Mulling the tradeoffs between saving a news org and saving the news ecosystem in the AI age: two different missions that don't necessarily map onto each other — and sometimes contradict each other. My latest post for the Tow-Knight Center at CUNY
restructurednews.substack.com/p/mission-co...
How many bots does it take to edit a story well? An experiment in adversarial editing in my latest post for the Tow-Knight Center for Journalism Futures at CUNY. Plus, some useful inhumanity.
bit.ly/4ubk1pN
Is Generative AI just overhyped, mindless math, or world-changing technology? What if the skeptics are right — and what if the skeptics of the skeptics are right? My post for the Tow-Knight Center at CUNY digs into what we don't know we don't know. restructurednews.substack.com/p/nothing-to...
Some thoughts on how our writing will evolve — no more scintillating leads and witty kickers? — if we're writing news mostly for machines and AI agents to read and summarize for human users instead of directly for humans. My latest post for the Tow-Knight Center at CUNY.
bit.ly/4akceOn
We wanted to make it easier to figure out what all the Nieman Lab predictions said, so we built a RAG (and a chatbot.) But really what we wanted to figure out was how to build a RAG and a chatbot. Follow along!
restructurednews.substack.com/p/the-future...
Dispatches from the (near) future: My latest post for the Tow-Knight Center for Journalism Futures at the Newmark School of Journalism at CUNY, about how the way we create and consume news may be changing faster than we think.
restructurednews.substack.com/p/dispatches...
The LLMs reply: What Claude and ChatGPT have to say about my last post about Structured Journalism.
restructurednews.substack.com/p/wwcd
Updating the circa-2010 idea of Structured Journalism for the LLM Age; my latest for the Tow-Knight Center for Journalism Futures at CUNY. (Claude and ChatGPT will push back later in the week...)
bit.ly/3Zh5ria
Trust, and verification: Why do people seem better disposed to chatbots than to news orgs, despite hallucinations? It may have less to do with LLM's features and more the way we interact with them. My latest post for the Tow-Knight Center at CUNY.
restructurednews.substack.com/p/you-talkin...
What's on my mind about AI and journalism: We have to be in two minds about it. A post for the Tow-Knight Center at CUNY about the contradictions we have to juggle when we strategize for an AI-intermediated world of news.
restructurednews.substack.com/p/both-sides...
My partner in crime at @towknight.bsky.social Tow-Knight Center for Journalism Futures, @adielkaplan.bsky.social Adiel Kaplan writes today about the changing — sometimes mysterious — world of metrics in an AI landscape. Plus: A survey! We know you'd love to fill it out. Please do!
bit.ly/4sCKP16
A conclave of editors: A look at how three different LLMs approach the same editing exercise, and how that might help us design systems to take advantage of their different strengths/approaches. My latest post for the Tow-Knight Center at CUNY. restructurednews.substack.com/p/a-conclave...
Musings about money, and the value of content in an AI world; it needs to be married with audience engagement and understanding in this new landscape. My last 2025 post for the Tow-Knight Center at CUNY's Newmark School.
restructurednews.substack.com/p/is-content...
Could we better serve more of our readers by building internal AI avatars of them to advocate for stories that speak to their needs and perspectives? Can we build machines to get us past our blind spots? My latest post for the Tow-Knight Center for Journalism Futures at CUNY.
bit.ly/4qfYnxE
Personalization is coming; that’s a byproduct of the technology. I don’t know that we can stop it, but can we both make it less bad - and also take advantage of the best opportunities that it offers?
I know this can get really bad - I talk often about filter bubbles of one. We’re already polarized, and it could get much worse worse. On the flip side, whose view of society should be privileged? Mine? The NYT’s? The WSJ’s? What’s the middle path that isn’t just gatekeeping?
With some caveats: Monolithic views by news orgs also cause harm; the NYT famously didn't cover LGBT issues in the 70s and ignored AIDS in the early 80s. Personalization has its issues; but so too does one-size-fits-all journalism. It's great when it reflects your views; not so much minorities.
Also, I don't mean to suggest that all parents in Florida care more about their child's education than on diversity initiatives; but that some will and some won't. The point is personal personalization, not proxy personalization. But as you know - the big question is how to do it well.
Thanks - I think - for the shoutout. And to be clear, I think personalization is coming, whether we like it or not. And like you, I think it's important that its done with public service in mind. And that (mostly) happens best when the readers control what they see and how they see it.
@ginaskchua.bsky.social: “If we don’t, we risk ceding the space to companies that will optimize for engagement, not civic participation.”
www.niemanlab.org/2025/12/pres...
Every year, Nieman Labs asks how journalism will change in the coming year. My prediction this year is about how audiences will change — and how we need to adapt if we want to stay relevant and fulfill our public service mission.
www.niemanlab.org/2025/12/pres...
Who should we think of as the core journalism user of AI tools: the savviest, or the laziest reporter in the newsroom — or someone else? Musings on meeting in the middle. My latest post for the Tow-Knight Center at CUNY's journalism school.
restructurednews.substack.com/p/meeting-in...
Some thoughts about where the audience for news is going — and what we can do about it. (And no, it's not for the Nieman Labs prediction issue, although hopefully that comes out soon too...) This one is for CJR's issue on journalism in 2025: www.cjr.org/analysis/wha...
If who you know matters as much as what you know, what's the most effective way to extract, store and present relationship data? An experiment with LLMs, a throwback to old projects, and thoughts about what we could try next. My post for this week. Please subscribe!
bit.ly/48X5ANf
This week's post on my substack: On a fascinating idea to try to ground LLMs in — if not truth, at least in sourcing — and what that could mean for journalism and public interesting information. Plus, structured journalism.
restructurednews.substack.com/p/the-facts-...
Claude has been responding to me this way for six months or so, and before I started using the Projects feature, so there seems to be something quasi-persistent, at least in the account.
Watching Claude adapt to your style as you work with it is both fascinating and terrifying — and speaks both to the ability to fine-tune it with use and the uncertainty about which version is working at any given time...
My latest post: Why my Claude isn't like your Claude
bit.ly/3JPwDAE
My latest post, about the multiple roles editors play in newsrooms, from steering reporters in the right direction, to helping them craft compelling narratives, to finding typos and nit-picking language. And whether AI can help on those tasks.
restructurednews.substack.com/p/in-praise-...
Some news: Starting today, the Tow-Knight Center will be the Tow-Knight Center for Journalism Futures-Knight, with a focus on how generative AI will change how people find, access and consume news and information, and how journalism needs to meet that moment to fulfill our mission
I'm moving my 15-year-old blog to Substack. My first post there — about that BBC/EBU study about the quality of AI-intermediated news, and what it does — and doesn't — say about how people are coming to news. Please subscribe. It's free!
restructurednews.substack.com/p/welcome
