A flaw in a widely used parking ticket system software gave a programmer access to the locations of thousands of citations – and their details. CC-licensed photo by Rachel Knickmeyer on Flickr.
You can sign up to receive each day’s Start Up post by email. You’ll need to click a confirmation link, so no spam.
A selection of 9 links for you. Not moving. I’m @charlesarthur on Twitter. On Threads: charles_arthur. On Mastodon: https://newsie.social/@charlesarthur. On Bluesky: @charlesarthur.bsky.social. Observations and links welcome.
Anthropic ‘destructively’ scanned millions of books to build Claude • The Washington Post
Aaron Schaffer, Will Oremus and Nitasha Tiku:
»
In early 2024, executives at artificial intelligence start-up Anthropic ramped up an ambitious project they sought to keep quiet. “Project Panama is our effort to destructively scan all the books in the world,” an internal planning document unsealed in legal filings last week said. “We don’t want it to be known that we are working on this.”
Within about a year, according to the filings, the company had spent tens of millions of dollars to acquire and slice the spines off millions of books, before scanning their pages to feed more knowledge into the AI models behind products such as its popular chatbot, Claude.
Details of Project Panama, which have not been previously reported, emerged in more than 4,000 pages of documents in a copyright lawsuit brought by book authors against Anthropic, which has been valued by investors at $183bn. The company agreed to pay $1.5bn to settle the case in August, but a district judge’s decision last week to unseal a slew of documents in the case more fully revealed Anthropic’s zealous pursuit of books.
The new documents, along with earlier filings in other copyright cases against AI companies, show the lengths to which tech firms such as Anthropic, Meta, Google and OpenAI went to obtain colossal troves of data with which to “train” their software.
…In June, District Judge William Alsup found that Anthropic was within its rights to use books for training AI models because they process the material in a “transformative” way. He likened the AI training process to teachers “training schoolchildren to write well.” The same month, District Judge Vince Chhabria found in the Meta case that the book authors had failed to show that the company’s AI models could harm sales of their books.
But companies can still get in trouble for how they went about acquiring books. In Anthropic’s case, the book-scanning project passed muster, but the judge found that the company may have infringed on authors’ copyright when it downloaded millions of pirated books free before launching Project Panama.
«
This is the problem that publishers and authors face: judges keep saying that the process of ingesting the books as “training” isn’t, per se, illegal, because what is done is transformative. Seems hard to argue against.
AI: the wrong kind of bubble • Breadcrumb.vc
Sameer Singh:
»
This post is bit of a history lesson, but an important one. Lately, there has been a lot of discussion about “productive” bubbles in technology — manias that focus capital and talent around a vision of the future. The argument is that bubbles create critical infrastructure that entrepreneurs leverage after the crash — the dotcom bubble is a key example. This is an oversimplified view of bubbles. All technology bubbles are not necessarily “productive”. They don’t always create infrastructure that is used after the bubble. Let’s take a look at whether this applies to the current AI bubble — and it is one, despite real utility.
…Why do I think AI is part of the maturity phase and not a whole new cycle? Simple, it carries all the hallmarks of a maturity innovation (h/t to Jerry Neumann) — the biggest incumbents in the technology world immediately jumped on it and adoption was lightning fast as the internet is already fully diffused. Whole new technology cycles begin on the fringes with hobbyists — like unknown nerds creating the Apple I. That is absolutely not what we’re seeing with AI.
A key takeaway from this model is that technology diffusion causes not one, but two types of financial excess — a mid-cycle bubble during the Frenzy phase and a late-cycle bubble during the Maturity phase. The two types of bubbles are strikingly clear in the chart below [in original article]. The chart shows the Shiller PE Ratio, also known as the Cyclically Adjusted PE Ratio (CAPE Ratio), for the S&P 500 over the past 150 years. This ratio swaps out one year earnings (the E in the PE Ratio) with 10 year average inflation adjusted earnings to smooth out temporary fluctuations and account for business cycles (earnings can also be inflated during bubbles).
…a late-cycle (Maturity) bubble looks very different. At this stage, the technology has won and the dominant companies of the era are so entrenched that they are thought to be infallible. It is this belief that leads to another wave of speculative excess and overinvestment.
Any new technology breakthroughs are adopted at lightning speed. The incumbents begin to invest aggressively to harness its potential and strengthen their positions. The strong cash position of established companies leads to a heavy amount of financial engineering. Exuberance also leads to risky debt issuances. This is exactly what is happening with AI — rapid adoption, aggressive investments from incumbents like Google, financial engineering with Nvidia investing in companies that buy its products, and companies like Oracle and Coreweave taking on unprecedented levels of high-risk debt to finance data center construction. The presence of companies with real earnings makes this bubble less violent on the upswing, but no less speculative.
«
OpenAI aims to ship its first device in 2026, and it could be earbuds • TechCrunch
Ivan Mehta:
»
Last November, Sam Altman described the potential device to be more “peaceful and calm” than iPhones. Previous reporting suggests the company wants to build a screen-free and pocketable device.
While the company is not spilling any details, more recent reporting from Asian publications and leakers suggests OpenAI’s first device could be a pair of earbuds. According to reports, this device is codenamed “Sweet Pea” and will have a unique design as compared to existing earbuds. The earbuds could work on a custom 2nm processor and handle AI tasks locally instead of sending requests to the cloud.
A separate report from a large Taiwanese newspaper noted that OpenAI was exploring a partnership with China-based Luxshare for manufacturing, but might eventually lean in favor of Taiwan’s Foxconn. The report also said in the first year of sales, OpenAI aims to ship 40 to 50 million units.
OpenAI’s ChatGPT has nearly a billion weekly users, but the company has to rely on other devices and platforms for distribution. With its own device, it might want to take more control of the development and distribution of the AI assistant and also release exclusive and purpose-built features.
However, replacing existing earbuds like AirPods in users’ daily lives is going to be challenging if there’s not a strong integration with operating systems.
Until now, there hasn’t been a standout AI device success story.
«
It still feels a little too early for AI device success stories, not because the tech isn’t advancing fast, but because errors are always going to be part of their response (doing nothing is actually better than doing the wrong thing) and because it’s going to be hard to get them to scale if people are wearing and using them all the time.
unique link to this extract
Does evidence even matter? • On my Om
Om Malik:
»
A smile is a smile in Chinese, Hindi, Portuguese, Spanish, and English. The emotional payload is universal. The visual web was emerging as a universal language. Photos were becoming the atomic unit of social platforms. I was excited. Millions of vantage points were creating a collective sight. I asked: How can we create a way for visuals to tell the near history of our time?
Minnesota is answering that question, and I don’t like what I see.
Cameras are everywhere. More than a billion surveillance cameras are installed worldwide. Last year, 1.2 billion smartphones with cameras were shipped. Two trillion photos are taken each year. In Minneapolis bystanders filmed everything. Reuters verified the footage, as did the BBC, the Wall Street Journal, and the Associated Press. The visual record is unambiguous. It contradicts the official story completely. Yet the official story continues.
I was right about cameras democratizing witnessing. Anyone can document reality. We bypassed gatekeepers. But I was wrong about accountability. I thought seeing would create it. I thought evidence would force consensus. That shared visual reality would make it harder to lie. The assumption was simple: if enough people saw the same thing, power couldn’t ignore it. Minnesota proved me wrong.
I keep thinking about The Circle, a wonderful book by Dave Eggers. It imagines a world dominated by one company, loosely modeled on Facebook, where cameras record everything and no one even pretends otherwise. “Secrets are lies,“ the company insists, and total transparency is sold as moral progress. This is the kind of nonsense I believed in my younger days.
In The Circle, a man is hunted in real time by crowdsourced tracking. The footage is not denied. It is reinterpreted, and it eventually loses moral weight. Truth does not disappear; it drowns in commentary, metrics, and ephemera. In 1984, George Orwell wrote, “The Party told you to reject the evidence of your eyes and ears.” Together, Eggers and Orwell reach the same unsettling conclusion. The crisis is no longer whether technology can show us what happened; it is whether society is still willing to believe its own eyes.
«
Our tribalism outweighs our rationality. It always has, but it has now reached a pitch where a few people can force our tribalism to do that again and again.
unique link to this extract
All Your Parking Tickets Are Belong to Me • Jack’s Blog
Jack Lafond realised there was an API which would let him figure out where every parking ticket in the US issued using a particular company’s software (Passport Parking) was issued. After being ignored, he decided to try to attract their attention by.. showing it:
»
Like any great chaos engineer, I decided to build a frontend to visualize the data, which is the magnum opus of this entire research project: parking.exposed is now live and visualizes a great deal of tickets that I’ve been able to collect over the last week or so. After fiddling around with ticket IDs I’ve managed to get within a 6 to 12 hour range of recently written tickets.
There’s a great heatmap visualization that allows you to see where tickets are primarily clustered. Unfortunately, ticket operators aren’t required to put in the location of where tickets are written, which meant I had to use some geocoding magic to try and figure it out. I’d say a good percentage are correct, but if you see big clusters in random areas (like the water) don’t blame me!
My personal favorite feature is the Pay Now button. Because I have the data of the city and the associated payment subdomain, you theoretically could pay someone’s parking ticket for them!
The site’s is live and is constantly refreshing, so if you want to see tickets stream in on the stats page that’s absolutely an option. It was such a blast building this and really puts into perspective the magnitude of this issue.
In an ideal world, someone from Passport Parking will reach out to me after this is published and I can finally get a response from them on working to fix the issue.
«
So far (five days later) no update, but plenty of time.
unique link to this extract
China’s labs pull ahead as global drugmakers invest in biotech pioneers • Financial Times
Aanu Adeoye and Patrick Temple-West:
»
Investors in western biotechs face the prospect of lower valuations as Chinese start-ups attract growing investment from global drugmakers looking to replenish their pipelines.
China has emerged in recent years as a hub for drug development, particularly early-stage candidates, with faster timelines allowing companies to reach proof of concept ahead of western rivals.
Because the country’s biotechs can run clinical trials more quickly and cheaply, rivals from elsewhere risk being “undercut in licensing or partnering discussions”, said Oliver Kenyon, senior director at life sciences investor RTW.
This was particularly the case for “fairly crowded therapeutic areas”, he added, noting the trend “might compress long-term returns” for investors.
“Chinese biotech clearly represents a structural shift in global drug development,” Kenyon said. “We don’t think it’s a cyclical phenomenon . . . they’re here to stay.”
Historically, China was known for its ability to quickly replicate new drugs developed elsewhere.
While that is still the case, Chinese groups are also developing new therapies. They are taking the lead, for example, on antibody-drug conjugates, which use antibodies to deliver chemotherapy in a much more targeted way.
Chinese companies account for more than half of new ADC drugs in early clinical trials, according to consultancy McKinsey. They are also making strides in developing next-generation Car-T treatments for autoimmune diseases, and small interfering RNA (siRNA) therapies. These can temporarily turn off harmful genes.
«
That point about running clinical trials faster and cheaper is unexplained, but has a hint of not quite going by the regulatory principles used in the West.
unique link to this extract
It’s time to think about 6G – yes, really • Light Reading
Anne Morris:
»
Right up until the very end of 2025, Red Hat’s Fran Heeran said he was still apologizing whenever he mentioned 6G in a conversation because there was “always that belief in the room that, yes, this is premature. Why … are we mentioning 6G when … we’re still going through the 5G Advanced phase?” – and what about the business case, too?
Now, he appears certain that 6G will become a mainstream conversation in 2026 and 2027, “whereas up until now, perhaps it was a little bit scary, given where we were with 5G.”
Heeran, a former Nokia and Vodafone executive who now holds the title of vice president and head of global telecom business at the open-source solutions specialist, was chatting to Gabriel Brown, senior principal analyst covering mobile networks at Omdia, during the opening session of Light Reading’s Telecom Trends, Digital Symposium, 2026 edition.
Titled the “Network Evolution Path from 5G to 6G,” the session explored the current state of play with 6G standards and more besides, in an attempt to build a picture of how the transition to 6G is currently playing out.
Brown himself echoed Heeran’s view that 6G is coming sooner than people might think. “Why do we actually need this? … I think it’s a pretty simple story,” he said.
“Most obviously, we have a chance to create a really amazing mobile communication system. 5G is great, but there’s a lot we can do better … it’s a fantastic opportunity,” he said.
«
I’m old enough to remember when 5G was going to bring us cars that would talk to each other on the road about traffic, and broadband everywhere wirelessly. Somehow that hasn’t happened? Or has it, but people just want a bigger number all the time?
unique link to this extract
New Iran videos show bodies piled up in hospital and snipers on roofs • BBC
Merlyn Thomas and Shayan Sardarizadeh:
»
Verified videos emerging from Iran show bodies piled up in a hospital, snipers stationed on buildings and CCTV cameras being destroyed, following the unprecedented crackdown on protests earlier this month.
BBC Verify has been tracking the spread of protests across Iran since they first erupted in late December, but the near total internet blackout imposed by the authorities has made it extremely difficult to document the scale of the state’s deadly crackdown on protesters.
The US-based Human Rights Activists News Agency (HRANA) says it has confirmed the killing of nearly 6,000 people, including 5,633 protesters, since the unrest began at the end of December. It says it is also currently investigating another 17,000 reported deaths received despite an internet shutdown after nearly three weeks.
Another group, the Norway-based Iran Human Rights (IHR), has warned that the final toll could exceed 25,000.
Iranian authorities said last week that more than 3,100 people were killed, but that the majority were security personnel or bystanders attacked by “rioters”.
The latest videos to emerge from the country are understood to have been filmed on 8 and 9 January, when thousands of people took to the streets following a call for nationwide protests from Reza Pahlavi, the exiled son of the late Shah.
They are thought to be the deadliest nights for protesters so far and these newly verified videos show how Iran’s security forces have been violently cracking down on protesters.
«
Some videos do trickle out, but this is far from the observation we would like to have.
unique link to this extract
Death of an Indian tech worker • Rest Of World
Parth MN:
»
Eighty-three% of India’s tech workers suffer from burnout, according to one recent survey. One in four clocks over 70 hours a week. In Karnataka state, home to Bengaluru, tech workers account for a starkly disproportionate 20% of patients seeking transplants due to organ failure, according to a leading regional newspaper. A study of tech employees in the IT hub of Hyderabad found that 84% had a liver disease linked to long hours of sedentary work and high stress.
Some of India’s tech leaders, meanwhile, are advocating 70-hour and even 90-hour workweeks, instead of the national legal maximum of 48.
IT workers in the World Trade Centre tower in Bengaluru. Sameer Raichur for Rest of World
Tech workers paint a picture of mounting anxiety. From junior software engineers to senior project managers, workers at firms across the industry told Rest of World they were buckling under the burden of deadlines. They had little time for themselves or their families, and worried about layoffs. Most said they feared conditions would only worsen with the rise of AI.The fate of India’s tech workers may foreshadow the future of a global workforce reckoning with the advent of AI. For decades, the country’s massive pool of outsourced tech workers have helped power global tech giants — the U.S. accounts for 62% of India’s IT outsourcing revenue. As employees worry that AI will threaten their jobs and demands for efficiency rise, an industry long known for 24/7 schedules and intense workloads is reaching a breaking point.
«
| • Why do social networks drive us a little mad? • Why does angry content seem to dominate what we see? • How much of a role do algorithms play in affecting what we see and do online? • What can we do about it? • Did Facebook have any inkling of what was coming in Myanmar in 2016? Read Social Warming, my latest book, and find answers – and more. |
Errata, corrigenda and ai no corrida: none notified
The Overspill: when there's more that I want to say 