as frequently as possible.”
• “Haggle over precise wordings of communications, minutes, resolutions.”
• “‘Misunderstand’ orders. Ask endless questions or engage in long correspondence about such orders. Quibble over them when you can.”
• “In making work assignments, always sign out the unimportant jobs first. See that the important jobs are assigned to inefficient workers of poor machines.”
• “To lower morale and with it, production, be pleasant to inefficient workers; give them undeserved promotions. Discriminate against efficient workers; complain unjustly about their work.”
• “Hold conferences when there is more critical work to be done.”
• “Multiply paperwork in plausible ways.” 
• “Make mistakes in quantities of material when you are copying orders. Confuse similar names. Use wrong addresses.”
• “Work slowly. Think out ways to increase the number of movements necessary on your job”
• “Pretend that instructions are hard to understand, and ask to have them repeated more than once. Or pretend that you are particularly anxious to do your work, and pester the foreman with unnecessary questions.”
• “Snarl up administration in every possible way. Fill out forms illegibly so that they will have to be done over; make mistakes or omit requested information in forms.”

The guide also suggests “general devices for lowering morale and creating confusion,” which include “Report imaginary spies or danger to the Gestapo or police,” “act stupid,” “Be as irritable and quarrelsome as possible without getting yourself into trouble,” “Stop all conversation when axis nationals or quislings enter a cafe,” “Cry and sob hysterically at every occasion, especially when confronted by government clerks.”

It is impossible to say why this book is currently going viral at this moment in time and why it may feel particularly relevant to a workforce of millions of people who have suddenly been asked to agree to be “loyal” and work under the quasi leadership of the world’s richest man, have been asked to take a buyout that may or may not exist, have had their jobs repeatedly denigrated and threatened, have suddenly been required to return to office, have been prevented from spending money, have had to turn off critical functions that help people, and have been asked to destroy years worth of work and to rid their workplaces of DEI programs. Maybe it's worth wondering why the most popular post in a subreddit for federal workers is titled “To my fellow Feds, especially veterans: we’re at war.” 

The narrative that OpenAI, Microsoft, and freshly minted White House “AI czar” David Sacks are now pushing to explain why DeepSeek was able to create a large language model that outpaces OpenAI’s while spending orders of magnitude less money and using older chips is that DeepSeek used OpenAI’s data unfairly and without compensation. Sound familiar?

Both Bloomberg and the Financial Times are reporting that Microsoft and OpenAI have been probing whether DeepSeek improperly trained the R1 model that is taking the AI world by storm on the outputs of OpenAI models. 

Here is how the Bloomberg article begins: “Microsoft Corp. and OpenAI are investigating whether data output from OpenAI’s technology was obtained in an unauthorized manner by a group linked to Chinese artificial intelligence startup DeepSeek, according to people familiar with the matter.” The story goes on to say that “Such activity could violate OpenAI’s terms of service or could indicate the group acted to remove OpenAI’s restrictions on how much data they could obtain, the people said.”

The venture capitalist and new Trump administration member David Sacks, meanwhile, said that there is “substantial evidence” that DeepSeek “distilled the knowledge out of OpenAI’s models.” 

“There’s a technique in AI called distillation, which you’re going to hear a lot about, and it’s when one model learns from another model, effectively what happens is that the student model asks the parent model a lot of questions, just like a human would learn, but AIs can do this asking millions of questions, and they can essentially mimic the reasoning process they learn from the parent model and they can kind of suck the knowledge of the parent model,” Sacks told Fox News. “There’s substantial evidence that what DeepSeek did here is they distilled the knowledge out of OpenAI’s models and I don’t think OpenAI is very happy about this.” 

I will explain what this means in a moment, but first: Hahahahahahahahahahahahahahahaha hahahhahahahahahahahahahahaha. It is, as many have already pointed out, incredibly ironic that OpenAI, a company that has been obtaining large amounts of data from all of humankind largely in an “unauthorized manner,” and, in some cases, in violation of the terms of service of those from whom they have been taking from, is now complaining about the very practices by which it has built its company. 

The argument that OpenAI, and every artificial intelligence company who has been sued for surreptitiously and indiscriminately sucking up whatever data it can find on the internet is not that they are not sucking up all of this data, it is that they are sucking up this data and they are allowed to do so. 

You can see the specific steps that a government agency is taking to comply with the Trump administration’s policies against diversity, equity, and inclusion on the agency’s GitHub, which shows it frantically deleting and editing various documents, employee handbooks, Slack bots, and job listings across everything the agency touches. 

18F is a much-hyped government agency within the General Services Administration that was founded under the Obama Administration after the disastrous rollout of Healthcare.gov. It more or less had the specific goal of attracting Silicon Valley talent to the federal government to help the government innovate and make many of its websites and digital services suck less. It is one of the “cooler” federal agencies, and has open sourced many of its projects on GitHub.

GitHub is a website for open source development that shows changes across different “commits,” or changes to code and documentation. In the first days of the Trump administration, 18F’s commit list is full of change logs detailing the administration’s attempts to destroy the concept of diversity, equity, and inclusion. 

The changes show that in the last 48 hours, 18F has edited text and wholesale deleted both internal and external web pages about, for example “Inclusive behaviors,” “healthy conflict and constructive feedback,” “DEIA resources,” and “Diversity, equity, inclusion, and accessibility.” It deleted a webpage about “psychological safety” (which now 404s) deleted all information about the “DE&I leads” at the agency, as well as language for employees that said "Anyone who has issues or concerns related to inclusion or equity in the 18F engineering chapter should feel empowered to reach out to the DE&I Leads.” It has deleted, in various places, the word “inclusion,” as well as the term “affinity groups.” 

It also deleted an internal Slack Bot called “Inclusion Bot,” which is described as being “integrated into Slack and passively listens for words or phrases that have racist, sexist, ableist, or otherwise exclusionary or discriminatory histories or backgrounds. When it hears those words, it privately lets the writer know and offers some suggested alternatives.” 

It has also notably deleted information intended for improving accessibility for blind and visually impaired employees, which asked employees to use “visual descriptions” when introducing themselves on Zoom meetings.

In a hiring document, the language “Teams should consider factors of equity and complexity of the research when determining compensation for participants on their project” has been changed to “team should consider other factors or complexity of the research.”

The Trump administration has not tried to hide that it is trying to delete web pages and employee information across the government. But seeing the change logs pop up as they’re happening on GitHub shows exactly how these changes are being done and how they’re rolling out.