404 Media

2026-01-06 20:28:53

Astronomers Discovered Something Near the Dawn of Time That Shouldn’t Exist

🌘
Subscribe to 404 Media to get The Abstract, our newsletter about the most exciting and mind-boggling science news and studies of the week.

Astronomers have discovered an ancient reservoir of gas that is too hot for cosmic models to handle, reports a study published on Monday in Nature.

By peering over 12 billion years through time to the infant cosmos, a team captured an unprecedented glimpse of a baby galaxy cluster called SPT2349-56. Cosmological models suggest that the gas strewn between galaxies in these ancient clusters should be much cooler than gas observed in modern galaxies, which has been heated up by the intense gravitational interactions that play out in clusters over billions of years.

But the new observations of SPT2349-56 reveal an inexplicably hot reservoir of this intracluster gas, with temperatures similar to those at the center of the Sun, a finding that is “contrary to current theoretical expectations,” according to the new study.

“It is a massive surprise,” said Dazhi Zhou, a PhD candidate at the University of British Columbia who led the study, in a call with 404 Media. “According to our current theory, this kind of hot gas inside young galaxy clusters should still be cool and less abundant, because these baby clusters are still accumulating and gradually heating their gas.”

“This one we discover is already pretty abundant and even hotter than many mature clusters that we see today,” he added. “So, it's a bit different and forces us to rethink our current understanding of how these large structures form and evolve in the universe.”

The first stars and galaxies emerged in the universe a few hundred million years after the Big Bang, during an era called cosmic dawn. Galaxies gradually accumulated together into large clusters over time; for instance, our Milky Way galaxy is part of the Laniakea supercluster which contains about 100,000 galaxies and stretches across hundreds of millions of light years.

As a baby cluster, SPT2349-56 is much smaller, measuring about 500,000 light years across, and containing about 30 luminous galaxies and at least three supermassive black holes. Zhou and his colleagues observed the cluster with Atacama Large Millimeter/submillimeter Array (ALMA), a highly sensitive network of radio telescopes in Chile, which allowed them to capture the first temp check of its intracluster gas.

“Because this gas is pretty distant, it's very challenging to see the light of the gas directly,” explained Zhou. To probe it, the team searched for what’s known as the thermal Sunyaev–Zeldovich signature, which is a detectable distortion of the oldest light in the universe as it passes through intracluster gas.

The results produced a thermal energy measurement of 10⁶¹ erg, which is about five times hotter than expected. While the heat source is still unknown, Zhou speculated that it could be caused by high levels of activity in the cluster, where stars are forming 5,000 times faster than in our own galaxy and huge energetic jets of matter spout out of galactic cores.

However, it will take more observations of these distant clusters to figure out whether the hot gas within SPT2349-56 is an aberration, or if super-hot gas is more common in early clusters than predicted.

“Like every first discovery, we have to be cautious and careful with big results,” Zhou said. “We need to test it further, with more independent observations and comparisons to other galaxy clusters at a similar time. This is what we hope that our community will do next, and we're also planning for follow up observations of other clusters to see whether there is a broader trend or if this system is an outlier.”

The new study is part of a wave of unprecedented observations of the early universe within the past few years. The James Webb Space Telescope, for example, has discovered massive galaxies much earlier in time than expected, pointing to a tantalizing gap in our knowledge about how our modern cosmos emerged from these ancient structures.

“It is starting to change our current understanding of how energetic the galaxy formation process was in such an early time,” Zhou said. “Galaxies were formed and evolved with much more violence, and were more active, more extreme, and more energetic than what we used to expect. The James Webb results are also consistent with our current discovery that these galaxies were very powerful in shaping their surroundings.”

Sunyaev–Zeldovich detection of hot intracluster gas at redshift 4.3 - Nature

Detection of hot intracluster gas in SPT2349−56 with the Atacama Large Millimeter/submillimeter Array provides new insights into early cluster formation.

^Nature

404 Media

2026-01-06 15:03:31

HarperCollins Will Use AI to Translate Harlequin Romance Novels

Book publisher HarperCollins said it will start translating romance novels under its famous Harlequin label in France using AI, reducing or eliminating the pay for the team of human contract translators who previously did this work.

Publisher’s Weekly broke the news in English after French outlets reported on the story in December. According to a joint statement from French Association of Literary Translators (ATFL) and En Chair et en Os (In Flesh and Bone)—an anti-AI activist group of French translators—HarperCollins France has been contacting its translators to tell them they’re being replaced with machines in 2026.
playlist.megaphone.fm?p=TBIEA2…
The ATFL/ En Chair et en Os statement explained that HarperCollins France would use a third party company called Fluent Planet to run Harlequin romance novels through a machine translation system. The books would then be checked for errors and finalized by a team of freelancers. The ATFL and En Chair et en Os called on writers, book workers, and readers to refuse this machine translated future. They begged people to “reaffirm our unconditional commitment to human texts, created by human beings, in dignified working conditions.”

HarperCollins France did not return 404 Media’s request for comment, but told Publisher’s Weekly that “no Harlequin collection has been translated solely using machine translation generated by artificial intelligence.” In its statement, it explained that the company turned to AI translations because Harlequin’s sales had declined in France.

“We want to continue offering readers as many publications as possible at the current very low retail price, which is €4.99 for the Azur series, for example,” the statement said. “We are therefore conducting tests with Fluent Planet, a French company specializing in translation for 20 years: this company uses experienced translators who utilize artificial intelligence tools for part of their work.”

According to Fluent Planet’s website, its translators “studied at the best translation universities or have decades of experience under their belt.” These human translators are aided by a proprietary translation agent Fluent Planet called BrIAn.

“When compared to standard machine translation systems that use neural networks, BrIAn can provide 2 to 3 times higher quality translations, that are more accurate, offer idiomatic phrasing, provide a deeper understanding of the meaning and a faithful representation of the style and emotions of the source text,” the site said. “BrIAn takes into account the author’s tone and intention, making it highly effective for complex literary or marketing content.”

Translation is a delicate work that requires deep knowledge of both languages. Nuances and subtleties—two aspects of writing AIs are notoriously terrible at—can be lost or deranged if not carefully considered during the translation process. Translation is not simply a substitution game. Idioms, jargon, and regional dialects come into play and need a human touch to work in another language. Even with humans, the results are never perfect.

“I will tell you that the author community is up in arms about this, as we are anytime an announcement arrives that involves cutting back on human creativity and ingenuity in order to save money,” romance author Caroline Lee told 404 Media. “Sure, AI-generated art is going to be cheaper, but it cuts out our cover artists, many of whom we've been working with for a decade or more (indie publishing first took off around 2011). AI editing can pick up on (some) typos, but not as well as our human editors can. And of course, we're all worried what the glut of AI-generated books will mean for our author careers.”

HarperCollins France is not the first major publisher to announce its giving some of its translation duties over to an AI. In March of 2025, UK Publisher Taylor & Francis announced plans to use AI to publish English-language books in other languages to “expand readership.” The publisher promised AI-translated books would be “copyedited and then reviewed by Taylor & Francis editors and the books’ authors before publication.”

In a manifesto on its website, In Flesh and Bone begged readers to “say no to soulless translations.”

“These generative programmes are fed with existing human works, mined as simple bulk data, without offering the authors the choice to give their consent or not,” the manifesto said. “Furthermore, the data processing remains dependent on an enormous amount of human labour that is invisibilized, often carried out in conditions that are appalling, underpaid, dehumanizing, even traumatizing (when content moderation is involved). Finally, the storage of the necessary data for the functioning and training of algorithms produces a disastrous ecological footprint in terms of carbon balance and energy consumption. What may appear as progress is actually leading to immense losses of expertise, cognitive skills, and intellectual capacity across all human societies. It paves the way for a soulless, heartless, gutless future, saturated with standardized content, produced instantaneously in virtually unlimited quantity. We are close to a point of no return that we would never forgive ourselves for reaching.”

The translation of the manifesto from French to English was done by the collective themselves.

Harlequin France to Test AI-Assisted Translation

Following backlash from the French Literary Translators Association, HarperCollins France confirmed it’s “conducting tests” with a French company that “uses experienced translators who utilize artificial intelligence tools,” to translate Harleq

^{PublishersWeekly.com}

404 Media

2026-01-06 14:00:35

Here is the Agreement Giving ICE Medicaid Patients' Data

A data sharing agreement between the Department of Homeland Security (DHS) and the Centers for Medicare & Medicaid Services (CMS), which was designed for ICE to receive the personal data of nearly 80 million Medicaid patients, was published as part of a lawsuit last year, with the public now able to see the exact text of that unprecedented agreement.

Last year, Freedom of the Press Foundation and 404 Media sued DHS for a copy of the agreement after the agency failed to turn it over in response to Freedom of Information Act (FOIA) requests. A U.S. attorney working on that case then flagged to our counsel that the document had been released in a separate lawsuit various states brought against the Department of Health and Human Services and DHS.

The full text of the agreement also shows the data promised to ICE included more granular data than previously reported, such as patients’ banking “routing number, account type, account number.”

“Access to this information will allow ICE to receive information concerning the identity and location of aliens in the United States, such as address, telephone number, banking information (routing number, account type, account number), email address, internet protocol (IP) addresses, or other information relevant to identifying and locating aliens in the United States,” the agreement reads.

💡
Do you know anything else about this data sharing? I would love to hear from you. Using a non-work device, you can message me securely on Signal at joseph.404 or send me an email at joseph@404media.co.

The existence of the data sharing agreement was reported at the time by the Associated Press and later WIRED. 404 Media has uploaded a copy here. At the end of December, a judge ruled that the Trump administration could resume sharing much of the data after it had been blocked from doing so, Politico reported. That means ICE can use Medicaid data in deportation cases starting Tuesday, Politico added.

Under a section called “description of the data that may be disclosed,” the agreement says that data includes “Medicaid recipients: Name, address, assigned Medicaid identification number, social security number (SSN), date of birth, sex, phone number, locality, ethnicity and race.” The data allowed to be given to ICE under the new ruling is slightly narrower than that, and includes citizenship, immigration status, address, phone number, date of birth, and Medicaid ID, and is limited to people living unlawfully in the U.S., Politico reported.

In June the Associated Press reported Medicaid officials unsuccessfully fought to block the transfer of data related to millions of Medicaid enrollees from California, Illinois, Washington state, and Washington D.C. Emails showed two top advisers to Health Secretary Robert F. Kennedy Jr. ordered the data transfer and CMS officials had 54 minutes to comply, the Associated Press added. At the time, the exact purpose of the data sharing was not known. Then the Associated Press reported on the agreement itself that said the sharing was for ICE to locate aliens in the country.
playlist.megaphone.fm?p=TBIEA2…
CMS did not respond to 404 Media’s request for comment.

DHS Assistant Secretary Tricia McLaughlin told 404 Media in a statement, “President Trump consistently promised to protect Medicaid for eligible beneficiaries. To keep that promise after Joe Biden flooded our country with tens of millions of illegal aliens CMS and DHS are exploring an initiative to ensure that illegal aliens are not receiving Medicaid benefits that are meant for law-abiding Americans.” Undocumented immigrants do not have access to federally funded healthcare coverage, including Medicaid, according to the non-partisan, non-profit organization American Immigration Council. Federal law mandates that hospitals provide emergency care regardless of the person’s immigration status, the organization says.

The agreement is part of a much wider practice of data sharing across the second Trump administration and its mass deportation effort. The IRS has funneled data to ICE; in November a court blocked that data sharing. This month the New York Times revealed the TSA was sharing multiple lists of people a week with ICE so immigration authorities could then detain them at airports.

Correction: due to an editing error, this article previously said we sued DHS earlier this year. It was last year. The copy has been updated to reflect this.

Court blocks IRS from sharing taxpayer addresses with ICE

A federal judge sided with plaintiffs who argued that the data-sharing pact violated privacy laws, putting immigrants at risk of their information being “impermissibly used” by ICE.

^{Matt Bracken (FedScoop)}