Tessera

"Data collectives and cooperatives, which let creators control the collection and distribution of their data, are emerging as preferred alternatives to big tech companies."

Can data collectives help strengthen vulnerable cultures in the face of AI?

Link: Fed up with Big Tech, communities turn to data collectives for control, by Rina Chandran at Rest of World It’s interesting to contrast the current moment to the “information wants to be free” era of Web 2.0, twenty or so years ago. Back then, everyone was talking about open APIs and open data. Now, it’s become clearer that communities need to control the terms of their data if they’re going to avoid being strip-mined for somebody else’s profit. > “Workers, producers, consumers, and others have been establishing cooperatives and other community-led associations to pool resources, share benefits, and address socioeconomic challenges for centuries. The United Nations marked 2025 as the year of cooperatives, positioning them as “essential solutions to today’s global problems,” kindling renewed interest in data collectives and cooperatives.” While there’s certainly an argument to be made that communities tend to over-estimate the value of their own data (looking at you, news), some of these datasets may be truly unique in ways that would add value to an AI service or model. As this article points out, collectively-owned data includes creative works in more than 20 African languages that aren’t recognized in mainstream linguistic frameworks. The danger, of course, is that putting these kinds of gates in front of underrepresented cultures just works to further marginalize them: in that potential future, if everyone’s using a model where those languages are missing, they become irrelevant. But there’s another one where data collectives can pull the levers they have to bring about the world they want to see. That’s exactly what the Nwulite Obodo Open Data License aims to do: data rights holders can negotiate to share their work and cultural heritage without losing their right to benefit from it. (Nwulite Obodo is Igbo for raising, reviving, and building the community.) In one model, vendors building non-extractive and responsibly trained models for public interest purposes get to use their data for free, but the closed-model big tech vendors have to pay. That’s what Meesum Alam did with voice data for 39 at-risk languages in Pakistan: the communities he worked with determined that the data was free for research and non-commercial purposes, but for-profit tech companies would need to negotiate terms (which Meta did). That potentially becomes more interesting: either OpenAI et al negotiate to license the data, or they lose functionality to their public interest competitors. There’s also a world where some communities proactively document their cultures and make them available specifically so that models, whoever they’re built by, won’t omit them. Either the world has more equitable AI or the communities financially benefit from their cultural heritage. Whatever happens, these communities certainly have the right to control their data however they see fit. What vendors do about it is the open question. But initiatives like Mozilla Data Collective make it more possible to have more substantive conversations about how data is provided and used, and that can only be a good thing.

werd.io

America's AI dominance is under threat; AI vendors put us at risk; meanwhile, everyone's staying in their jobs for the health insurance.

Notable links: July 24, 206

_Most Fridays, I share a handful of pieces that caught my eye at the intersection of technology, media, and society._ _Did someone forward this to you?_ Subscribe for free_._ * * * ### China delivers a one-two punch to America’s AI dominance AI models, as a product in themselves, have very little moat beyond what amounts to brand loyalty and superficial switching costs. Instead, the moat is in the enterprise services that sit around them: the deals and contracts, connectivity with enterprise systems, and quality of life features in an enterprise context. If we consider the models themselves, it’s easy to switch between them: someone could be using ChatGPT today and Claude tomorrow, with very little impact on their workflows. This is particularly true in the engineering world, where models are accessed via API: you can swap out the API and use the same prompt. Those companies can make deals to lock their customers in, but in practice there’s very little long-term technical incentive to use one vendor over another. You pick the best model for your needs and change models and vendors if another one becomes better. The US government has placed export controls on GPUs. There are also strong regulations that (reasonably) prevent sharing certain kinds of data with Chinese servers. The result is that while Chinese companies have enough compute to _train_ models, they can’t really provide the kinds of global-scale centralized services that we see from OpenAI and Anthropic — at least, not in the same way. And open almost always wins when it comes to infrastructure adoption. Open technologies can be used permissionlessly and therefore can be at the center of more innovation. You can host them where you want, experiment with them, alter them, and tweak to fit your use case. Open weights models are not open source, but they _are_ portable and permissionless. With all this in mind, it makes sense for China to release its AI models openly. It turns a US-created compute disadvantage into a distribution advantage; it commoditizes the layer where American companies make money; and it creates a far more effective global ecosystem than could be established through locked-in, centralized services. It’s obvious to me that there are ecosystem benefits throughout China, from manufacturing to scientific research; every sector can just plug in these models. The saving grace for American companies has been that US frontier models have outperformed open ones. That gap is now closing: > “Moonshot and Alibaba unveiled models they claim can go toe-to-toe with the best from OpenAI and Anthropic at a fraction of the cost. The rapid-fire releases suggest America’s lead at the AI frontier is increasingly tight, just as the technology is becoming central to national security, economic power, and geopolitical influence.” Even without these new capabilities, the strategy has already been working. a16z partner Martin Casado noted in the Economist that there’s an 80% chance that any given startup is using Chinese models, and Chinese models are poised to take the lead. It’s worth taking a step back and considering the surprising underlying dynamics. We think of China as being a locked-down society — and it is in many ways. I have serious concerns about how these models might reflect Chinese government perspectives (try asking them about Tiananmen Square). But it’s American companies that are keeping tight control of their technology rather than releasing it as openly as possible. This is in stark contrast to the strategy behind US government support for the open internet, for example. Locked-down business practices for a technology with no real moat but significant potential ecosystem benefits is an obviously losing strategy; permissively releasing it with an open, collaborative approach is obviously a winning one. But the incentives in the US aren’t there: instead, these companies are forced to chase first-order profits rather than ecosystem benefits, and the government tries to put its finger on the scale through forcible measures like tight export controls. We should consider what would need to change to make those incentives more aligned. That’s particularly important given how much of the US economy is currently driven by AI spending. If the bottom falls out of that spending — and I think it clearly will, given the dynamics — the outcome could be severe. I care about having open technology that can be run in the public interest, aligned with the public’s values. Threads like public AI, federated services, and open research have traction but need backing. Getting there in the US needs more nuanced strategy and support than we’re seeing today. * * * ### This Conversation Is Being Recorded. They All Are. I’ve been thinking about this story for days. > “A Zoom call isn’t complete without an artificial-intelligence note taker. Phones are out at meetings, capturing every word. During impromptu conversations with co-workers, someone might turn on the Granola transcription app, which can turn the interactions into one-page summaries or a list of action items. Even at bars and on dates, people are using AI-infused listening apps to analyze conversations later on.” The story goes on to talk to a woman who uses Granola to record her dates, then pours the transcripts into Claude to give her feedback about how she could have done better. And there’s account after account of people using it in meetings without asking for consent or revealing that they’re recording. Certainly in Silicon Valley, a societal shift seems to be underway. It’s likely much more widespread than that. I’ve been present in meetings outside the tech industry where Granola’s watermarking was visible but I wasn’t asked to consent. The watermarking is optional; I have to assume I’ve been in meetings where I’ve been recorded without my knowledge. Pair this trend with the story that the Trump Administration actively sought the phone records of journalists — and their families — who reported on the new Qatari-gifted Air Force One. Subpoenas were issued to the phone carriers, and the Department of Justice notified the newsroom a week later. In some cases, subpoenas can be issued to carriers and service providers privately, allowing the data to be retrieved without the newsroom’s knowledge; in this case, the DoJ did try to gag the phone company from alerting the newsroom. A world in which every conversation is recorded and transcribed is one where every conversation can be subpoenaed or surveilled. Here, the surveillance is decentralized through people who actively want to conduct it for their own benefit, but the data is still stored centrally and available for authorities to subpoena or someone else to mine. Granola’s security page makes clear that the data is accessible to them — and therefore to a third party that compels them to hand it over — and notes that: > “Granola trains on your anonymized data so we can keep making Granola better. You can opt out of this in your Settings.” Granola makes a point of saying that audio is not stored, but given that transcriptions _are_ , this seems moot: the words in a conversation carry its meaning. Subpoenas for your conversations go to it, not to you, and you may never know they were served. If you record someone’s conversation without letting them know, you’re putting them at risk. Don’t get me wrong: I would _love_ to have an automatic summary of meetings I’ve taken part in. I have also run meetings on non-sensitive topics where I’ve asked for consent before starting transcription. It’s the ubiquity and covert nature of the transcription that bothers me, paired with its central storage in what amounts to a honeypot for subpoenas and hackers. Recording a conversation with someone without their consent is illegal in many states and countries, so this behavior may be forced to change. California is one of them, and Granola appears to be thriving there, so there is a world where the law changes to meet the new ubiquitous surveillance norm. Until the dust settles one way or the other, anyone who wants to talk about a sensitive topic, particularly in Silicon Valley, will need to be more wary than usual. * * * ### How OpenAI’s human mistake led to the AI-powered hack on Hugging Face The biggest technology story this week was how a combination of OpenAI models hacked into third-party AI provider Hugging Face and breached its production database. The incident was initially spun as a sort of partnership between the two companies, but it seems like that’s not what went down at all. > “OpenAI failed to properly configure what it called a ‘highly isolated environment,’ allowing a testing sandbox that should have been completely secluded from the internet to actually connect to the internet.” That’s actually one of at least two lapses here: not only did OpenAI fail to properly isolate its models, but Hugging Face’s production database was in a state where those models could hack into it. The whole thing does not speak well of security practices at AI vendors overall. We’re being asked to share more and more private information with model vendors. The standard protection they offer — at least, to their paying customers — is that your data will not be used for model training purposes. That’s all well and good, but your data is still hitting their servers, potentially being logged for an extended period in such a way that it could, in theory, be accessed by their employees. Even before we bring in the possibility of hackers, that leaves your private data open to being accessed via subpoena, an unscrupulous employee, or, indeed, an unscrupulous _vendor_. (Consider that Uber breached at least one journalist’s privacy and considered hiring an opposition research firm. Do we really think AI vendors are more ethical? Why?) Leaving a production database in a state where it could be breached is the icing on the cake. In this case, the models weren’t even harnessed to hack Hugging Face — they did so autonomously to _cheat a test_. Imagine what they might do if they were _intentionally_ pointed that way. Hacking is becoming cheaper and easier: once the preserve of talented technologists, this story proves that the latest frontier models can effectively find and exploit vulnerabilities in systems. And apparently, AI vendors can’t be trusted to secure their own infrastructure. The combination of us being encouraged to share more and more data, the inherent risks of centralizing that data, the dubious security of the places we’re being asked to share it, and the obvious shadiness of some of the companies involved should give us all pause. For newsrooms and anyone dealing with sensitive or private data, particularly relating to source materials or journalism in progress, this weak security environment is not enough. We need zero data retention contracts at minimum, but the only real way to be sure nobody can access our information is to use confidential computing environments and, ultimately, local models. Anything less leaves our work open to cowboys, hackers, and, apparently, misconfigured robots. * * * ### Staying in a job for the health insurance? About 1 in 4 Americans do, a survey says This is a striking, but not necessarily surprising, figure from a new survey by the West Health-Gallup Center on Healthcare in America: > “A new report finds that nearly a quarter of workers who get health insurance through their jobs report staying in unwanted jobs for health insurance — a figure that's risen dramatically in the last five years.” The figure rises to 41% of people with three or more chronic health conditions. The figure has risen wildly in part because Affordable Care Act subsidies were allowed to expire. I’ll get the soapbox out of the way first: having spent around thirty years of my life in the UK before moving to the US, the thing I miss most is the NHS. It’s been treated like a political football since I left and is apparently a shell of its former self — not because the idea is bad and can’t work but because conservative politicians, some of whom have received funding from private healthcare companies, have deliberately sabotaged it. But it’s hard to explain the _lack of fear_ of walking into a doctor’s office or a hospital. You know for a fact that there won’t be an onerous bill. You can just get seen. That security allowed me to found my first startup, which in turn has set the stage for my entire career. If you’re in the US, you may have heard some less pleasant things about socialized healthcare: it turns out much of it was a deliberate disinformation campaign by private health insurers, which I think says a lot about how the whole American healthcare system actually works. That soapbox out of the way, I also want to highlight how the private healthcare system creates perverse incentives for employers and dampens innovation. If employees have freedom of movement between companies, the incentive for employers is to create the best working conditions possible: higher wages, great benefits, a nurturing working environment. If, on the other hand, some employees are effectively chained to their desks by their need to have healthcare, employers have less of a need to provide those things. As long as they provide a reasonable health plan, wages and working conditions are secondary. As the West Health-Gallup Center themselves assert, the effect is lower wages and worse work. In turn, fewer innovators are empowered, which is a disaster for industries like news that desperately need innovation. Often, innovators will find themselves constrained by their existing employers for various reasons and want to leave to explore a new idea that has the potential to change their industry. (That was my experience leaving the university sector to build a social platform for learning, which was ultimately used by Ivy Leagues, non-profits, and governments around the world.) If they _can’t_ because they’re tethered to employers who won’t greenlight their ideas, those innovations will never see the light of day. So not only does socialized healthcare allow people to be healthier by removing the fear of going to the doctor in the first place, it improves wages, creates more competitive working conditions, and promotes innovation. Even a representative for the Cato Institute — a libertarian think tank — has this to say in the piece: > “Favoring employer-sponsored health insurance creates coverage gaps, reduces income mobility, and is crying out for reform.” When even the libertarians want reform, you know it’s a bad deal. We need a different healthcare system. While the libertarians would likely disagree, my vote — having experienced and enjoyed it for much of my life — is for universal healthcare. The only real downside to it is that a bunch of companies that have entrenched their positions taking advantage of ordinary people will be denied a little profit. Which, you know. Pardon me while I find my tiny violin. * * * ### And more: Here are some of the stories I didn't get a chance to go into in depth this week. #### Protecting our FLOSS commons from LLMs The source code repository hosting service Codeberg has banned LLM-generated code. Time will tell whether that's a move that solidifies a niche as a place for hand-crafted software, or whether it just turns people back to GitHub. #### Google search traffic to leading UK publishers set to halve by Q3 2027 The march towards Google Zero continues apace. I believe the most effective way to build resilience against this trend is by building stronger relationships – not just one-way audience strategies, but real community. #### The Fourth Circuit Says Border Agents Can Search Your Phone By Hand, No Suspicion Required A court upheld that border agents have the right to search your phone. Newsrooms should build strong, repeatable guidance for journalists who might want to cross borders with source information. #### Kaiser Permanente nurses say technology is making their jobs — and patient care — worse Despite what vendors and management say, the people who are actually on the ground providing healthcare report that AI is having a detrimental effect on the care they can provide. That will eventually come to a head – particularly if it starts to reveal itself in patient outcome statistics.

werd.io

There were two big lapses in the Hugging Face / OpenAI hacking story: the lack of security protections on OpenAI's end, and Hugging Face's vulnerable production database. That should worry anyone who uses AI with sensitive data.

AI vendors can't be trusted to secure their systems. Newsrooms need to act accordingly

Link: How OpenAI’s human mistake led to the AI-powered hack on Hugging Face, by Lorenzo Franceschi-Bicchierai in TechCrunch The biggest technology story this week was how a combination of OpenAI models hacked into third-party AI provider Hugging Face and breached its production database. The incident was initially spun as a sort of partnership between the two companies, but it seems like that’s not what went down at all. > “OpenAI failed to properly configure what it called a ‘highly isolated environment,’ allowing a testing sandbox that should have been completely secluded from the internet to actually connect to the internet.” That’s actually one of at least two lapses here: not only did OpenAI fail to properly isolate its models, but Hugging Face’s production database was in a state where those models could hack into it. The whole thing does not speak well of security practices at AI vendors overall. We’re being asked to share more and more private information with model vendors. The standard protection they offer — at least, to their paying customers — is that your data will not be used for model training purposes. That’s all well and good, but your data is still hitting their servers, potentially being logged for an extended period in such a way that it could, in theory, be accessed by their employees. Even before we bring in the possibility of hackers, that leaves your private data open to being accessed via subpoena, an unscrupulous employee, or, indeed, an unscrupulous _vendor_. (Consider that Uber breached at least one journalist’s privacy and considered hiring an opposition research firm. Do we really think AI vendors are more ethical? Why?) Leaving a production database in a state where it could be breached is the icing on the cake. In this case, the models weren’t even harnessed to hack Hugging Face — they did so autonomously to _cheat a test_. Imagine what they might do if they were _intentionally_ pointed that way. Hacking is becoming cheaper and easier: once the preserve of talented technologists, this story proves that the latest frontier models can effectively find and exploit vulnerabilities in systems. And apparently, AI vendors can’t be trusted to secure their own infrastructure. The combination of us being encouraged to share more and more data, the inherent risks of centralizing that data, the dubious security of the places we’re being asked to share it, and the obvious shadiness of some of the companies involved should give us all pause. For newsrooms and anyone dealing with sensitive or private data, particularly relating to source materials or journalism in progress, this weak security environment is not enough. We need zero data retention contracts at minimum, but the only real way to be sure nobody can access our information is to use confidential computing environments and, ultimately, local models. Anything less leaves our work open to cowboys, hackers, and, apparently, misconfigured robots.

werd.io

1 in 4 American workers stay in their jobs because of the healthcare. It suppresses wages and working conditions - and prevents innovators from heading out on their own.

Private healthcare makes industries less innovative. It's time for change.

Link: Staying in a job for the health insurance? About 1 in 4 Americans do, a survey says, by Joseph Kim at NPR This is a striking, but not necessarily surprising, figure from a new survey by the West Health-Gallup Center on Healthcare in America: > “A new report finds that nearly a quarter of workers who get health insurance through their jobs report staying in unwanted jobs for health insurance — a figure that's risen dramatically in the last five years.” The figure rises to 41% of people with three or more chronic health conditions. The figure has risen wildly in part because Affordable Care Act subsidies were allowed to expire. I’ll get the soapbox out of the way first: having spent around thirty years of my life in the UK before moving to the US, the thing I miss most is the NHS. It’s been treated like a political football since I left and is apparently a shell of its former self — not because the idea is bad and can’t work but because conservative politicians, some of whom have received funding from private healthcare companies, have deliberately sabotaged it. But it’s hard to explain the _lack of fear_ of walking into a doctor’s office or a hospital. You know for a fact that there won’t be an onerous bill. You can just get seen. That security allowed me to found my first startup, which in turn has set the stage for my entire career. If you’re in the US, you may have heard some less pleasant things about socialized healthcare: it turns out much of it was a deliberate disinformation campaign by private health insurers, which I think says a lot about how the whole American healthcare system actually works. That soapbox out of the way, I also want to highlight how the private healthcare system creates perverse incentives for employers and dampens innovation. If employees have freedom of movement between companies, the incentive for employers is to create the best working conditions possible: higher wages, great benefits, a nurturing working environment. If, on the other hand, some employees are effectively chained to their desks by their need to have healthcare, employers have less of a need to provide those things. As long as they provide a reasonable health plan, wages and working conditions are secondary. As the West Health-Gallup Center themselves assert, the effect is lower wages and worse work. In turn, fewer innovators are empowered, which is a disaster for industries like news that desperately need innovation. Often, innovators will find themselves constrained by their existing employers for various reasons and want to leave to explore a new idea that has the potential to change their industry. (That was my experience leaving the university sector to build a social platform for learning, which was ultimately used by Ivy Leagues, non-profits, and governments around the world.) If they _can’t_ because they’re tethered to employers who won’t greenlight their ideas, those innovations will never see the light of day. So not only does socialized healthcare allow people to be healthier by removing the fear of going to the doctor in the first place, it improves wages, creates more competitive working conditions, and promotes innovation. Even a representative for the Cato Institute — a libertarian think tank — has this to say in the piece: > “Favoring employer-sponsored health insurance creates coverage gaps, reduces income mobility, and is crying out for reform.” When even the libertarians want reform, you know it’s a bad deal. We need a different healthcare system. While the libertarians would likely disagree, my vote — having experienced and enjoyed it for much of my life — is for universal healthcare. The only downside to universal healthcare is that a bunch of companies that have entrenched their positions taking advantage of ordinary people will be denied a little profit. Pardon me while I find my tiny violin.

werd.io

Apps like Granola make it easy to transcribe conversations without asking for consent. Those transcripts are a subpoena honeypot.

People are transcribing your conversations without asking. That puts you at risk.

Link: This Conversation Is Being Recorded. They All Are., by Katherine Bindley at the Wall Street Journal I’ve been thinking about this story for days. > “A Zoom call isn’t complete without an artificial-intelligence note taker. Phones are out at meetings, capturing every word. During impromptu conversations with co-workers, someone might turn on the Granola transcription app, which can turn the interactions into one-page summaries or a list of action items. Even at bars and on dates, people are using AI-infused listening apps to analyze conversations later on.” The story goes on to talk to a woman who uses Granola to record her dates, then pours the transcripts into Claude to give her feedback about how she could have done better. And there’s account after account of people using it in meetings without asking for consent or revealing that they’re recording. Certainly in Silicon Valley, a societal shift seems to be underway. It’s likely much more widespread than that. I’ve been present in meetings outside the tech industry where Granola’s watermarking was visible but I wasn’t asked to consent. The watermarking is optional; I have to assume I’ve been in meetings where I’ve been recorded without my knowledge. Pair this trend with the story that the Trump Administration actively sought the phone records of journalists — and their families — who reported on the new Qatari-gifted Air Force One. Subpoenas were issued to the phone carriers, and the Department of Justice notified the newsroom a week later. In some cases, subpoenas can be issued to carriers and service providers privately, allowing the data to be retrieved without the newsroom’s knowledge; in this case, the DoJ did try to gag the phone company from alerting the newsroom. A world in which every conversation is recorded and transcribed is one where every conversation can be subpoenaed or surveilled. Here, the surveillance is decentralized through people who actively want to conduct it for their own benefit, but the data is still stored centrally and available for authorities to subpoena or someone else to mine. Granola’s security page makes clear that the data is accessible to them — and therefore to a third party that compels them to hand it over — and notes that: > “Granola trains on your anonymized data so we can keep making Granola better. You can opt out of this in your Settings.” Granola makes a point of saying that audio is not stored, but given that transcriptions _are_ , this seems moot: the words in a conversation carry its meaning. Subpoenas for your conversations go to it, not to you, and you may never know they were served. If you record someone’s conversation without letting them know, you’re putting them at risk. Don’t get me wrong: I would _love_ to have an automatic summary of meetings I’ve taken part in. I have also run meetings on non-sensitive topics where I’ve asked for consent before starting transcription. It’s the ubiquity and covert nature of the transcription that bothers me, paired with its central storage in what amounts to a honeypot for subpoenas and hackers. Recording a conversation with someone without their consent is illegal in many states and countries, so this behavior may be forced to change. California is one of them, and Granola appears to be thriving there, so there is a world where the law changes to meet the new ubiquitous surveillance norm. Until the dust settles one way or the other, anyone who wants to talk about a sensitive topic, particularly in Silicon Valley, will need to be more wary than usual.

werd.io

China's open-weights AI strategy is winning: its companies are taking the lead. America's closed-first, locked-down strategy is doomed to failure - and it could take the US economy down with it.

American AI is locked down and proprietary. It's losing.

Link: China delivers a one-two punch to America’s AI dominance, by Robert Hart in The Verge AI models, as a product in themselves, have very little moat beyond what amounts to brand loyalty and superficial switching costs. Instead, the moat is in the enterprise services that sit around them: the deals and contracts, connectivity with enterprise systems, and quality of life features in an enterprise context. If we consider the models themselves, it’s easy to switch between them: someone could be using ChatGPT today and Claude tomorrow, with very little impact on their workflows. This is particularly true in the engineering world, where models are accessed via API: you can swap out the API and use the same prompt. Those companies can make deals to lock their customers in, but in practice there’s very little long-term technical incentive to use one vendor over another. You pick the best model for your needs and change models and vendors if another one becomes better. The US government has placed export controls on GPUs. There are also strong regulations that (reasonably) prevent sharing certain kinds of data with Chinese servers. The result is that while Chinese companies have enough compute to _train_ models, they can’t really provide the kinds of global-scale centralized services that we see from OpenAI and Anthropic — at least, not in the same way. And open almost always wins when it comes to infrastructure adoption. Open technologies can be used permissionlessly and therefore can be at the center of more innovation. You can host them where you want, experiment with them, alter them, and tweak to fit your use case. Open weights models are not open source, but they _are_ portable and permissionless. With all this in mind, it makes sense for China to release its AI models openly. It turns a US-created compute disadvantage into a distribution advantage; it commoditizes the layer where American companies make money; and it creates a far more effective global ecosystem than could be established through locked-in, centralized services. It’s obvious to me that there are ecosystem benefits throughout China, from manufacturing to scientific research; every sector can just plug in these models. The saving grace for American companies has been that US frontier models have outperformed open ones. That gap is now closing: > “Moonshot and Alibaba unveiled models they claim can go toe-to-toe with the best from OpenAI and Anthropic at a fraction of the cost. The rapid-fire releases suggest America’s lead at the AI frontier is increasingly tight, just as the technology is becoming central to national security, economic power, and geopolitical influence.” Even without these new capabilities, the strategy has already been working. a16z partner Martin Casado noted in the Economist that there’s an 80% chance that any given startup is using Chinese models, and Chinese models are poised to take the lead. It’s worth taking a step back and considering the surprising underlying dynamics. We think of China as being a locked-down society — and it is in many ways. I have serious concerns about how these models might reflect Chinese government perspectives (try asking them about Tiananmen Square). But it’s American companies that are keeping tight control of their technology rather than releasing it as openly as possible. This is in stark contrast to the strategy behind US government support for the open internet, for example. Locked-down business practices for a technology with no real moat but significant potential ecosystem benefits is an obviously losing strategy; permissively releasing it with an open, collaborative approach is obviously a winning one. But the incentives in the US aren’t there: instead, these companies are forced to chase first-order profits rather than ecosystem benefits, and the government tries to put its finger on the scale through forcible measures like tight export controls. We should consider what would need to change to make those incentives more aligned. That’s particularly important given how much of the US economy is currently driven by AI spending. If the bottom falls out of that spending — and I think it clearly will, given the dynamics — the outcome could be severe. I care about having open technology that can be run in the public interest, aligned with the public’s values. Threads like public AI, federated services, and open research have traction but need backing. Getting there in the US needs more nuanced strategy and support than we’re seeing today.

werd.io

At a time when journalism is increasingly under attack, we need PIT Crews for news.

Notable links: July 17, 2026

* * * _Most Fridays, I share a handful of pieces that caught my eye at the intersection of technology, media, and society._ _Did someone forward this to you?_ Subscribe for free_._ * * * ### Mamdani invests in tech capacity to “solve real problems” There’s a lot that newsrooms can learn from Zohran Mamdani’s mayoral administration in New York City. His latest announcement is the Public Interest Technology (PIT) Crew, a set of dynamic, cross-disciplinary digital teams that will solve problems across the city using a rapid, human-centered approach. As Pamela Herd notes here, this is a shift from contracting out to building internal capacity: > “Traditionally, the conventional wisdom since the 1990s and before was that governments could buy tech products like an off-the-shelf product. This led to a massive turn to contracting out, which was great for consultants but bad for government capacity. The outsourced approach often cost too much, delivering too little and too late. > > […] What people who know tech and government have been screaming for years is that building good tech needs in-house capacity, even when you are using contractors. It requires the government owning the design, development and delivery of technology, relying on rapid iteration to fix problems in a way that is impossible when contractors are running things.” This dynamic is also highly prevalent in newsrooms, resulting in the same problems. If you rely too heavily on buying existing technology or working with outside contractors, you are building operational, functional, and intellectual dependencies on those organizations. You import their values and ways of working, which in the case of some vendors may be catastrophic in itself, but you also put yourself on their timelines and make yourself subject to their feature priorities and interests. And that’s before you consider security and trust profiles, which may radically differ between newsrooms and the vendors that serve them. New York City isn’t alone; other governments are beginning to shift from outsourcing back to internally owned technology. The article links to a report explaining Colorado’s move back to internally-run IT, which states the issue plainly: > “There is an alignment problem: the issue is not effort, but that we have organized around internal structures rather than outcomes, and that misalignment has made excellent work harder.” Mamdani’s PIT Crew sounds a lot like how a product team should work: directed groups of experts rapidly prototyping solutions to concretely defined problems anchored in real people’s needs. By doing it internally, he can make sure these solutions are built exactly the way the city needs, build institutional capacity and knowledge, and, theoretically at least, do it far more cheaply in the long run. As these sorts of civic measures succeed, I think (or, perhaps, I _hope_) we’ll see more newsrooms translate those outcomes to their own businesses and begin to understand that they need to prioritize technical capacity too. All the same reasons apply here. Of course, most newsrooms don’t have the budget of the New York City Mayor’s office. I think the solution to that is third entities: non-profit organizations that exist to provide shared technical capacity across newsrooms, based on newsroom needs, _that behave as if they were part of newsroom teams_. Think of it as a kind of PIT Crew for news, operated independently but in deep collaboration with newsrooms. By using a radically open source approach, newsrooms can pool resources together and solve shared technical problems more easily, on their terms and according to their values. While there are always places for startups and tech platforms, the idea that the tech industry can always serve needs better than building institutional capacity is fundamentally broken; it’s also fundamentally right-wing. I’m delighted to see the New York City Mayor’s office move in a more productive direction. I hope it becomes an example for everyone. * * * ### We are not alone I was delighted to be included in this roundup by Adiel Kaplan, the Program Director at the Tow-Knight Center for Journalism Futures at the Craig Newmark Graduate School of Journalism at CUNY. As Adiel says: > “Having a say in what the future of news looks like will likely require not just that collaboration across newsrooms, but also outside them, with other institutions that want to shape a future with informed communities at its center — which is, after all, the mission. Right? > > […] It will also require a different way of thinking about our role in this ecosystem, beyond creating content and distributing it. It might mean getting more involved in building technology, or joining forces in new ways with government-funded institutions.” This is exciting to me: I’ve been saying for a while now that news needs to get more involved in building technology. My flippant line is that _news treats technology as something that happens to it, like an asteroid_ — but it’s actually a creative work, like an article. Although many newsrooms are too small to build a strong capacity in themselves, it’s perfectly possible for news _as an industry_ to build capacity and create the technology that is unique to its use cases on its terms. So I think it’s a very good thing that news institutions are talking about this need. The people listed in the article are exceptional. I’m just happy to be on the list in such fine company. Don’t sleep on any of them; I feel most connected to Ivan Sigal’s ambitious and vital work at the Modal Foundation and what Trei Brundrett is building (in collaboration with Blaine Cook and others) at New_ Public. But these are all worthy endeavors: the Library Newsroom Project is a genius on-the-ground effort to create local newsrooms based in every public library in the US, and Sannuta Raghu’s news atoms embed meaning and provenance in natural language articles. All are promising. We need to move forward. There are certainly more people who could have been added to such a list; my hope is that if one were written a year from now, it would be exponentially longer. Let’s innovate. * * * ### White House Directed Patel to Oversee Investigation Involving Times Reporting The White House personally directed FBI Director Kash Patel to issue subpoenas to journalists reporting on the President’s new Qatari-gifted Air Force One. > “The White House’s deep involvement in the case came after officials said that President Trump was enraged about the coverage of the Qatari-donated plane, which The Times reported Thursday lacks the same defensive countermeasures of the previous Air Force One.” These subpoenas were delivered by hand to some of the reporters at home, echoing the FBI’s raid of a Washington Post engagement reporter’s home earlier this year. In both cases, it’s highly likely that these were attempts to discover who leaked information to their respective newsrooms. There’s lots to say about first amendment issues here, and commentators like Dan Kennedy at Media Nation have pertinent thoughts. It’s clear that journalism is under attack by the administration, and they rescinded rules that protected journalists in leak investigations last year. The US Press Freedom Tracker is a sobering read. But it’s also important to take a moment to talk about the technology side of this story. When the administration wants to issue a subpoena to a newsroom, it has a few avenues available to it. The first is to issue it directly to the newsroom or to its reporters, as they did here. In some ways, this is the best outcome: then the newsroom knows about the subpoena and can actively fight it in court. The other avenue is to subpoena the newsroom’s service providers. If source information is stored unencrypted on a service like Google Workspace, the administration could subpoena Google. If a gag order is added — which might well happen if it’s a criminal subpoena or labeled a matter of national security — then the newsroom would never find out and have the chance to fight it. This is true even if the service provider nominally promises to notify the newsrooms about subpoenas: a gag order is a gag order. Larger newsrooms have strong data security practices for this reason: they know to create policies and architectures which force subpoenas to come through them. But not every newsroom has the capacity to build a strong security strategy. Which means for every story we hear about that involves these newsrooms, there may be many more that took place in secret. The Freedom of the Press Foundation maintains digital security resources and runs training for newsrooms and specific advice about source protection. The EFF also has some great resources. More resources are out there. But there is more of a need than ever for every newsroom to make sure they have access to someone who can advise them on digital security both holistically and on a case-by-case basis. Not every newsroom can afford a permanent member of staff, but finding access to some kind of resource is vital. Likewise, journalism funders should focus on providing access to experts, understanding that these issues are existential for the organizations they fund. Not only is this an attack on press freedoms, but it’s also an attack on trust. Every newsroom can do its reporting because sources feel safe to reach out to it; if their safety is in question, they may be less likely to leak, and we may be less likely to read the stories that help us make good democratic decisions. That’s what the administration seems to be banking on. * * * ### Trump dismantled a federal climate website. These women rebuilt it. This shouldn’t have been necessary, but is still wonderful to see. Climate.gov had been the go-to resource for climate data, but it went offline when the Trump Administration radically cut NOAA’s funding. At that point: > “[Rebecca] Lindsey joined forces with former NOAA employees Anna Eshelman, and Mary Lindsey, her older sister, to become the core team behind the deactivated site’s successor, Climate.us, preserving over 15 years of key climate data and resources. The trove features key maps, educational materials and climate indicator reports, including the now-deleted Fifth National Climate Assessment, the government’s most comprehensive analysis of climate change that was at risk of being lost to the public.” This is possible because US government data is public domain by law. Had it not been available under a permissive license, the administration’s act of vandalism would have meant the data was gone for good. But because it was, the datasets can find a new home. It’s a joy to use. Check out the climate dashboard, which tracks numbers like the total area of the Arctic Ocean that was at least 15% ice-covered each September. It also hosts a set of resources for teaching climate and energy. The dataset gallery includes crucial information like the NOAA’s archive of oral histories from people whose lives were affected by climate change. But it’s also precarious. The whole thing relies on donations to keep it afloat, which is really what tax dollars are for. Still, for the moment it’s wonderful to see people pick up the slack when government is no longer doing its job. In the absence of government support, archives like this are works of journalism in themselves: ways to help us make stronger decisions. They deserve stronger support, and ultimately, we all deserve the restoration of such important government infrastructure. * * * ### A Leak of San Francisco Police Drone Footage Exposes the New Reality of Urban Surveillance I’m not sure I agree with this article’s implication that the problem with SFPD’s drone policing was that it accidentally leaked the data. > ““There’s a certain trust given to the police to use these things correctly,” says Curry. “When you're watching a drone feed live, you can look into dozens of different apartments, you can see police zooming in on people, you can see arrests. The fact that all of this was exposed feels like a really big issue from a privacy perspective.”” I’d humbly submit that the privacy problem exists regardless of whether the footage was leaked or not: this is ubiquitous surveillance of a city’s citizens from above. That footage can be analyzed, both by humans and software, to track people and target them for any reason. There is very little oversight, and because the police department is using a private company to run it, the teams there presumably have access to an enormous amount of private footage. The thing is, none of this actually makes us safer. As the ACLU of Northern California points out in its Seeing Through Surveillance report: > “The evidence is clear that while surveillance has increased exponentially, public safety has not. On the contrary, surveillance systems often make people less safe, especially for groups that have historically been in the government’s crosshairs. Modern surveillance technology makes it possible for the government to track who we are, where we go, what we do, and who we know. It fuels high-tech profiling and perpetuates systems of biased policing. It facilitates deportations, chills speech, and imperils the rights of activists, religious minorities, and people who need reproductive and gender-affirming care.” Most importantly, it doesn’t actually help. As the report points out, the city of San Francisco itself learned that adding cameras to its highest-crime neighborhoods _had no impact on crime_. Regardless, it added more funding to the program and voted to remove oversight in 2023. The result is more money spent, less privacy, with no impact on public safety. And now we know that the footage is being accidentally leaked, the privacy footprint is obviously even worse. In a world that is becoming markedly more authoritarian, it’s unconscionable that supposedly permissive cities would add more surveillance. It doesn’t work, it misuses funds that could be spent helping the vulnerable, and it’s data that could be used for undemocratic purposes. It needs to stop — and to do that, we need to apply pressure to our elected representatives and raise awareness of how backwards it is.

werd.io

"Allies, archives and infrastructure in the AI age" - a list of people with the potential to push news forward.

To innovate, news needs allies

Link: We are not alone, by Adiel Kaplan at the Tow-Knight Center I was delighted to be included in this roundup by Adiel Kaplan, the Program Director at the Tow-Knight Center for Journalism Futures at the Craig Newmark Graduate School of Journalism at CUNY. As Adiel says: > “Having a say in what the future of news looks like will likely require not just that collaboration across newsrooms, but also outside them, with other institutions that want to shape a future with informed communities at its center — which is, after all, the mission. Right? > > […] It will also require a different way of thinking about our role in this ecosystem, beyond creating content and distributing it. It might mean getting more involved in building technology, or joining forces in new ways with government-funded institutions.” This is exciting to me: I’ve been saying for a while now that news needs to get more involved in building technology. My flippant line is that _news treats technology as something that happens to it, like an asteroid_ — but it’s actually a creative work, like an article. Although many newsrooms are too small to build a strong capacity in themselves, it’s perfectly possible for news _as an industry_ to build capacity and create the technology that is unique to its use cases on its terms. So I think it’s a very good thing that news institutions are talking about this need. The people listed in the article are exceptional. I’m just happy to be on the list in such fine company. Don’t sleep on any of them; I feel most connected to Ivan Sigal’s ambitious and vital work at the Modal Foundation and what Trei Brundrett is building (in collaboration with Blaine Cook and others) at New_ Public. But these are all worthy endeavors: the Library Newsroom Project is a genius on-the-ground effort to create local newsrooms based in every public library in the US, and Sannuta Raghu’s news atoms embed meaning and provenance in natural language articles. All are promising. We need to move forward. There are certainly more people who could have been added to such a list; my hope is that if one were written a year from now, it would be exponentially longer. Let’s innovate.

werd.io

Zohran Mamdani has unveiled a radically collaborative, cross-disciplinary approach to building internal technology capacity. News has a lot to learn from it.

We need a PIT Crew for news

Link: Mamdani invests in tech capacity to “solve real problems”, by Pamela Herd in Can We Still Govern? There’s a lot that newsrooms can learn from Zohran Mamdani’s mayoral administration in New York City. His latest announcement is the Public Interest Technology (PIT) Crew, a set of dynamic, cross-disciplinary digital teams that will solve problems across the city using a rapid, human-centered approach. As Pamela Herd notes here, this is a shift from contracting out to building internal capacity: > “Traditionally, the conventional wisdom since the 1990s and before was that governments could buy tech products like an off-the-shelf product. This led to a massive turn to contracting out, which was great for consultants but bad for government capacity. The outsourced approach often cost too much, delivering too little and too late. > > […] What people who know tech and government have been screaming for years is that building good tech needs in-house capacity, even when you are using contractors. It requires the government owning the design, development and delivery of technology, relying on rapid iteration to fix problems in a way that is impossible when contractors are running things.” This dynamic is also highly prevalent in newsrooms, resulting in the same problems. If you rely too heavily on buying existing technology or working with outside contractors, you are building operational, functional, and intellectual dependencies on those organizations. You import their values and ways of working, which in the case of some vendors may be catastrophic in itself, but you also put yourself on their timelines and make yourself subject to their feature priorities and interests. And that’s before you consider security and trust profiles, which may radically differ between newsrooms and the vendors that serve them. New York City isn’t alone; other governments are beginning to shift from outsourcing back to internally owned technology. The article links to a report explaining Colorado’s move back to internally-run IT, which states the issue plainly: > “There is an alignment problem: the issue is not effort, but that we have organized around internal structures rather than outcomes, and that misalignment has made excellent work harder.” Mamdani’s PIT Crew sounds a lot like how a product team should work: directed groups of experts rapidly prototyping solutions to concretely defined problems anchored in real people’s needs. By doing it internally, he can make sure these solutions are built exactly the way the city needs, build institutional capacity and knowledge, and, theoretically at least, do it far more cheaply in the long run. As these sorts of civic measures succeed, I think (or, perhaps, I _hope_) we’ll see more newsrooms translate those outcomes to their own businesses and begin to understand that they need to prioritize technical capacity too. All the same reasons apply here. Of course, most newsrooms don’t have the budget of the New York City Mayor’s office. I think the solution to that is third entities: non-profit organizations that exist to provide shared technical capacity across newsrooms, based on newsroom needs, _that behave as if they were part of newsroom teams_. Think of it as a kind of PIT Crew for news, operated independently but in deep collaboration with newsrooms. By using a radically open source approach, newsrooms can pool resources together and solve shared technical problems more easily, on their terms and according to their values. While there are always places for startups and tech platforms, the idea that the tech industry can always serve needs better than building institutional capacity is fundamentally broken; it’s also fundamentally right-wing. I’m delighted to see the New York City Mayor’s office move in a more productive direction. I hope it becomes an example for everyone.

werd.io

"After losing their jobs at NOAA, Rebecca Lindsey, her sister and another colleague teamed up to rebuild a pivotal resource the Trump administration took offline."

Climate.gov was destroyed. Open data saved it.

Link: Trump dismantled a federal climate website. These women rebuilt it., by Jenae Barnes at The 19th This shouldn’t have been necessary, but is still wonderful to see. Climate.gov had been the go-to resource for climate data, but it went offline when the Trump Administration radically cut NOAA’s funding. At that point: > “[Rebecca] Lindsey joined forces with former NOAA employees Anna Eshelman, and Mary Lindsey, her older sister, to become the core team behind the deactivated site’s successor, Climate.us, preserving over 15 years of key climate data and resources. The trove features key maps, educational materials and climate indicator reports, including the now-deleted Fifth National Climate Assessment, the government’s most comprehensive analysis of climate change that was at risk of being lost to the public.” This is possible because US government data is public domain by law. Had it not been available under a permissive license, the administration’s act of vandalism would have meant the data was gone for good. But because it was, the datasets can find a new home. It’s a joy to use. Check out the climate dashboard, which tracks numbers like the total area of the Arctic Ocean that was at least 15% ice-covered each September. It also hosts a set of resources for teaching climate and energy. The dataset gallery includes crucial information like the NOAA’s archive of oral histories from people whose lives were affected by climate change. But it’s also precarious. The whole thing relies on donations to keep it afloat, which is really what tax dollars are for. Still, for the moment it’s wonderful to see people pick up the slack when government is no longer doing its job. In the absence of government support, archives like this are works of journalism in themselves: ways to help us make stronger decisions. They deserve stronger support, and ultimately, we all deserve the restoration of such important government infrastructure.

werd.io

In the face of declining trust in media, communities are part of the solution.

Notable links: July 10, 2026

_Most Fridays, I share a handful of pieces that caught my eye at the intersection of technology, media, and society._ _Did someone forward this to you?_ Subscribe for free_._ * * * ### Chicago Public Media launching community website — chicago.com — in the fall In an increasingly AI-dominated information landscape, real trusted communities and relationships will be the way to build trust and loyalty — I’m convinced of it, and organizations like New_ Public agree. So it was exciting to see Chicago Public Media take a huge step towards building a community platform. > The site will include Chicago-area information, civic and cultural resources, community-sourced knowledge and opportunities for audience participation, the nonprofit said Wednesday. It will also curate headlines from the Sun-Times, WBEZ and other news sources. This will be a familiar argument to regular readers: > For independent journalism to “truly service the public … we should have digital infrastructure that is also steered by public media companies,” Chicago Public Media CEO Melissa Bell said. The news industry “has ceded a lot of distribution to places like Facebook and X, formerly known as Twitter, and I think that has done a disservice to centering civic discourse in a healthy way.” We’ve seen other platforms release similar efforts effectively. The Newsmast Foundation builds community-first social media apps on open protocols for newsrooms that include The Bristol Cable and Find Out Media. Flipboard’s Surf platform powers curated social feeds, again built on open social web protocols, for the likes of 404 Media and Rolling Stone (as well as my own curated non-profit US news feed). And Canada’s Village Media serves 13 local social networks through its SPACES platform. But this is the first time we’ve seen a single social platform rolled out by a public media company at this scale. Chicago Public Media was gifted the underlying chicago.com domain and will be rolling it out to neighborhoods and suburbs throughout the area. It sounds like each community will be highlighted (perhaps with its own feed), with an attached hub that covers the entire region. Clearly, this is an experiment, but I’m delighted to see a public media innovator explore these ideas at this scale. I see it as vindication for the idea that building stronger community applications into the public media model is a path towards a more trusted future for local journalism. I’ll be watching very closely, and I’m curious to see who dives in next. * * * ### Your SaaS Metrics Are A Result, Not A Strategy I still subscribe to sites like Crunchbase News from my time in startup-land; although it’s been a while since I’ve run the financial side of a business, I’m interested, and I know that I’ll run one again. I see stories like this and wonder: what would it look like for a newsroom to think this way? In startups, these metrics are known top to bottom, but I’ve rarely heard business teams talk about LTV (customer Life Time Value), CAC (Customer Acquisition Cost), or even ARR (Annual Recurring Revenue). This may be happening in finance and fundraising teams, but the culture of talking about customers / donors in teams more widely often simply isn’t there: metrics aren't communicated, dashboards aren't made available, the concepts of the metrics themselves are not explained. Not everyone should be thinking about this all the time – the firewall between business and editorial is important to maintain – but in order to make sharp prioritization and experimentation decisions, the business side should be much more customer / donor focused than they often are. Beyond that, this piece points out, rightly, that metrics are not strategy: they’re the measurable outcome of your strategy. They’re important tools to help you figure out cause and effect and improve your revenue efficiency, but they are not the underlying mechanism. Interesting provocation here from the author: > “The Rule of 4 adds a simple durability check: ARR growth divided by annual customer churn should be above four. If it is low, growth may be hiding a leaking bucket. > > [The board should ask:] are we growing on top of a loyal customer base, or replacing customers we should have kept?” Growth in annual recurring revenue — the portion of your revenue that is from recurring customers like subscribers or monthly / annual donors — is expressed as a percentage. So is churn: what percentage of customers (paid subscribers, members, recurring donors) cancel their commitments and don’t return? How many newsrooms have those numbers handy? What would it take to measure them? Which systems are missing that would let you do that? There is _so much_ that newsrooms — including non-profit publications — can learn from for-profit startups and other businesses. There’s a lot to be gained by sharing knowledge from those other domains. Figuring out which metrics successful businesses track and mapping the data gaps inside a newsroom is a good place to start. * * * ### How Kalshi infects the news Kalshi’s deals with newsrooms seem to be paying dividends for the company: > “Since December CNBC has published 58 articles that do little more than advertise the existence of a Kalshi market related to a news event. […] Since April, CNBC has employed a dedicated reporter to produce these articles. CNBC also maintains a page on its website featuring Kalshi prediction markets selected by CNBC editors, along with its web coverage. […] In at least 22 cases, CNBC has written about Kalshi and not disclosed its financial conflict.” CNN doesn’t pay for access, and instead is paid to exclusively promote Kalshi. CNBC reporting carries a disclosure which states that its relationship goes further: “CNBC and Kalshi have a commercial relationship that includes customer acquisition and a minority investment.” CNBC will gain financially if its coverage leads to more signups or a growth in Kalshi’s valuation. CNN’s is a simpler paid placement, but both deals are aggressive ways for Kalshi to compete with Polymarket, which has been making similar deals with newsrooms like Yahoo Finance. This is even happening when markets are not significant enough to be newsworthy. As the New Yorker noted in December: > “When Enten lauded the benefits of analyzing betting odds, on air the other day, he failed to mention that only several hundred thousand dollars had been bet on that particular market. Kalshi’s odds provided good fodder for television, but, statistically speaking, they didn’t say much.” It reminds me of the deals Twitter made with newsrooms relatively early in its life. Suddenly, almost out of nowhere, anchors read out tweets on the news, and shows promoted their official Twitter accounts over their websites. This didn’t happen organically: Twitter partnerships teams made deals behind the scenes to ensure their product was showcased well. It was one of the first times that a web startup impactfully executed on a media strategy, and startups have built on that pattern ever since. Here, rather than serving a social network, money is changing hands for newsrooms to promote gambling markets — and in CNBC’s case, they will make more money if more people gamble. It’s obviously weirder, and the incentives here would pull at traditional newsroom ethics in an uncomfortable way even if adequate disclosures were published. This comes at an unfortunate time when trust in news is falling quickly, and newsrooms like CNN are increasingly seen as serving their owners rather than bastions of trustworthy reporting. These Kalshi deals are weird, and an obvious conflict of interest that will likely drive people to trust the news even less than they do today. The Reuters Institute’s 2026 Digital News Report found that 70% of respondents think media owners and corporate parents exert undue influence on the news. As more of these sorts of deals are made, and as trust in news continues to decline, newsrooms are going to need to more overtly state that their coverage is free from this sort of sponsored content. Stronger, more transparent ethics statements, and louder conversations about how reporting decisions are made, will help some newsrooms to explain how they stand apart from these dynamics. In the meantime, CNN and CNBC are helping to drive trust in media into the gutter. * * * ### AI Content Is Everywhere on Social Media, Especially LinkedIn This is one of the core effects of AI: even when people are not engaging with AI-generated content directly, it’s hard to avoid. Our feeds are increasingly full of AI slop. > “AI-generated content appeared across all social media platforms in our data set. The average AI rate across all scanned items was 13.8%, but specific rates varied by platform and item length. On four out of five platforms, longer content was more likely to be AI-generated than shortform content. Across all platforms, one in four longform items (25.72% of items over 250 words) were fully AI-generated.” Specifically, long-form content on LinkedIn was 41% likely to be AI-generated, which shouldn’t surprise anyone who’s browsed LinkedIn lately. Medium was 31% likely and X was 29% likely. Open social web platforms like Bluesky and Mastodon don’t seem to have been a part of the dataset, but I think it would be foolish to assume they’re immune. Some caveats here: the analysis was done by Pangram, which builds a browser extension and back-end tech that attempts to detect AI-generated content. That’s an imperfect process, and there are no tools that are completely reliable at making this distinction. False positives and false negatives have been common with these tools, although Pangram claims a 0.01% false positive rate. So take it with a pinch of salt, but it’s reasonable to assume that these numbers are _directionally_ true. All of this serves to drive trust in these platforms even lower. Increasingly, people on platforms like LinkedIn are being lazy writers and using AI to produce content that you don’t want to put the effort into. I generally think that if you can’t be bothered to write something, it’s not reasonable to ask people to read it; still, there may be some value in AI _assisted_ writing, depending on the piece and how it was produced. (That kind of AI content, by the way, was not really measured by this test.) But AI has also led to a lot of outright spam making its way into people’s feeds in order to shamelessly build clout and advertising revenue. Both things are making these platforms unusable, which in turn is driving people to smaller communities and group chats with people they _know_ they can trust. I believe that’s going to be a big trend: AI leading to a noticeable drop in quality that drives people away from the platforms where it’s allowed to thrive. In that world, platforms that foster trusted relationships and communities will win. _Via_ 404 Media, which has characteristically great coverage of the story_._

werd.io

Chicago Public Media is launching a region-wide social platform as a way to bolster its journalism. I believe it represents a path to the future.

Communities will build trust and loyalty for local public media. Chicago Public Media is taking a big leap forward.

Link: Chicago Public Media launching community website, by Amy Yee at the Chicago Sun-Times In an increasingly AI-dominated information landscape, real trusted communities and relationships will be the way to build trust and loyalty — I’m convinced of it, and organizations like New_ Public agree. So it was exciting to see Chicago Public Media take a huge step towards building a community platform. > The site will include Chicago-area information, civic and cultural resources, community-sourced knowledge and opportunities for audience participation, the nonprofit said Wednesday. It will also curate headlines from the Sun-Times, WBEZ and other news sources. This will be a familiar argument to regular readers: > For independent journalism to “truly service the public … we should have digital infrastructure that is also steered by public media companies,” Chicago Public Media CEO Melissa Bell said. The news industry “has ceded a lot of distribution to places like Facebook and X, formerly known as Twitter, and I think that has done a disservice to centering civic discourse in a healthy way.” We’ve seen other platforms release similar efforts effectively. The Newsmast Foundation builds community-first social media apps on open protocols for newsrooms that include The Bristol Cable and Find Out Media. Flipboard’s Surf platform powers curated social feeds, again built on open social web protocols, for the likes of 404 Media and Rolling Stone (as well as my own curated non-profit US news feed). And Canada’s Village Media serves 13 local social networks through its SPACES platform. But this is the first time we’ve seen a single social platform rolled out by a public media company at this scale. Chicago Public Media was gifted the underlying chicago.com domain and will be rolling it out to neighborhoods and suburbs throughout the area. It sounds like each community will be highlighted (perhaps with its own feed), with an attached hub that covers the entire region. Clearly, this is an experiment, but I’m delighted to see a public media innovator explore these ideas at this scale. I see it as vindication for the idea that building stronger community applications into the public media model is a path towards a more trusted future for local journalism. I’ll be watching very closely, and I’m curious to see who dives in next.

werd.io

Long-form LinkedIn posts are 41% slop. Other social networks are not far behind.

As social networks fill up with AI slop, trusted relationships and communities will win.

Link: AI Content Is Everywhere on Social Media, Especially LinkedIn, by Max Spero at Pangram This is one of the core effects of AI: even when people are not engaging with AI-generated content directly, it’s hard to avoid. Our feeds are increasingly full of AI slop. > “AI-generated content appeared across all social media platforms in our data set. The average AI rate across all scanned items was 13.8%, but specific rates varied by platform and item length. On four out of five platforms, longer content was more likely to be AI-generated than shortform content. Across all platforms, one in four longform items (25.72% of items over 250 words) were fully AI-generated.” Specifically, long-form content on LinkedIn was 41% likely to be AI-generated, which shouldn’t surprise anyone who’s browsed LinkedIn lately. Medium was 31% likely and X was 29% likely. Open social web platforms like Bluesky and Mastodon don’t seem to have been a part of the dataset, but I think it would be foolish to assume they’re immune. Some caveats here: the analysis was done by Pangram, which builds a browser extension and back-end tech that attempts to detect AI-generated content. That’s an imperfect process, and there are no tools that are completely reliable at making this distinction. False positives and false negatives have been common with these tools, although Pangram claims a 0.01% false positive rate. So take it with a pinch of salt, but it’s reasonable to assume that these numbers are _directionally_ true. All of this serves to drive trust in these platforms even lower. Increasingly, people on platforms like LinkedIn are being lazy writers and using AI to produce content that you don’t want to put the effort into. I generally think that if you can’t be bothered to write something, it’s not reasonable to ask people to read it; still, there may be some value in AI _assisted_ writing, depending on the piece and how it was produced. (That kind of AI content, by the way, was not really measured by this test.) But AI has also led to a lot of outright spam making its way into people’s feeds in order to shamelessly build clout and advertising revenue. Both things are making these platforms unusable, which in turn is driving people to smaller communities and group chats with people they _know_ they can trust. I believe that’s going to be a big trend: AI leading to a noticeable drop in quality that drives people away from the platforms where it’s allowed to thrive. In that world, platforms that foster trusted relationships and communities will win. _Via_ 404 Media, which has characteristically great coverage of the story_._

werd.io

Three short fiction pieces

IndieWeb Fiction Carnival: May 2026 Roundup

Back in May, I signed up to host the IndieWeb Fiction Carnival, and kicked it off with a prompt: sticks and stones will break my bones_._ I'm incredibly late with my roundup: just as soon as I'd kicked it off, life became chaotic in a way that won't settle down until September. Which is a poor excuse, because I received some great submissions. Daniel Miller wrote Pull: > The first chamber was larger than the second. Wren heard and felt the pneumatic doors close behind them and was careful to place one foot next to the other, in a stable stance, and took hold of the handrails. They felt their boots lock onto the walkway. The loud, mechanical horn blared its single warning and clouds of thin white mist filled the room. Wren stared straight ahead but could see in the peripheral vision possible through their helmet’s visor the particles cling to their suit. In seconds they covered the visor as well. Zachary Kai wrote Tear Me Down, Build Me Back Up: > They repeated the usual version of this phrase, turned on him in the second person, until it became so pervasive they might've well have tattooed it on the insides of his eyelids. April wrote Lara's Neighbour: > Being a young lady in the neighborhood. One of the elderly man came around one day towards Lara and asked when is she getting hitched? Flabbergasted by his question - she said she won't be getting married and then this oldie guy got on her nerves by saying that a young lady is a BURDEN on the family and getting married removes that burden. _Also syndicated to IndieNews._

werd.io

I've rarely seen newsroom business teams discuss metrics like LTV, CAC, and even ARR outside fundraising or finance. They should be commonly known.

Startups track business metrics. Newsrooms should learn from them.

Link: Your SaaS Metrics Are A Result, Not A Strategy, by Itay Sagie in Crunchbase News I still subscribe to sites like Crunchbase News from my time in startup-land; although it’s been a while since I’ve run the financial side of a business, I’m interested, and I know that I’ll run one again. I see stories like this and wonder: what would it look like for a newsroom to think this way? In startups, these metrics are known top to bottom, but I’ve rarely heard business teams talk about LTV (customer Life Time Value), CAC (Customer Acquisition Cost), or even ARR (Annual Recurring Revenue). This may be happening in finance and fundraising teams, but the culture of talking about customers / donors in teams more widely often isn’t there. While the firewall between business and editorial is important to maintain, the business side should be more customer / donor focused. Beyond that, this piece points out, rightly, that metrics are not strategy: they’re the measurable outcome of your strategy. They’re important tools to help you figure out cause and effect and improve your revenue efficiency, but they are not the underlying mechanism. Interesting provocation here from the author: > “The Rule of 4 adds a simple durability check: ARR growth divided by annual customer churn should be above four. If it is low, growth may be hiding a leaking bucket. > > [The board should ask:] are we growing on top of a loyal customer base, or replacing customers we should have kept?” Growth in annual recurring revenue — the portion of your revenue that is from recurring customers like subscribers or monthly / annual donors — is expressed as a percentage. So is churn: what percentage of customers (paid subscribers, members, recurring donors) cancel their commitments and don’t return? How many newsrooms have those numbers handy? What would it take to measure them? Which systems are missing that would let you do that? There is _so much_ that newsrooms — including non-profit publications — can learn from for-profit startups and other businesses. There’s a lot to be gained by sharing knowledge from those other domains. Figuring out which metrics successful businesses track and mapping the data gaps inside a newsroom is a good place to start.

werd.io

Some newsrooms are making money from deals with prediction markets. It will only serve to further destroy the public's trust in news.

CNN and CNBC promote gambling to make a cheap buck

Link: How Kalshi infects the news, by Aaron Rupar and Judd Legum in Public Notice and Popular Information Kalshi’s deals with newsrooms seem to be paying dividends for the company: > “Since December CNBC has published 58 articles that do little more than advertise the existence of a Kalshi market related to a news event. […] Since April, CNBC has employed a dedicated reporter to produce these articles. CNBC also maintains a page on its website featuring Kalshi prediction markets selected by CNBC editors, along with its web coverage. […] In at least 22 cases, CNBC has written about Kalshi and not disclosed its financial conflict.” CNN doesn’t pay for access, and instead is paid to exclusively promote Kalshi. CNBC reporting carries a disclosure which states that its relationship goes further: “CNBC and Kalshi have a commercial relationship that includes customer acquisition and a minority investment.” CNBC will gain financially if its coverage leads to more signups or a growth in Kalshi’s valuation. CNN’s is a simpler paid placement, but both deals are aggressive ways for Kalshi to compete with Polymarket, which has been making similar deals with newsrooms like Yahoo Finance. This is even happening when markets are not significant enough to be newsworthy. As the New Yorker noted in December: > “When Enten lauded the benefits of analyzing betting odds, on air the other day, he failed to mention that only several hundred thousand dollars had been bet on that particular market. Kalshi’s odds provided good fodder for television, but, statistically speaking, they didn’t say much.” It reminds me of the deals Twitter made with newsrooms relatively early in its life. Suddenly, almost out of nowhere, anchors read out tweets on the news, and shows promoted their official Twitter accounts over their websites. This didn’t happen organically: Twitter partnerships teams made deals behind the scenes to ensure their product was showcased well. It was one of the first times that a web startup impactfully executed on a media strategy, and startups have built on that pattern ever since. Here, rather than serving a social network, money is changing hands for newsrooms to promote gambling markets — and in CNBC’s case, they will make more money if more people gamble. It’s obviously weirder, and the incentives here would pull at traditional newsroom ethics in an uncomfortable way even if the adequate disclosures were published. This comes at an unfortunate time when trust in news is falling quickly, and newsrooms like CNN are increasingly seen as serving their owners rather than bastions of trustworthy reporting. These Kalshi deals are weird, and an obvious conflict of interest that will likely drive people to trust the news even less than they do today. The Reuters Institute’s 2026 Digital News Report found that 70% of respondents think media owners and corporate parents exert undue influence on the news. As more of these sorts of deals are made, and as trust in news continues to decline, newsrooms are going to need to more overtly state that their coverage is free from this sort of sponsored content. Stronger, more transparent ethics statements, and louder conversations about how reporting decisions are made, will help some newsrooms to explain how they stand apart from these dynamics. In the meantime, CNN and CNBC are helping to drive trust in media into the gutter.

werd.io

The apparatus to cement centralized, undemocratic power has already been built. Regulations have a part to play, but they can't just be recommendations.

Governance can prevent AI from being used to undermine democracy. But only if it has teeth.

Link: In Geneva, the World Can Anchor AI Governance in Free Expression, by Isabelle Anzabi in Tech Policy Press I like that this is happening, will appreciate the recommendations that arise from it, and know with complete certainty that no major AI vendor will adhere to them unless they are forced: > “The United Nations is convening its Global Dialogue on Artificial Intelligence Governance in Geneva July 6-7, and the stakes are high: what rights will anchor the future of how the world governs AI? Generative AI systems mediate, filter, and generate the information we encounter every day. They are, at their core, a technology of expression and access to information. How we govern them will have downstream effects on what people can say, seek, and know. Geneva is where we can get that right.” A previous UNESCO meeting recommended governance that “bars AI systems from being used for social scoring or mass surveillance, and requires member states to ensure that AI actors respect rights in the AI lifecycle”. As we all know, that has _definitely_ happened. Both AI vendors and the two countries that predominantly house them, the United States and China, are famously against surveillance and social scoring, and in favor of maintaining rights. Certainly none of them have, for example, used AI with the most sensitive personal information government has access to, or worked with the private sector to create a surveillance apparatus for policing that includes the broad detainment of immigrants. Neither of them is famous for social scoring — in China by government or in the United States through private enterprise. Job done! Sarcasm aside, it’s important to have these meetings, and it’s important to continue to add pressure to protect rights and prevent abuses of this technology. But we should be clear-eyed about the fact that any recommendations almost certainly won’t actually be followed unless major nations that represent real dollars to these vendors enshrine them in law and hold the line when countries like the US apply pressure to undermine them. It can’t just come down to individual governments. The linked article argues that we should also protect freedom of expression, in particular from government (although I’d argue we should also worry about the influence of wealthy private entities). Because these are black box systems, it’s vital that we prevent governments (or anyone with power) from opaquely tweaking their answers and outputs to benefit their agendas. They should not be allowed to be opaque systems; full transparency and auditability should be requirements. The danger is that, in the wrong hands, AI can be used to cement centralized, undemocratic power. To prevent this, ideally, organizations like the UN should apply real sanctions to nations that don’t obey transparency rules and tweak AI systems in service of undemocratic goals — but the UN’s history of doing this effectively is not strong. Part of the problem is the “AI is the information technology of the future” framing used here. If you believe that AI is the future of information systems, you are also more likely to believe that your nation will miss out if you don’t embrace it completely. (That’s been the marketing: if you don’t jump into AI, you will be left behind.) But the reality is, of course, far more nuanced. AI will absolutely change industries like software engineering, and has already made an impact there, although it’s far from an existential transformation. I’m less receptive to the idea that it will transform entire nations. AI vendors need global markets more than global markets need AI. Governments should understand the power they have, and use it.

werd.io

Let's keep building.

Happy 250th birthday, America

Celebrating the 250th anniversary of a declaration that had a passage decrying slavery removed because southern states complained by eating carcinogenic mass produced meat tubes while the President tells half of us we’re evil Marxists while we swelter through a man made heat wave is incredibly American. Happy 4th! All snark aside, genuinely, I do love America. The Declaration of Independence - 250 years ago! - was genuinely important. Most places you go, the people are kind, community-minded, and optimistic. One day, eventually, they will have an equitable government that redresses wrongs and provides real support. My entire political worldview is: what if you could maintain the optimism and energy of America but add an inclusive culture of support that gives everyone education, healthcare, a real safety net and a springboard for their lives, in a way that builds communities rather than extracts from them?It's not our current reality, but I'm certain we can get there. One day, we'll look back, and the corporatism, the militarism, the racism, and the subjugation will be a thing of the past. In their place will be thriving communities. And, finally, all of this will be great.

werd.io

AI, surveillance, open tech, and news as a business.

Notable links: July 3, 2026

_Most Fridays, I share a handful of pieces that caught my eye at the intersection of technology, media, and society._ _Did I miss something important?__Send me an email_ _to let me know._ _Did someone forward this to you?_ Subscribe for free_._ * * * ### OpenAI proposes handing Trump administration 5% stake In order to ward off backlash against AI and curry favor with the Trump administration, Sam Altman has floated the idea of giving 5% of OpenAI to a wealth fund that pays dividends to both the government and citizens — and that every leading AI vendor should do the same. > “Sam Altman, chief executive of the ChatGPT maker, has argued that giving the public a financial stake in the company is the best way to share the upside of AI and has suggested a stake of this size in early conversations with the administration, according to two people familiar with the talks.” It’s transparently a way to align everyone with AI vendor profits. If the sector increases in value, the government and the voting population benefit. If it _decreases_ in value … well, the government is incentivized to prevent that from happening. It also wouldn’t be without precedent: it’s modeled on the Alaska Permanent Fund, which does this with oil profits for Alaskan residents. Intel is also now 10% government-owned, and the administration has reversed course to be behind it since gaining that stake. Would a government whose revenues are directly linked to the performance of a sector be likely to enact hard regulations on that sector? Perhaps not. It’s not a slam dunk, though: for example, the UK receives significant tax revenue on fossil fuels, but still promoted electric cars. There are lots of factors at play, and profit alignment isn’t necessarily outweighed by the effects of other harms. (See also: cigarettes, which are taxed but also tightly controlled as an addictive carcinogen.) Meanwhile, Bernie Sanders has pushed for closer to 50% ownership through a sovereign wealth fund. At this much lower stake, Sam Altman’s proposal uses Sanders’s democratic socialist “share the wealth” language as a way to launder OpenAI’s profits through a thin veneer of good ethics. What’s also interesting to me is that all of these arguments assume that AI is going to be an enormous driver of wealth and innovation — but what if it isn’t? It’s another great way to advertise the technology as something world-changing that everybody must get behind right now. Even if AI turns out to be what the people heavily invested in its success say it will be, it doesn’t stand alone as a sea change innovation. The personal computer, the iPhone, word processors, and spreadsheets were pretty transformational technologies. Should there have been a wealth fund attached to each of those? What, exactly, makes AI different? The answer is that it represents labor displacement: people will lose their jobs. And if that’s actually going to be the case, we need bigger, more structural safety nets and reforms. Dividends from 5% of a sector aren’t going to replace wages at scale — and are heavily dependent on valuations continuing to rise. This proposal ties the welfare of people who have lost their jobs to the success of the companies that drove those losses. The incentives are perverse. We shouldn’t accept this proposal. Instead, we should push for stronger protections and stronger regulation. If a sector can’t succeed without real damage to working communities, then it must not be allowed to. And if these claims turn out not to be true, then it’s an empty gesture designed to add credibility to a self-interested science fiction view of the future. * * * ### Companies Are Making Claude and Codex Talk Like Cavemen to Stop AI’s Soaring Costs I find this very funny: > “Companies are deliberately making their AI tools speak like cavemen in an attempt to stop burning through AI tokens and curb their massive expenditure on AI, 404 Media has found. The tool turns the usually verbose outpost of LLMs like Claude Code, Codex, or Gemini into a much more to the point answer. Think less “you’re right to push back, I was wrong,” and more “Hulk smash.”” If only we had other limited-vocabulary lexicons designed to talk to computers efficiently! I think we’re circling a few different possibilities that may show up over the next few years: * Literally LLM-specific “programming languages” that humans can use to talk to models more efficiently, of which Caveman is the hilarious first step * A proprietary bytecode-like language for LLMs that makes interactions more efficient but also just happens to be owned by one of the major vendors and creates a hitherto-unobtainable moat for their business * This all becomes moot when local models become viable for most businesses without insanely high hardware prices or configuration costs * LLM costs eventually fall to a fraction of their existing level But who knows? Maybe enterprise businesses will continue to talk in stilted caveman language to achieve their business goals forever. * * * ### Journalism Has the Receipts. It Won’t Use Them. Arts organizations learned long ago to prove their economic value with hard numbers: attendance, tourism revenue, multiplier effects. News, as Yoni Greenbaum argues here, likes to cling to civic virtue and assume that the work should speak for itself. > “Journalism operated on a commercial advertising revenue model for over 150 years. Publishers sold readers to advertisers, while editors fretted about maintaining a church-and-state divide between the newsroom and business desk. Journalists saw themselves as watchdogs, not wealth generators. Pitching our value based on our own economic impact felt gauche, too close to an advertorial.” Yoni points out that this is starting to change. We know that news deserts cost communities at least $1.1B a year, for example, because of a report by Rebuild Local News and the University of Illinois Chicago. But newsrooms themselves tend to shy away from reporting their own economic impact — even though they already have the tools to do so. It’s not obvious to me that this accounting would work as an argument across the board for newsrooms, and particularly for those with a national focus. Does ProPublica (my employer until the end of the month) save anyone money? It certainly does prevent corruption, and there are instances with real dollar amounts attached to them: Intuit, for example, paid back $141 million to its customers over deceptive marketing. But I’m not sure that its impact can be quantified easily overall, despite the newsroom’s obvious public benefit. On the other hand, for _local_ newsrooms, this makes a lot of sense to me: at their best, they act as connective tissue for their communities. That $1.1B a year was _just_ increased interest costs from lenders who felt they could charge more to unmonitored governments. They just need to get more comfortable at telling the economic side of their stories. And there’s a wider point here, which is that almost all nonprofit newsrooms need to be able to get more concrete and detailed about their business models. If you’re running an organization that wants to be sustainable, it’s not enough to care about the journalistic process, and your business accounting cannot be limited to activities like events and merchandise. You actually have to care about building a business holistically, and everything that entails. * * * ### Cascade PBS launches Local Public as standalone streaming tech company I’ve got some complicated feelings about this announcement from Cascade PBS: > “Cascade PBS, the non-profit television station serving western Washington state, has spun out its technology division into a separate company that will help similar public broadcasters carve out their own streaming and digital identities. > > The new company, Local Public, will help develop streaming applications for connected TVs, mobile devices and the web, allowing public television stations to offer locally branded streaming experiences featuring their own programming alongside national PBS content.” On one hand, I absolutely love that they were able to spin out their technology division. Most public media companies don’t have the resources or skills to build their own tech, and building this capability outside of any one station so that all of them can take advantage of it makes a lot of sense to me. The Local Public site itself also makes the ROI transparent. WETA, the public media station for Greater Washington, ran the numbers and said that it would break even in the first year, and a calculator is available for other stations that want to check for themselves. The pricing hinges on Passport-eligible donors: those giving at least $60 a year. Local Public charges $60,000 a year for stations with fewer than 15,000, $75,000 for up to 40,000, and $100,000 a year for everyone else — which is not out of bounds. It all seems like a decent business, run in the public interest as a subsidiary of Cascade PBS, that will genuinely help public media stations. I want to see more of this. But I do wish it was fully based on open technology. While stations gain the right to modify the source code of their _apps_ after a year, they remain locked into Local Public’s back-end services. For the CMS, which builds network effects the more stations use it, stations can only get support, maintenance, and customization through Local Public. Over time, that lock-in does not incentivize great support, and Local Public will need to work hard to buck the trends. NPR’s CMS, for example, is notorious among the stations that have to use it. I’m certain the will is there to do better, but they will need to proceed with intention. In my opinion it would be better if, at least after establishing a customer base, they open sourced their back-end CMS too. I tend to think that _any_ technology provided to support the public interest should be fully open. That doesn’t mean there isn’t a tidy business available to its creator — ask Ghost, which is generating millions of dollars off the back of its open source CMS. If there’s a class of organization that absolutely doesn’t deserve to be locked into a technology stack, it’s public service broadcasters. This isn’t Cascade PBS’s fault. It needs its spinout to be sustainable, and this model feels like it will hit that goal. The best scenario, in my mind, would be if there were central funders who bankrolled open tech that the whole ecosystem could use. But, of course, it’s 2026, and central funding for _anything_ public media is hard to come by. Still, this is wonderful to see, and anything that encourages collaboration on a technical level between public service media organizations deserves support. * * * ### Emergency Mode for news Emergency Mode is a set of resources, tools, and training that aims to prepare small newsrooms for various disasters. It’s a co-production between OpenNews, NC Local, and Newspack. They’ve done a great job. As their about page puts it: > “Emergency Mode for News equips local journalists and their newsrooms with the tools they need to respond to climate disasters. With a disaster reporting action pack, software and a learning community, Emergency Mode is designed to help journalists act nimbly and creatively to serve their communities when the unexpected happens.” Toolkits include things like a practical checklist for newsrooms covering wildfires and a template for maintaining source lists during an emergency. There’s also a hands-on workshop series and tools like WordPress plugins for live rolling news updates and providing bandwidth-light versions of sites. Most of all, I really appreciate the practical nature of all of it. Rather than hand-waving about principles and ideas, as many newsroom-facing resources do, everything here is a concrete tool that can actively be used in the field. Newsrooms are more squeezed than they’ve ever been, so it doesn’t hurt that it’s all free. I’d love to see this level of concrete specificity for _the normal working of a newsroom_. Wouldn’t it be cool to have a list of business model checklists you could pull from? Or disaster recovery plans? Or data protection policies? Just as the tools on this site are going to be concretely useful to any newsroom that covers a disaster, checklists, tools, and training for standard operational practices could be really meaningful — particularly for smaller newsrooms that don’t have the ability to hire CTOs, CFOs, and so on. In other words: more, please. This is lovely. * * * ### The Quiet Erosion of Collective Action Under Digital Surveillance The most important outcome of increased surveillance is a chilling effect on free speech and expression. As Gina Romero, the United Nations Special Rapporteur on the Rights to Freedom of Peaceful Assembly, notes here, that extends to the organizations that have been established to protect those rights: > “As organizations operate under the constant assumption that they are being monitored, their core functions are profoundly affected. Their ability to serve as watchdogs, provide rights-based services, protect victims of human rights abuses, and educate the public is severely constrained. Ultimately, the very possibility of advancing and protecting rights, democracy and the rule of law is undermined.” Civil society organizations and advocates have been mislabeled as national security threats around the world. It’s true in some of the nations that we’ve long thought of as being authoritarian, but it’s also true in the United States. Even places like the United Kingdom have tried to apply pressure to technology companies so that they can gain access to backdoors. Tools like Signal have become all the more important. We need more easy to use end to end encrypted systems so that we can communicate and organize with each other without fear of government surveillance. That also allows whistleblowers and sources for journalists to reach out with less of a fear they they will suffer repercussions. But those tools don’t stop you from being surveilled in the real world. Cameras and microphones are everywhere; license plate readers are now commonplace; even AI-enabled drones have been deployed for events like the World Cup. It’s generally true that if government _can_ do something, it will. So the only way to stop this kind of widespread surveillance is to make it impossible. Romero calls for legislative prevention that takes into account the whole systemic impact of surveillance rather than just the immediate first-order effects. Her report also calls out that it can be very difficult to challenge these systems because what they are and who owns them tends to be complicated or obfuscated: > “The study reveals a lack of transparency surrounding the relationship between state power and non-state actors, creating an information vacuum that makes surveillance practices exceedingly difficult to challenge through litigation. As a result, the right to an effective remedy is fundamentally weakened.” So I think we also need more technical capabilities that interfere with how these systems of surveillance actually work. We need more spaces that are designated privacy-first and enforce an anti-surveillance rulebook. And, just as communities have taken it upon themselves to dismantle Flock cameras, we need to take back our streets. * * * ### OpenAI will delay GPT-5.6 after Trump administration request I’ve got (at least) two worries about the story that the Trump Administration halted the release of models from both Anthropic and OpenAI. Anthropic recently pulled its Fable model release in response to the government. Now it turns out that OpenAI has done something similar: > “The Information reported that OpenAI CEO Sam Altman told employees Wednesday in a company Q&A that it would release GPT-5.6 in limited preview form — granting access only to a small group of enterprise customers — in compliance with a request from the federal government. During that preview period, the Trump administration itself would reportedly approve access for customers on a case-by-case basis.” In some ways, what a coup for the AI industry. This technology is so powerful that the government doesn’t think anyone should have it — and when it does inevitably release into the public’s hands, what a valuable product that will be. Get the tech that’s too dangerous to be released! This magical product can be yours for an unbelievable price! So one worry is that this is, in essence, great marketing for these vendors. But it’s worth remembering that these AI models are black boxes that respond to information queries in opaque ways. The more people rely on them for knowledge, the more powerful the models become. The argument being presented is that they can be used in ways that might present traditional security threats — but consider that some versions of the truth, to the wrong kind of authoritarian-minded government, might also be considered a threat. (Remember that “extremism on migration, race, and gender” and hostility to “traditional American views” are now considered markers of domestic terrorism.) This is a golden opportunity, in other words, to hit pause on frontier model releases, at a time when models are becoming more prevalent, in order to make sure models are shaped to represent a certain version of the world. The administration has already signaled a willingness to do this; there is nothing to say they aren’t. The only way to _prove_ that they aren’t is to open source not just the models but the training process and make the whole thing transparent and verifiable. The industry is a long way off from doing that. * * * ### Now we’re getting AI fake news complaining about how AI fake news is the death of real news A bunch of people — including, unfortunately, me — were taken in by this AI-generated newsroom earlier this week. The story was decently written and seemed to be well-cited, but it turned out to be nonsense. Ironically, it was about a would-be media empire that purchased struggling papers, fired their staff, and replaced them with AI, leading to the death of each newsroom. All false. So the big question about The Editorial is: why does it exist? As Joshua Benton put it: > “Fake news isn’t new, obviously. And while AI-generated slop is newer, it’s hardly unfamiliar at this point. But why would a spam site bother making up a story about Alabama weekly newspapers, of all things? Whose interest is it in to get that niche?” Here’s my theory: I think it’s a two-headed LLM poisoning scheme. On one hand, most of the content relates to Chinese-specific interests: articles about Taiwan or African nations where China is making inroads. These are all articles from a China-friendly perspective. If an LLM were to ingest them and _trust the site_ , it might start repeating the assertions made in each one as fact. One way to make sure a site is trusted is to get other, trusted sources to point to it. That’s where the stories about journalism come in: there are few things that journalists engage in more than stories about their own industry. Get enough patsies (like, again, to my chagrin, _me_) to point links in their direction and journalists might post them in high-trust communities on high-trust sites like Reddit, as well as their own, and Bob’s your uncle. We already know that it takes as little as 13 words to poison an LLM with falsehoods. Of course, that might not be it at all. Frank, the site’s owner (who lists himself as CEO of Nordiso Group on LinkedIn), at least appears to be a Finnish solopreneur. If he wanted to clear the air, he could write a post (himself) about what he was up to. It might be that he’s running an experiment to see how easily an LLM can be poisoned with propaganda! Until then, I think it’s reasonable to assume that something underhand is going on.

werd.io

We need to see more technical collaborations between public service media organizations. It's also really important that they're based on open technology that doesn't lock them in.

A new media spinout provides streaming apps for public service broadcasters. I just wish it was open.

Link: Cascade PBS launches Local Public as standalone streaming tech company, by Matthew Keys at The Desk I’ve got some complicated feelings about this announcement from Cascade PBS: > “Cascade PBS, the non-profit television station serving western Washington state, has spun out its technology division into a separate company that will help similar public broadcasters carve out their own streaming and digital identities. > > The new company, Local Public, will help develop streaming applications for connected TVs, mobile devices and the web, allowing public television stations to offer locally branded streaming experiences featuring their own programming alongside national PBS content.” On one hand, I absolutely love that they were able to spin out their technology division. Most public media companies don’t have the resources or skills to build their own tech, and building this capability outside of any one station so that all of them can take advantage of it makes a lot of sense to me. The Local Public site itself also makes the ROI transparent. WETA, the public media station for Greater Washington, ran the numbers and said that it would break even in the first year, and a calculator is available for other stations that want to check for themselves. The pricing hinges on Passport-eligible donors: those giving at least $60 a year. Local Public charges $60,000 a year for stations with fewer than 15,000, $75,000 for up to 40,000, and $100,000 a year for everyone else — which is not out of bounds. It all seems like a decent business, run in the public interest as a subsidiary of Cascade PBS, that will genuinely help public media stations. I want to see more of this. But I do wish it was fully based on open technology. While stations gain the right to modify the source code of their _apps_ after a year, they remain locked into Local Public’s back-end services. For the CMS, which builds network effects the more stations use it, stations can only get support, maintenance, and customization through Local Public. Over time, that lock-in does not incentivize great support, and Local Public will need to work hard to buck the trends. NPR’s CMS, for example, is notorious among the stations that have to use it. I’m certain the will is there to do better, but they will need to proceed with intention. In my opinion it would be better if, at least after establishing a customer base, they open sourced their back-end CMS too. I tend to think that _any_ technology provided to support the public interest should be fully open. That doesn’t mean there isn’t a tidy business available to its creator — ask Ghost, which is generating millions of dollars off the back of its open source CMS. If there’s a class of organization that absolutely doesn’t deserve to be locked into a technology stack, it’s public service broadcasters. Still, this is wonderful to see, and anything that encourages collaboration on a technical level between public service media organizations deserves support.

werd.io

Newsrooms like to spend their time on the journalistic process and assume that the value of their work will speak for itself. They need to start selling themselves.

Newsrooms need to get comfortable expressing their business value - and raising money on it.

Link: Journalism Has the Receipts. It Won’t Use Them., by Yoni Greenbaum in Backstory & Strategy Arts organizations learned long ago to prove their economic value with hard numbers: attendance, tourism revenue, multiplier effects. News, as Yoni Greenbaum argues here, likes to cling to civic virtue and assume that the work should speak for itself. > “Journalism operated on a commercial advertising revenue model for over 150 years. Publishers sold readers to advertisers, while editors fretted about maintaining a church-and-state divide between the newsroom and business desk. Journalists saw themselves as watchdogs, not wealth generators. Pitching our value based on our own economic impact felt gauche, too close to an advertorial.” Yoni points out that this is starting to change. We know that news deserts cost communities at least $1.1B a year, for example, because of a report by Rebuild Local News and the University of Illinois Chicago. But newsrooms themselves tend to shy away from reporting their own economic impact — even though they already have the tools to do so. It’s not obvious to me that this accounting would work as an argument across the board for newsrooms, and particularly for those with a national focus. Does ProPublica (my employer until the end of the month) save anyone money? It certainly does prevent corruption, and there are instances with real dollar amounts attached to them: Intuit, for example, paid back $141 million to its customers over deceptive marketing. But I’m not sure that its impact can be quantified easily overall, despite the newsroom’s obvious public benefit. On the other hand, for _local_ newsrooms, this makes a lot of sense to me: at their best, they act as connective tissue for their communities. That $1.1B a year was _just_ increased interest costs from lenders who felt they could charge more to unmonitored governments. They just need to get more comfortable at telling the economic side of their stories. And there’s a wider point here, which is that almost all nonprofit newsrooms need to be able to get more concrete and detailed about their business models. If you’re running an organization that wants to be sustainable, it’s not enough to care about the journalistic process, and your business accounting cannot be limited to activities like events and merchandise. You actually have to care about building a business holistically, and everything that entails.

werd.io

A wealth fund that shares 5% of AI success with government and voters is either based on hype or not nearly enough to cover the damage. Either way, the incentives are perverse.

OpenAI wants to give us 5% of its success. It's a bad bargain.

Link: OpenAI proposes handing Trump administration 5% stake, by Cristina Criddle in the Financial Times In order to ward off backlash against AI and curry favor with the Trump administration, Sam Altman has floated the idea of giving 5% of OpenAI to a wealth fund that pays dividends to both the government and citizens — and that every leading AI vendor should do the same. > “Sam Altman, chief executive of the ChatGPT maker, has argued that giving the public a financial stake in the company is the best way to share the upside of AI and has suggested a stake of this size in early conversations with the administration, according to two people familiar with the talks.” It’s transparently a way to align everyone with AI vendor profits. If the sector increases in value, the government and the voting population benefit. If it _decreases_ in value … well, the government is incentivized to prevent that from happening. It also wouldn’t be without precedent: it’s modeled on the Alaska Permanent Fund, which does this with oil profits for Alaskan residents. Intel is also now 10% government-owned, and the administration has reversed course to be behind it since gaining that stake. Would a government whose revenues are directly linked to the performance of a sector be likely to enact hard regulations on that sector? Perhaps not. It’s not a slam dunk, though: for example, the UK receives significant tax revenue on fossil fuels, but still promoted electric cars. There are lots of factors at play, and profit alignment isn’t necessarily outweighed by the effects of other harms. (See also: cigarettes, which are taxed but also tightly controlled as an addictive carcinogen.) Meanwhile, Bernie Sanders has pushed for closer to 50% ownership through a sovereign wealth fund. At this much lower stake, Sam Altman’s proposal uses Sanders’s democratic socialist “share the wealth” language as a way to launder OpenAI’s profits through a thin veneer of good ethics. What’s also interesting to me is that all of these arguments assume that AI is going to be an enormous driver of wealth and innovation — but what if it isn’t? It’s another great way to advertise the technology as something world-changing that everybody must get behind right now. And even if AI turns out to be what the people heavily invested in its success say it will be, it doesn’t stand alone as a sea change innovation. The personal computer, the iPhone, word processors, and spreadsheets were pretty transformational technologies. Should there have been a wealth fund attached to each of those? What, exactly, makes AI different? The answer is that it represents labor displacement: people will lose their jobs. And if that’s actually going to be the case, we need bigger, more structural safety nets and reforms. Dividends from 5% of a sector aren’t going to replace wages at scale — and are heavily dependent on valuations continuing to rise. This proposal ties the welfare of people who have lost their jobs to the success of the companies that drove those losses. The incentives are perverse. We shouldn’t accept this proposal. Instead, we should push for stronger protections and stronger regulation. If a sector can’t succeed without real damage to working communities, then it must not be allowed to. And if these claims turn out not to be true, then it’s an empty gesture designed to add credibility to a self-interested science fiction view of the future.

werd.io

I was taken in by a fake AI news site. I think it may be an attempt to poison LLMs with propaganda.

I have a theory about AI fake news site The Editorial

Link: Now we’re getting AI fake news complaining about how AI fake news is the death of real news, by Joshua Benton in Nieman Lab A bunch of people — including, unfortunately, me — were taken in by this AI-generated newsroom earlier this week. The story was decently written and seemed to be well-cited, but it turned out to be nonsense. Ironically, it was about a would-be media empire that purchased struggling papers, fired their staff, and replaced them with AI, leading to the death of each newsroom. All false. So the big question about The Editorial is: why does it exist? As Joshua Benton put it: > “Fake news isn’t new, obviously. And while AI-generated slop is newer, it’s hardly unfamiliar at this point. But why would a spam site bother making up a story about Alabama weekly newspapers, of all things? Whose interest is it in to get that niche?” Here’s my theory: I think it’s a two-headed LLM poisoning scheme. On one hand, most of the content relates to Chinese-specific interests: articles about Taiwan or African nations where China is making inroads. These are all articles from a China-friendly perspective. If an LLM were to ingest them and _trust the site_ , it might start repeating the assertions made in each one as fact. One way to make sure a site is trusted is to get other, trusted sources to point to it. That’s where the stories about journalism come in: there are few things that journalists engage in more than stories about their own industry. Get enough patsies (like, again, to my chagrin, _me_) to point links in their direction and journalists might post them in high-trust communities on high-trust sites like Reddit, as well as their own, and Bob’s your uncle. We already know that it takes as little as 13 words to poison an LLM with falsehoods. Of course, that might not be it at all. Frank, the site’s owner, at least appears to be a Finnish student. If he wanted to clear the air, he could write a post (himself) about what he was up to. It might be that he’s running an experiment to see how easily an LLM can be poisoned with propaganda! Until then, I think it’s reasonable to assume that something underhand is going on.

werd.io

Me think this hilarious sign of times.

AI's costs are going through the roof - so businesses are telling LLMs to talk like cavemen

Link: Companies Are Making Claude and Codex Talk Like Cavemen to Stop AI’s Soaring Costs, by Joseph Cox at 404 Media I find this very funny: > “Companies are deliberately making their AI tools speak like cavemen in an attempt to stop burning through AI tokens and curb their massive expenditure on AI, 404 Media has found. The tool turns the usually verbose outpost of LLMs like Claude Code, Codex, or Gemini into a much more to the point answer. Think less “you’re right to push back, I was wrong,” and more “Hulk smash.”” If only we had other limited-vocabulary lexicons designed to talk to computers efficiently! I think we’re circling a few different possibilities that may show up over the next few years: * Literally LLM-specific “programming languages” that humans can use to talk to models more efficiently, of which Caveman is the hilarious first step * A proprietary bytecode-like language for LLMs that makes interactions more efficient but also just happens to be owned by one of the major vendors and creates a hitherto-unobtainable moat for their business * This all becomes moot when local models become viable for most businesses without insanely high hardware prices or configuration costs * LLM costs eventually fall to a fraction of their existing level But who knows? Maybe enterprise businesses will continue to talk in stilted caveman language to achieve their business goals forever.

werd.io

Civil society actors in nearly every region of the world now operate under the assumption that they are being surveilled. The result is a less democratic world for everyone.

They tell us surveillance makes us safer. It undermines our democratic rights.

Link: The Quiet Erosion of Collective Action Under Digital Surveillance, by Gina Romero in Tech Policy Press The most important outcome of increased surveillance is a chilling effect on free speech and expression. As Gina Romero, the United Nations Special Rapporteur on the Rights to Freedom of Peaceful Assembly, notes here, that extends to the organizations that have been established to protect those rights: > “As organizations operate under the constant assumption that they are being monitored, their core functions are profoundly affected. Their ability to serve as watchdogs, provide rights-based services, protect victims of human rights abuses, and educate the public is severely constrained. Ultimately, the very possibility of advancing and protecting rights, democracy and the rule of law is undermined.” Civil society organizations and advocates have been mislabeled as national security threats around the world. It’s true in some of the nations that we’ve long thought of as being authoritarian, but it’s also true in the United States. Even places like the United Kingdom have tried to apply pressure to technology companies so that they can gain access to backdoors. Tools like Signal have become all the more important. We need more easy to use end to end encrypted systems so that we can communicate and organize with each other without fear of government surveillance. That also allows whistleblowers and sources for journalists to reach out with less of a fear they they will suffer repercussions. But those tools don’t stop you from being surveilled in the real world. Cameras and microphones are everywhere; license plate readers are now commonplace; even AI-enabled drones have been deployed for events like the World Cup. It’s generally true that if government _can_ do something, it will. So the only way to stop this kind of widespread surveillance is to make it impossible. Romero calls for legislative prevention that takes into account the whole systemic impact of surveillance rather than just the immediate first-order effects. Her report also calls out that it can be very difficult to challenge these systems because what they are and who owns them tends to be complicated or obfuscated: > “The study reveals a lack of transparency surrounding the relationship between state power and non-state actors, creating an information vacuum that makes surveillance practices exceedingly difficult to challenge through litigation. As a result, the right to an effective remedy is fundamentally weakened.” So I think we also need more technical capabilities that interfere with how these systems of surveillance actually work. We need more spaces that are designated privacy-first and enforce an anti-surveillance rulebook. And, just as communities have taken it upon themselves to dismantle Flock cameras, we need to take back our streets.

werd.io