Notable links: May 15, 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.

After the Feed

I think this research-based presentation about the future of the information ecosystem in the age of AI is important for publishers, product leaders, and social platform builders to read and understand. If you assume that AI will dominate how people receive their information, its conclusions are sensible, well thought-through, and even optimistic in some ways. I think all signs — AI adoption curves, data about social media use, qualitative evidence about how people are using AI to gather information today — point to the fact that it will.

This is the crux:

“Agentic interfaces are the new intermediaries for information about the world around you. This looks like a chat with Claude or a briefing from your personal AI agent — an interface built for an audience of one.

These agentic interfaces will increasingly become the nexus through which you access information and connection.”

For me, the framing of AI’s effects on existing information ecosystems while it establishes a new one was helpful. It’s, frankly, brutal: social networks, other online spaces, and the web itself are getting filled with bots and slop as people compete for engagement and eyeballs. In these spaces, AI-powered harassment, doxxing, and cheap, automated content are becoming more prevalent, while AI models are simultaneously making it easier to extract signal from those same spaces.

AI vendors are clearly the “new new gatekeepers”. Like the previous ones, they will dominate how we learn about the world even while some of us turn to open source and liberatory alternatives. But they may not dominate how we connect and share our experiences of the world, and that’s the core of the opportunity: how do we design pro-social frameworks and spaces that sit alongside an agentic information ecosystem?

I’m biased towards New_ Public’s point of view: pro-social spaces, pro-democracy technology, and community as an ingredient for trust are all my jam. But everything laid out in this presentation is already happening. People are already getting AI-generated information summaries; they are already retreating into trust-based group chats and small spaces; much more software is already being produced, straining platforms like GitHub; social platforms are already declining. But the opportunities are genuinely emerging too: I’ve written before about the opportunity for open protocols as building a foundation for bespoke micro-communities, and the core need on the internet has always been to connect with other people.

How this plays out is not yet written, although new defaults are currently being established by the AI vendors. We need more research, more experimentation, and more dedicated space to explore pro-social spaces, trust, and connection. And we need builders. Communities and trust are going to be very central to my work and research over the next year; I’m grateful for this encapsulated research, which I think will help to guide us all.

Writers are fleeing the Substack Tax

If you weren’t all that bothered about Substack platforming and compensating Nazis, The Verge reports that there’s a new reason to be worried: it costs more and its much-touted network doesn’t count for much if you’re not one of its featured writers.

Sean Highkin of The Rose Garden Report is quoted in the piece:

““When I first joined up, [Substack] gave me a big push and featured me and funneled a lot of traffic to me, which led to a good amount of growth,” Highkin says. “But once I wasn’t one of the ‘new recruited talent’ they could tout, they stopped featuring me and I saw my growth stagnate.””

Ghost (with Ryan Singel’s Outpost) cost less than half and drove a significant increase in subscribers. It’s mentioned here alongside Beehiiv and Kit, but is the only truly open-source alternative. That means you can use Ghost’s services (as I do), but if you’re dissatisfied, you can move to another provider.

This is in stark contrast with Substack, which has been promoting social media style following relationships over true subscriptions, and only allows creators to export their subscribers should they choose to move. Similarly, Beehiiv starts with open protocols like RSS switched off by default, locking readers into its ecosystem.

That freedom is important. As Casey Newton says in the piece:

“The more important thing is that we have a home on the open web that we control, and whatever anti-creator changes Substack is forced to make in the future to live up to its valuation we won’t be affected by.”

Every media company, publisher, and individual creator needs to maintain their platform independence if they want to make independent business decisions. It’s good to see more people taking this step, and it’s good to see that they have options.

Radical Speed Month — The Reader Meets the Fediverse

We’re closer to the entire web being a social environment than ever before. That’s very exciting to me on two fronts. The first is that it’s always been the promise of the web that anyone could publish and be heard, and baking in social functionality is a huge part of that. The second is that it undermines the stranglehold that traditional social media platforms have had on the public discourse and democracy itself. We need movements like these to grow.

So I think it’s cool that WordPress.com just shipped some major improvements to its core reader:

“The Radical Speed Month bet: ship three protocol adapters in four weeks, and prove the Reader can become a universal aggregator. RSS / Google Reader API (so any reader app can use WordPress.com as a sync backend), ActivityPub (so Mastodon, Pixelfed, and friends show up natively), and ATProto / Bluesky (because that’s where a real chunk of the social-web conversation has gone). One Reader, every protocol you care about.”

In practice, that means that you can read updated content from the web via RSS, the Fediverse, and ATproto from the WordPress dashboard — and connect any compatible reader app to that dashboard to make reading more seamless. (I’m a die-hard fan of Reeder Classic, and it sounds like that works.) WordPress is now compatible with reading the whole open social web.

But, of course, it’s WordPress, which is a publishing environment at its heart. It’s supported RSS forever, and has supported the Fediverse for a while. Now it supports Bluesky, too. Unlike most readers, which are read-only environments, you can interact with those sources right from your feed, including by publishing posts and replying to other people’s.

That’s something the indie web community has been thinking about forever: people like Aaron Parecki have been building their own interactive readers using open web standards, and I remember working on a simple prototype at an IndieWebCamp in Portland.

But it’s also an idea that has become more powerful as the open social web has grown. There are millions of people to interact with – all of whom might be publishing from their own websites, on their terms, free from intermediation. May it continue to grow and spread.

The First Year

I could include Corey Ford’s posts in my link roundups every single week. Each one is genuinely gold — and I’ve had the pleasure of knowing and working with Corey in various ways for over a decade, so I also know they work. I use many of them in my own day-to-day practice, and I’ll have them front of mind as I move on to my next chapter later this year.

I also want to say: posting every week on the same day, at the same time, for a whole year is an achievement in itself. I’ve been blogging since 1998 and I’m not convinced I’ve ever been that consistent. As he says, consistency compounds:

“I embraced constraints and forced myself to ship every week, without a long-term plan. Half sheet by half sheet. The first few posts felt like shouting into the void. (And if I'm being honest, I sometimes still wonder whether anyone has time to read these long posts at all.) But then I would run into someone in person at a conference. Or I would catch up with an old student on Zoom. And I would hear the same thing, over and over: Thank you for sharing these frameworks. I just sent our latest one to my team.”

I’ve been quietly sharing his posts in our internal #product-reads channel on Slack, which I set up to share links that I think are either inspiring or will be useful in our work. I’ve been in board meetings at other orgs where his work has come up organically and I’ve been able to enthusiastically +1. If you’re not following him, there’s still time to correct that. He’s the real deal, has changed my life multiple times, and has been similarly influential for others. And if you get a chance to work with him, including as a coach or a consultant for your team and culture — run at it.

Bridging on a Budget

I’ve been in awe of Ryan Barrett since I first met him over a decade ago. He cofounded Google App Engine and led engineering at Color Health. His Bridgy tool, which allows people on different protocols and networks to follow and converse with each other, is now the basis of A New Social, the open social web non-profit that he runs with Anuj Ahooja. (Disclosure: I’m on the board.)

This post about how he reduced Bridgy costs is brilliantly detailed. It’s a good look into what’s involved when you need to refactor and reduce cost at scale — and what’s remarkable is how effective this work actually was.

“The end result of all of this is that we grew from 2k users to almost 150k, added a ton of heavy new functionality, and still managed to optimize and cut down costs from $.15 per active user per month to just $.03 or so.”

But it didn’t come easily. When you’re connected to the kinds of firehoses that Bridgy needs to be, and serving the kind of traffic it’s starting to handle, every optimization really counts. Because it’s open-source, you can dig down into individual optimizations and follow along each exploration. It’s painstaking work and a demonstration of their commitment to financial responsibility. Try vibe coding that.

Bridgy (and its parent A New Social) exists to help make the individual protocols less important: everyone should be able to collaborate with everyone else regardless of which platform they’re using. It’s the kind of thing that feels easy in the moment — but as this post proves, it’s far from simple under the hood.

'The Biggest Student Data Privacy Disaster in History': Canvas Hack Shows the Danger of Centralized EdTech

I started in edtech. When I graduated with my Computer Science degree, I returned to the university to work at the Media and Learning Technology Service. There, I discovered that all the edtech software at the time was so bad — the learners hated it, the teachers hated it, the administrators hated it, and I have to assume the people who made it also had a deep-seated contempt for it — that it actively made learning worse. Worse, these platforms were charging institutions huge amounts of money for the privilege.

Because I was an avid blogger at that time and knew that people were learning from each other on the web all the time, I built a prototype social network for learning and tried to give it to them. They told me they didn’t want it (in a way that was much ruder than that). So I quit my job and ended up releasing it under an open source license so it wouldn’t be centralized and hold institutions hostage. That act of hubris set up the entirety of the rest of my career.

Which brings me to this article:

“Thursday afternoon, millions of students at thousands of universities and K-12 schools were locked out of Canvas, a piece of catch-all education technology software that has become the de facto core of many classes. ShinyHunters, a ransomware group, hacked Canvas’s parent company and apparently stole “billions” of messages and accessed more than 275 million individuals’ data, according to the hacking group. The group also locked students out of Canvas.”

Ian Linkletter — a librarian who has been an active, and in my opinion, unceasingly correct edtech critic — is quoted as calling this “the biggest student data privacy disaster in history”. It need not have been the case; Canvas is theoretically open source. But you can’t make money with open source alone, and self-hosting is not something most institutions want to undertake. Canvas is a huge codebase with real quirks that is non-trivial to self-host, and the maintenance and infrastructure costs are real.

It’s also not clear that self-hosted infrastructure would be more resilient: a university could be subject to a ransomware attack with very little recourse. At the same time, the centralized nature of Canvas’s core offering means every institution that uses it, including over half of all US higher education institutions, were in a hard place right in the middle of final exam season. Access is coming back, but at the time of writing, it hasn’t been fully restored. It’s a hard lesson about the dangers of putting everything in the hands of a single cloud provider.

You couldn't create a more anti-news internet if you tried

Matt Pearce, Director of Policy for Rebuild Local News, writes a behavioral economics inspired take on why our current embodiment of the internet is so bad for news and information.

In particular, he sees the introduction of “nudges” as being a pro-information feature that search engines, LLM interfaces, and social media platforms could introduce:

“Social media, too, could choose to feature quality news outlets as “defaults” or provide subtle “nudges” on content that prompt users to donate or subscribe to the news outlets providing high quality news videos on platforms like Instagram, which don’t pay for themselves.”

I happen to particularly agree with his implied criticism of newsrooms going deep on Instagram, which usually leads to vanity metrics going up and to the right but not necessarily to conversions, impact, or revenue. And I think it’s true that nudges across all these platforms would have the effect he’s hoping for. But I think the tragedy is that there’s no real reason why any of these platforms would actually do it.

The internet as it stands is perfectly optimized for the needs of these platforms: engagement, advertising revenue, and rapid growth. Adding pro-social nudges would add friction to their well-oiled loops and take users off-platform. That’s exactly why Google has moved from leading people to the best websites for a query to answering those questions on-page: its own needs are best served by keeping users in one place. For them to make different choices, they would need to be far more benevolent architects than they are.

So, one path forward is that they need to be forced to do it. This would need regulations to govern the features an information platform can provide, and could have very adverse side effects. We’re seeing increased regulations with respect to things like age verification, so introducing regulation is possible — but that age verification tech has become a surveillance layer that impacts freedom of speech for vulnerable groups. And if publishers go too far in that direction, for example by dictating that platforms share more ad revenue, the networks might simply stop supporting news content at all, as we’ve seen in places like Canada.

Another is to build new platforms that make better choices for the whole ecosystem: more interesting for readers, more supportive of publishers. We’re already seeing a resurgence in new open social web platforms as well as a regrowth in older technologies like RSS. But the incumbent platforms aren’t going to simply go away; any new pro-social platform has to directly compete with them while also building an ecosystem. Still, I think it’s more promising, particularly in a world where incumbent platforms are losing goodwill with the public. The kind of thinking that Matt’s done here is very useful in helping to design what those new platforms might look like.

We’re not in a great place and there’s a hard road ahead. I’m sure of one thing: asking existing platforms to do better is not going to work. So we need to take matters into our own hands.