Prompt-Brush 1.0 ist ein Projekt des Grafikdesigners Pablo Delcan, das das wachsende Feld KI-generierter Kunst persifliert. Was er als „das erste nicht-KI-generative Kunstmodell“ bezeichnete, lud Delcan die Öffentlichkeit ein, Textansagen, wie einen Prompt einzureichen, die er dann manuell mit Pinsel und schwarzer Tinte illustrierte. Das Ergebnis ist als Buch erhältlich. Tolle Idee.

In response to the rise of AI-generated art, designer Pablo Delcan takes a human-centered approach in this interpretive, illustrative collection of humorous, poignant, and universal ideas based on text prompts from the public.

(Direktlink, via swissmiss)

A data hoarder on Reddit used AI to create a searchable database of more than 8,100 files about Jeffrey Epstein released by the House Oversight Committee, making it one of the easiest ways to search through a very messy batch of files. 

The project, called Epstein Archive and released on Github, allows people to search the database to find files that mention specific people, organizations, locations, and dates. 

Thousands of people are named in the more than 33,295 pages worth of files released by the House Oversight Committee last month as part of its investigation into Epstein. The files, which are partial and redacted (as in, they are not the “full Epstein files”), were subpoenaed by the committee from the Department of Justice. The files were released to the public in an extremely poorly organized Google Drive folder, and were released primarily as jpg and tif images of documents, which are not in any discernible order. Because of this, they have been an absolute nightmare to search.

The Redditor, nicko170, said they had a large language model transcribe, collate, and summarize the documents and built the database to make the files more easily searchable. 

“It processes and tries to restore documents into a full document from the mixed pages - some have errored, but will capture them and come back to fix,” they wrote. “Not here to make a buck, just hoping to collate and sort through all these files in an efficient way for everyone.”

On the project’s GitHub, they further explain how it was built:

“This project automatically processes thousands of scanned document pages using AI-powered OCR to:

• Extract and preserve all text (printed and handwritten)
• Identify and index entities (people, organizations, locations, dates)
• Reconstruct multi-page documents from individual scans
• Provide a searchable web interface to explore the archive

This is a public service project. All documents are from public releases. This archive makes them more accessible and searchable.”

The database only features OCR’ed transcripts of the files and not images of the files, though it does tell users the filename so they can go and download the actual document themselves. Like essentially all OCR and LLM projects, there are some errors. Some of the transcripts are gibberish, presumably caused by blurry or illegible type and handwriting on the source documents. But the project represents a pretty good use of AI technology, because the source documents themselves are so messy and were released in such a terrible format that is extremely time consuming to go through them. 

The database is indeed pretty usable; I was able to quickly find files that mention Donald Trump (which have already been widely reported on).

To be clear, there are no new files included in this tool, but for people who are looking to explore what has been released in a coherent, straightforward way, this is one of the better options out there. Besides releasing the database and the code for it on Github, they have also turned the entire project into a torrent file, meaning it cannot be easily deleted from the internet. 

A catastrophic breach has impacted Discord user data including selfies and identity documents uploaded as part of the app’s verification process, email addresses, phone numbers, approximately where the user lives, and much more. 

The hack, carried out by a group that is attempting to extort Discord, shows in stark terms the risk of tech companies collecting users’ identity documents, and specifically in the context of verifying their age. Discord started asking users in the UK, for example, to upload a selfie with their ID as part of the country’s age verification law recently.

“This is about to get really ugly,” the hackers wrote in a Telegram channel, which 404 Media joined, while posting user data on Wednesday. A source with knowledge of the breach confirmed to 404 Media that the data is legitimate. 404 Media granted the source anonymity to speak candidly about a sensitive incident.

Scientists have developed a unique nanoparticle vaccine that prevented the development of multiple forms of cancer in mice, reports a study published in Cell Reports Medicine on Thursday. 

Eighty percent of mice that received the novel vaccine and were subsequently exposed to cancerous cells did not develop tumors and survived to the end of the 250-day long experiment. In contrast, all of the mice that received different vaccine formulations, or remained unvaccinated, developed tumors and none survived longer than 35 days.

It’s too early to know if this breakthrough will ever be applicable to human cancer prevention or treatment, but the successful demonstration in mice is a promising result for the team’s so-called “super-adjuvant” vaccine. This approach uses nanoparticles made of fatty molecules to deliver two distinct “adjuvants,” which are substances in vaccines that enhance an immune response.

“The results that we have are super exciting, and we're really looking forward to pushing forward to the next steps,” said Griffin Kane, a postdoctoral research associate at the University of Massachusetts Amherst and first author on the paper, in a call with 404 Media. “But I think that the translation of these types of therapies from preclinical mouse models to the clinic is a very humbling experience for a lot of people and teams.”

“It’s these highlights that make it worth coming to work,” added Prabhani Atukorale, assistant professor of biomedical engineering in the Riccio College of Engineering at UMass Amherst and corresponding author on the paper, in the same call. “But I agree that the translation of these findings is key. We are not satisfied with simply publishing a paper. We want to get these into patients, and it is a humbling process because there are significant gaps.”

Scientists have been working on nanoparticle-based drug designs for decades, and the field has experienced rapid progress in recent years alongside advances in nanotechnology and drug delivery pathways. Nanoparticles provide a stable platform for carrying vaccine components to key targets, increasing the efficiency of delivery to specific sites in the body and uptake by the immune system. 

Atukorale’s team previously published a study on a similar vaccine that shrank and cleared tumors from mice. In their new study, the researchers adapted the nanoparticle design to achieve prophylactic protection from melanoma, pancreatic, and triple-negative breast cancer in mice, with support from the Institute for Applied Life Sciences at UMass Amherst, UMass Chan Medical School, and the National Institutes of Health.

Vaccines consist of two main components: antigens, which are substances that trigger an immune response, and adjuvants, which enhance the immune response. Like other cancer vaccines, the nanoparticle treatment delivers antigens that activate white blood cells in the immune system to help fight off specific types of tumors. 

What’s new in this study is that the nanoparticles accommodated two distinct adjuvants that target different immune pathways known as STING (stimulator of interferon genes) and TLR4 (Toll-like receptor 4), which further boosts the immune response to introduced cancer cells. 

Adjuvants often require very different drug delivery systems, but the nanoparticles, which are about 30 to 60 nanometers across, are big enough to house different adjuvants in their unique environments, while remaining small enough to enter lymph nodes where they can activate key immune cells.  

“The big picture is that we need better adjuvants for our vaccines,” Atukorale said. “We think that we can build them using nanoparticles. This is an example in a tumor.”

One of the most exciting surprises from the study turned out to be the prolonged protection against the spread of cancer provided by the nanoparticle vaccine. The vaccinated mice that did not develop tumors during their first exposure to melanoma cells were then later injected with new metastatic cancer cells, and their immune systems fought those off too, preventing the development and spread of the tumors.

“There's long-term robust memory immunity,” said Kane.

Moreover, while the team focused on certain cancers in their experiment, the nanoparticle platform could deliver a range of specialized antigen-adjuvant combinations to target different types of tumors. 

“We think that this is one of the true strengths of these strategies,” said Atukorale. “They will have much broader reach than many of the cancer-specific treatments out there.”

That said, Kane and Atukorale cautioned that their team’s work is still in early stages—and, of course, focused on mice and not people. They also noted that only a handful of cancer vaccines have been clinically approved out of thousands in development. While the new study represents an intriguing step forward, the dream of wide-ranging prophylactic cancer vaccines is many years away, assuming it can materialize at all. 

“A lot of very elegant technologies have come out of labs and have not fully succeeded in patients,” Kane said. “We believe that we're building this technology towards something that would improve on what current cancer vaccines are able to deliver.”

Over the last few years, some of our more meaningful (and unfortunately bleakest) reporting has been on the many ways in which the right wing has systematically targeted libraries, schools, authors, and educators over the things they teach, specifically with regard to the teaching of systemic racism, LGBTQ+ issues, science, and sex education. These targeting efforts have led to a widespread, highly successful effort to ban books, restrict curricula, harass and oust teachers and librarians, and broadly censor the educational system. This movement has leveraged these successes to seize power not just in city councils and local school boards but has succeeded in making censorship and “anti wokeness” one of the dominant political ideologies in the United States.

We have successfully gained access to public records that show, for example, how a local group in Idaho successfully got a police officer to go hunting for “obscene” books at the public library, the playbook behind getting "Drag Storytime" library events canceled, how superintendents in Florida couldn’t figure out how to comply with the state’s “Don’t Say Gay” law, and have spoken to numerous librarians, scientists, and professors to learn how educational freedom, free access to information, and historic archives are under attack. Today—which happens to be the fourth day of Banned Books Week—we are proud and excited to announce that we will be continuing and ramping up this work over the next year with the help of a grant from our friends and colleagues at government transparency nonprofit Muckrock, with support from the Filecoin Foundation for the Decentralized Web. (We’re also excited to partner with Muckrock on this new piece of limited edition merch it made for Banned Books week).

From our proposal: “Book banning and educational censorship (the banning of LGBTQIA+ studies, the study of slavery and systemic racism, the war on “DEI” and trans people) has become a political cudgel and core rallying point for the current administration. These bans have been pushed through by organized groups such as Moms for Liberty and high-profile politicians, and impact the daily lives, careers, and future prospects of students, their families, and teachers, while simultaneously managing to become a core part of the culture war. These documents about censorship are themselves difficult to obtain and are at risk of being memory holed and forgotten about without a systematic effort to obtain, publish, and archive them. This project will show how censorship works and will shed light on the sheer scale of these censorship efforts, at a time when public trust in the government is at an all-time low.”

Over the next few weeks, we will be filing hundreds of public records requests with state, local, and federal governments and school districts with the hope of unearthing more information about the groups, politicians, and monied interests that have been pushing book bans and educational censorship on American public schools and libraries. As we get these documents back over the course of the next few months, we will be making them available to the public through Document Cloud, with the hopes of creating an enduring archive of public records about educational censorship in the United States. We will also, of course, be reporting on the documents we get back and will be turning them into articles that you can read on 404 Media. 

As always, we will need some help from our readers. We need help deciding what to look for, which school districts and cities to seek public records from, and need leads on where we should point our reporting efforts. During the height of the pandemic, many city councils made their meeting minutes and meeting transcripts searchable, so we have a good sense of the types of organizations and communities that have been most severely affected by educational censorship and book bans, and have a good idea of where to get started. But if you are a librarian, teacher, educator, parent, local politician, or activist who is aware of systemic efforts to ban books, censor curricula, defund libraries, or otherwise attack educational freedom, please let us know by emailing jason@404media.co or by reaching out to Jason securely over Signal at jason.404. And if you want to further support this work, you can do so by becoming a paid subscriber or by donating to our tip jar.

The FBI accused a former National Guardsman living in Tulsa, Oklahoma of trying to sell 3D printed guns to Al Qaeda. According to an indictment unsealed by the Justice Department in September, 25-year-old Andrew Scott Hastings used a Discord server to plan a Caliphate in America and shipped more than 100 3D printed machine gun conversion kits to an undercover FBI operative who claimed he had contacts in the terrorist organization.

The Army Times first reported the story after the DOJ unsealed the charging documents. According to the court records, Hastings first landed on the radar of authorities in 2019 when a co-worker at an Abuelo’s restaurant in Tulsa called the police to report he’d been talking about blowing things up. When the cops interviewed Hastings, he told them he was just interested in chemistry and The Anarchist’s Cookbook. In 2020, the cops interviewed his mom. “Hastings’ mother, Terri, told TPD that her son was on the [autism] spectrum, was socially active online, and had converted to Islam.”

According to Terri, odd incidents piled up. She said that someone mailed Hastings a Quran, that he’d once received an order of chicken wings paid for by someone in Indonesia, and that he’d once threatened his family with a can of gasoline. “She also mentioned an incident in the family home where Hastings became enraged when she cooked bacon, and thereafter called someone she described as his ‘handler,’” according to court records.

The charging documents said the FBI got involved in 2024 because of a Discord server called “ARMY OF MUHAMMAD.” Discord cooperated with the FBI investigation and granted access to some of Hastings’ records to authorities. The FBI alleged that Hastings met with several other people on the Discord server and plotted terror attacks against Americans. At this time, Hastings worked for the National Guard as an aircraft powertrain repairer and held a SECRET-level national security clearance.

The charging documents detailed Hastings' alleged plot to establish a caliphate in the US via Discord. “[T]he most important theater right now is cyberspace…we need an actionable plan we can start work on--something slow and Ling(sic) term not hasty and slapdash,” Hastings allegedly said on Discord. “I think it would be best if we create a channel and I’ll list a physical training routine.”

“If we get 9-10 guys maybe inshaAllah we can …we could put headquarters in the USA cuz yk [you know] if we are fighting them the military is prohibited from operations on the homeland only ntnal [sic] guard and agencies can operate within borders…[y]ou need to contest air land and cyberspace…what my plan addresses is how to contest all of these at once while providing more aid than harm we can do in collateral and taking out targets of higher strength.”

According to the FBI’s version of events, Hastings talked about moving the group off of Discord and onto Signal because he believed Discord wasn’t secure. He also bragged about police interrogating him about explosives and “claimed to have made a firearm and discussed making a nuclear rocket.”

“We don’t want democracy lol,” he said on Discord, according to court records. “We want caliphate”

Hastings talked about other groups he was in contact with on Signal, offered to make training videos about weapon handling, and told others on the Discord server that he knew how to make firearms and was willing to ship them to like-minded militants. “I already have some small arms components partially finished and nearly ready to issue,” he said, according to the charging document. “I’ll send one photo but wanna remain kinda anonymous.”

The FBI said it slipped an “Online Covert Employee” (OCE) into Hastings’ life on March 26, 2025. Posing as a person on eBay, the FBI employee told Hastings he had contacts with Al Qaeda. “The OCE then recommended they move the conversation to Telegram or Signal, the latter of which Hastings said did not even have ‘a backdoor,’ meaning it could not be hacked or intercepted by law enforcement.”

The issue, of course, is that Hastings was speaking with an FBI employee. Over the next few months, Hastings spoke with the OCE about using a 3D printer to manufacture weapons for them with the eventual goal of getting them in the hands of Al Qaeda. Hastings allegedly told the OCE that he’d been discharged from the military and needed to make money.

In the summer of 2025, the FBI alleged that Hastings started mass printing Glock parts and switch conversion kits for Al Qaeda. “Hastings told the OCE he was moving out of his parent’s home in July 2025 after they complained about the noise and smell created when he 3D printed weapons,” the court documents said. The FBI allegedly has video of Hastings at a post office shipping multiple packages that summer that authorities said contained more than 100 3D printed switches, two 3D printed lower receivers for a Glock, and one 3D printed Glock slide.

The FBI has charged Hastings with attempting to provide material support or resources to designated foreign terrorist organizations and illegal possession or transfer of a machinegun. The Justice Department considers every single 3D conversion kit Hastings shipped an individual machinegun, even when they’re not installed.