For Governing, Jule Pattison-Gordon on how librarians are handling the slop trying to enter the public catalog.

When you get a suspicious book, “you look it up to see what else the [author] has written. And you see there’s 30 things, and none of them have any reviews, and they were all written in the last two years,” DeMeester-Lane says. Typically, AI-generated books list an author with little to no digital footprint. And while a small or lacking online presence could indicate a new author just starting out, it takes humans time to write. One book every year or so is normal; dozens of titles per year start looking quite suspicious.

From WikiProject AI Cleanup, a guide on spotting fake writing on Wikipedia.

This list is not a ban on certain words, phrases, or punctuation. No one is taking your em-dashes away or claiming that only AI uses them. Not all text featuring the following indicators is AI-generated, as the large language models that power AI chatbots are trained on human writing, including the writing of Wikipedia editors. This is simply a catalog of very common patterns observed over many thousands of instances of AI-generated text, specific to Wikipedia. While some of its advice may be broadly applicable, some signs—particularly those involving punctuation and formatting—may not apply in a non-Wikipedia context.

More on em-dashes:

While human editors and writers often do use em dashes (—), LLM output tends to use them more often than nonprofessional human-written text of the same genre, and uses them in places where humans are more likely to use commas, parentheses, colons, or (misused) hyphens (-). LLMs especially tend to use em dashes in a formulaic, pat way, often mimicking “punched up” sales-like writing by over-emphasizing clauses or parallelisms. LLMs overuse em dashes because they were trained (sometimes illegally) on novels, and novelists have always used em dashes more often than is typical of a layperson.

This sign is most useful when taken in combination with other indicators, not by itself.

I think I’ve been subconsciously using more commas these days.

The rapper Curtis Jackson became 50 Cent a couple of decades ago, but the dollar is not what it used to be. Accounting for inflation, Brian Moore shows Jackson’s current value, which is 109 Cent, as of August 2025.

One might argue the other direction where 50 Cent stayed static name-wise, in which case he is currently 50 Cent and rewinding back to 1994, he would be 23 Cent.

These are the day’s tough decisions.

Om Malik interviewed Rodney Brooks, the founder of iRobot and former director of the MIT Computer Science and Artificial Intelligence Laboratory.

You can do so much more computation, sensing, some actuation, but people underestimate the long tail of the natural environment. That’s what we see with autonomous vehicles. I first attended a talk on autonomous vehicles in 1979 in Tokyo. By 1990, Ernst Dickmanns in Germany had his truck driving on the Autobahn at 100 kilometers an hour. He took it to Paris, and an autonomous vehicle drove around Paris in 1990. Then in 2007, 2008, people saw the DARPA autonomous vehicle and said, “Oh, it’s going to be everywhere instantly.” But it’s taken almost 20 years, and it’s still only in little tiny geographical areas because of the long tail of all the things that can happen.

There’s a tendency to go for the flashy demo, but the flashy demo doesn’t deal with the real environment. It’s going to have to operate in the messy reality. That’s why it takes so long for these technologies.

Historical context and the reality of building things for a physical world is almost a shocking point of view these days. It usually feels like we’re either headed towards the end of the world or some kind of automated utopia. Brooks offers optimism, but a view in between doom and the box of dreams.

This week we sort the dots and look closer at the things we can control.

There’s been a lot going on and we almost forget that the federal government is approaching its third week of shutdown. The administration has used the time to cancel and pause billions in grants in the places you might guess. Tony Romm and Lazaro Gamio for the New York Times have the analysis.

The Times conducted its analysis by examining federal funding records, which include details about the city and state where each grant recipient is based. The projects include new investments in clean energy, upgrades to the electric grid and fixes to the nation’s transportation infrastructure, primarily in Democratic strongholds, such as New York and California.

For the Washington Post, Craig Whitlock, Lisa Rein, and Caitlin Gilbert examine court records and disability claims.

“One inherent problem [VA] must contend with, is that it operates on an honor system with its veterans,” Justice Department prosecutors wrote in a 2021 court filing for a case involving an Army veteran who defrauded VA of more than $200,000. “… The result is that some veterans can, and unfortunately some veterans do take advantage.”

While most claims are probably legit, there seems to be an alarming increase over the last few years for hay fever, irritable bowel syndrome, and penile deformity.

When there is an excess of solar energy during the day, it can be stored in batteries for use in the evening when the sun goes down. Financial Times, in a chart forward piece, describes how this applies at scale, using California as a prime example.

Here in California it’s become more common to have home batteries than not in some places. During really hot days, you can send energy to the grid and receive reimbursements, which is nice. But the main selling point for me is when there is an outage. The battery kicks in and you don’t even notice.

I’m a fan. Of the battery approach and the hatching in the FT charts.

The cost of driving an electric vehicle varies by where you charge and how much drive. It’s a lot cheaper to charge a car at home during off hours than it is to charge on the road during peak times. For NYT’s the Upshot, Francesca Paris made a straightforward calculator to estimate the true cost of going electric versus driving a gas car.

The big AI companies are making deals with each other, promising and distributing hundreds of billions of dollars over the next few years. It’s difficult to keep track, but Bloomberg has this network diagram that shows the moves.

And from today, add Broadcom to the mix dealing with OpenAI, because sure why not.

Everyone loves reading about millionaires who don’t have enough liquidity to buy things, but the seven-figure status is not what it once was. For Bloomberg, Andre Tartar, Ben Steverman, and Stephanie Davidson show the increasing number of millionaire households who have most of their net worth tied up in housing and retirement accounts.

The mosaic plot above breaks down assets by millionaire income groups. About half of households fall within the $1M to $2M range and only 17% of their net worth counts as liquid assets (in green). The liquid share increases with net worth.

Cue the violins.

Parlays in sports betting are booming and people are losing billions of dollars to online sportsbooks, like FanDuel and DraftKings. For the Washington Post, Emily Giambalvo, Kati Perry, and Aaron Steckelberg show how the bets work with playable simulations.

Like most gambling games, the odds are completely against you in the long run and virtually guarantee that sportsbooks get all your money.

So instead of tossing your money into the sports machine, you can just run simulations. It’s not nearly as fun, but at least you’ll be able to watch sports again for the spirit of competition and love of the game.

Kurzgesagt, popular for illustrative explainer videos on YouTube, describes the current situation with generative AI slop. There is a flood of random, fake videos on the rise.

It is all fun and games at first, but when there is no way to distinguish between fake and real online, this useful internet thing turns into something else.
Read More

Visualization has been around for a few minutes, and while the tech, format, devices, and mediums change, we can still apply classic patterns in these new places.

Our World in Data compared causes of death in the United States against how much those causes are covered by the New York Times, Washington Post, and Fox News. The results are about what you would expect, based on coverage data from Media Cloud.

Rarer events, like homicide and drug overdose, are reported more heavily, whereas everyday causes, like cancer and heart disease, are reported less.

This, of course, is because the news covers things that are out of the ordinary, which is what readers and viewers are looking for on a by-the-minute timeline. Not many people care that mortality rates, which take more time to estimate than reporting on single events, are the same as yesterday.

But, as residents of this planet, it is beneficial to know that life is not always getting worse. It’s good to get a reminder sometimes.

Matthew Inman, the cartoonist who illustrates The Oatmeal, breaks down his feelings towards AI-generated art. When you remove the work, pain, mistakes, intention, and ultimately, the person from the art, the true value is lost.

OpenAI released Sora, which lets users easily generate videos with a prompt. The videos present in a TikTok-like feed and provides us with another source of endless scroll. Thank you, internet gods for your benevolence. Casey Neistat, known for his YouTube-ing, voices his concerns.
Read More

As many of us have learned first-hand, TikTok (and its scrolling video ilk) is addictive and can pull people in to stare at their phones for hours. However, TikTok is secretive about how their viewing algorithm works. So the Washington Post recruited readers to track usage and send the data, which gave WaPo a sample of how the infinite scroll kept people watching over a six-month period.

New users start with general music videos and the feed grows more specific. A few minutes per day can easily increase to hours. Your brain craves the dopamine all the time. It gets harder to interact with others. You can’t concentrate on bigger tasks. All in all, pretty amazing.

Check out the full project, and then maybe delete TikTok.

Dell Cameron, for Wired, on the system currently in its planning phase:

Throughout, ICE has leaned on Palantir’s Investigative Case Management system to combine disparate streams of data into a single investigative platform. Recent contract updates show the system lets agents search people using hundreds of categories, from immigration status and country of origin to scars, tattoos, and license-plate reader data. Each surveillance contract ICE signs adds another layer—location trails, social networks, financial records, biometric identifiers—feeding into Palantir’s hub. ICE’s new initiative is about scaling up the human side of the equation, stationing analysts around the clock to convert the firehose of data into raid-ready leads.

I hope people really are realizing that our gift of data and attention to social media is not in our best interest.