Folks who work at finding solutions to the local news crisis are understandably frustrated at what a difficult, frustrating slog it can be. Earlier this week, Elizabeth Hansen Shapiro, the former executive director of the National Trust for Local News, gave Richard J. Tofel a preview of a report she’s written for Press Forward and said, “I think the challenges now are so systemic that the only way to do responsible, impactful funding going forward is to look at system solutions rather than newsroom-based ones.”

Follow my Bluesky newsfeed for additional news and commentary. And please join my Patreon for just $6 a month. You’ll receive a supporters-only newsletter every Thursday.

I’m looking forward to reading Hansen Shapiro’s report. (She’s featured in our book, “What Works in Community News,” and has been on our podcast.) And yet there really is no substitute for solving this problem one community at a time. For all the talk you hear about scale, that’s really not the way to go unless you’re talking about obvious things like finding a common tech platform so that every local news publisher doesn’t have to reinvent the wheel — or, in this case, the content management system. In the early days of the hyperlocal news movement, a group of publishers got together and formed an organization called Authentically Local. Its spot-on message: “Local Doesn’t Scale.”

Authentically Local is no longer around, but other organizations — especially LION (Local Independent Online News) Publishers — have risen up to take its place. LION current lists 445 members, many of them tiny hyperlocals comprising one, two or three people. And yes, LION provides its members with ideas for common solutions, but fundamentally its members are independent,  locally based entrepreneurs, for-profit and nonprofit, who are engaged in the hard but rewarding job of bringing news and information to their communities.

What brings all of this to mind is the growing perception that AI can help solve the local news gap. To some extent, it can if used responsibly and ethically. As I wrote recently, I think Chris Quinn is taking it too far at Cleveland.com and The Plain Dealer. By turning over the task of actually writing stories to AI, he’s eliminating the vitally important work of having the journalists who reported those stories sift through the nuances and emphasis points that readers need. But at least he says he’s enforcing some ethical guidelines, the most important of which is human review before publication.

I wonder, though, if that’s going to be the case at Patch. Liz Skalka reports for the Columbia Journalism Review that the nationwide network of local news sites is embracing AI in a big way, using AI to produce newsletters that “rely heavily on aggregation, automated event calendars, and posts from Nextdoor.” She quotes Patch chief executive Warren St. John as saying “This is a utility. This is not the high church of journalism. This is about creating a foothold in a relationship and meeting a need.” I find it a little difficult to see exactly what need is being met. I mean, Nextdoor posts?

Patch has been through several iterations and ownership changes over the years. At one time, maybe 12 to 15 years ago, Patch had one actual journalist in each community it covered, and they would go out and cover stories; I recall encountering a Patch editor who was working out of a Borders bookstore. I still see some of that in our local Patch, although each site now covers multiple communities. And in New Hampshire, Tony Schinella continues to provide award-winning coverage for Patch. But I have to wonder how long that will continue with Patch’s new AI overlords taking charge. Maybe they’ll do both. I hope so.

A happier story is told by Alexandra Bruell in The Wall Street Journal. Bruell reports on news organizations that are using AI to extend the reach of their actual journalists. Among them: The Philadelphia Inquirer, a for-profit regional daily owned by the nonprofit Lenfest Institute. She writes:

Reporters are using artificial-intelligence tools to scan community meetings for topics that may prompt news, such as a zoning issue related to an ICE detention facility and a proposal for a new data center. The effort is partly funded by a partnership between OpenAI and Microsoft and the nonprofit Lenfest Institute, which owns the Inquirer.

Using AI to ferret out stories that reporters can then follow up on is exactly the sort of task that is best handled by robots. Bruell adds that AI “promises a way to monitor police scanners and town meetings — the time-consuming bread and butter of much local journalism — more efficiently, and even opens the door to expanding coverage.” Indeed, that’s exactly what’s happening at the Midcoast Villager in Camden, Maine, where the paper’s small staff is using AI to summarize governmental meetings in the 43 towns it covers to see what might be worth the intervention of a human journalist.

The need for local solutions to local problems is greater than ever because of another AI phenomenon: its incorporation into Google search. Google now produces AI-generated summaries in response to queries, and even though those results include links, most users don’t bother to click. Never mind that the answers may be flat-out wrong.

Anna Nicolaou of the Financial Times reported last week that magazine publisher Condé Nast has been hit with such a drop in incoming traffic that Google will soon be inconsequential to its magazines, which include titles such as The New Yorker and Vogue. How bad is it? Over the past year, chief executive Roger Lynch said, Google has declined from providing a majority of incoming traffic to just 25%. Yet Condé Nast’s revenues actually grew in 2025, Lynch said, mainly because of the company’s success in selling digital subscriptions.

Which brings me back to where I started. Condé Nast may not be engaged in the local news business, but it’s succeeding by building direct relationship with its audiences. At the local level, publishers need to move beyond tech platforms, which were never their friends, and serve their communities. Every farmer’s market is an opportunity to sign people up for their newsletters. For-profits need to get readers to subscribe — and give them a reason to do so. Nonprofits need to drive voluntary membership fees. Events bring people together and raise visibility.

Publishers can learn from each other and develop best practices, but ultimately they need to apply these lessons in a way that makes sense for their own individual projects.

Local doesn’t scale.

Massachusetts has long been notorious for being one of the least progressive states with regard to government transparency. The state’s public records law is alone in exempting the governor’s office, the Legislature and the judiciary, leaving cities, towns, counties and the state’s executive agencies as the only government bodies that may be compelled to produce documents when requested to do so by journalists or members of the public.

Follow my Bluesky newsfeed for additional news and commentary. And please join my Patreon for just $6 a month. You’ll receive a supporters-only newsletter every Thursday.

What’s worse, there are few penalties for failing to comply with the law. As John Hilliard observes (sub. req.) in The Boston Globe:

In Massachusetts, the state law’s deadlines for fulfilling records requests can be ignored, workers can conspire to overestimate costs, elected officials can spend years fighting requests in court, or not bother releasing records at all. No one tracks whether local governments like cities and school districts follow the law; state agencies self-report requests, but not the reasons why they refuse them.

Michael Morisy, the chief executive of Boston-based MuckRock, who’s been helping people file public records requests for years, told the Globe: “It’s among the worst states when it comes to public records access.”

That could be about to change, although a lot of pieces will need to fall into place. State Auditor Diana DiZoglio is pushing a question that may appear on the ballot this fall that would end the exemption for the governor’s office and the Legislature. Granted, there’s nothing in the text of the measure that would strengthen enforcement of the existing law — and a previous DiZoglio-backed law to allow her to audit the Legislature is bogged down in legal challenges even though it was overwhelmingly approved by voters in 2024.

Still, subjecting the governor and the Legislature to the public records law would be a step in the right direction. Chris Lisinski of CommonWealth Beacon reported earlier this week on a hearing before a special legislative committee that is considering a range of proposed ballot initiatives. Some legislators remain hostile to the idea, Lisinski noted, which shows that they are not going to embrace greater transparency without a clear message from voters. State Sen. Cindy Friedman, D-Arlington, co-chair of the committee, cast her opposition as an affront to her ability to speak candidly with other legislators:

The Legislature is a deliberative body. It is our job to think of things, to have many discussions, to talk over what our differences are. If every one of those is public, if everything we do is public, we will not be able to do our work because someone will take something out of that and turn it into a completely different message, and that’s what we will end up responding to.

She continued:

I need to be able to talk to my folks, my other senators, and be able to have a real and honest conversation. We are not hiding anything. We are not trying to go and do nefarious, nasty, bad things. We are not spending huge amounts of money on yachts, right? None of that.

DiZoglio, who was also there, had this to say in testimony before the committee: “Knowledge is power, and when we as elected officials believe that knowledge belongs only to a select few and not to everybody, what does that say? It says we believe that we should just have all the power.”

The Globe story made reference to a survey the paper conducted in 2025, when it asked all 200 members (sub. req.) of the Legislature whether they would support extending the public records law to themselves. “The vast majority of lawmakers — 78 percent — did not respond,” the Globe reported. “And many of those who did respond did not offer a clear answer on their positions on the issue.”

At Northeastern, we found the same in 2020. That’s when we assigned our intermediate reporting students to contact not just incumbents but all 257 candidates for legislative seats. We reported our results in The Scope, a digital publication that’s part of our School of Journalism. The findings: Only 72% of those contacted responded to our survey despite repeated emails and phone calls directed at each candidate.

Of those who did respond, “about 72% of respondents said they favored expanding the public records law to cover the Legislature. Only 7% were opposed, while another 21% provided responses that did not fit neatly into either ‘yes’ or ‘no.’” As we noted at the time:

Despite the state’s progressive image, Massachusetts has a dubious record when it comes to governmental openness. In 2015, the Center for Public Integrity awarded Massachusetts an “F” for failing to provide public access to information. A reform law approved in 2016 did not address the legislative exemption, which also extends to the judiciary and the governor’s immediate staff.

The story in The Scope led with comments by Erika Uyterhoeven, a Somerville Democrat who was running for a House seat and who supported ending the exemption. “Democracy is not just about voting,” she told our students. “It’s about ongoing political engagement and understanding of the issues, understanding of the debate.”

Uyterhoeven won, and now she’s running for the state Senate seat being vacated by retiring Democratic Sen. Pat Jehlen. Uyterhoeven has been at least partly successful in bringing more openness to the Legislature, pushing successfully for a measure to make committee votes public. But she’s just one of 200.

It seems likely that the voters will have the ultimate decision on whether to end the public records exemption, although it’s possible that state Attorney General Andrea Campbell will rule that it’s unconstitutional, as Central Connecticut University Professor Jerold Duquette argued in testimony before the special committee. Duquette said such a law would infringe upon the separation of powers, Lisinski wrote for CommonWealth, since it “could force the House and Senate’s ‘involuntary cooperation’ with the secretary of state, the elected official who enforces the law.”

With an entrenched Democratic majority that appears to be resistant to change, a moribund Republican minority, and the very real possibility that an affirmative vote by the public would be challenged in court, it appears that we remain a long way from guaranteeing that the Legislature and the governor conduct the public’s business in public. I hope I’m wrong.

It’s been a month since I launched whenwords, and since then there’s been a flurry of experiments with spec driven development (SDD): using coding agents to implement software using only a detailed text spec and a collection of conformance tests.

Github Could Use a ‘Docs Review’ UI

First off, despite whenwords being a couple Markdown docs and a YAML test set, people have submitted valuable PRs. Mathias Lafeldt spotted a disagreement about rounding, where the spec instructed the agent to round up in several scenarios, but three tests were rounding down. Others have suggested there should be some [CI][ci] (despite their being no code) and wonder what that should be.

There’s been enough action on the repo to give us an idea of what open source collaboration could look like in a SDD world. And it feels more like commenting in and marking up a Google Doc than code merges. I would love to see Github lean into this and build richer Markdown review, like Word or Google Docs, allowing for easier collaboration and accessibility to a wider audience.

Emulation & Porting are the Low-Hanging SDD Use Case

By far, the hardest part of starting a SDD project is creating the tests. Which is why many developers are opting for borrowing existing test sets or deriving by referencing a source of truth.

Here’s a few examples:

• Anthropic wrote a C compiler in Rust. They used existing test suites and used GCC as a source of truth for validation and generating new tests.
• Vercel created a bash emulator in TypeScript. They created and curated an amazing set of shell script spec tests and have been feeding these to Ralph. (To make this even more meta, I’ve been following their commits and Clauding them into Python).
• Pydantic created a Python emulator…in Python. This sounds silly, but it’s useful in the same way Vercel’s just-bash is: it’s a super lightweight sandbox for AI agents. (In fact, I’ve already wrapped it in a CodeInterpretter for use with DSPy’s RLM module)

Now… It’s worth noting that most of these examples didn’t emerge perfectly. Anthropic’s C-compiler just kinda punted on the hard stuff and admits the generated code is inefficient¹. Pydantic’s Python emulator lacks json, typing, sys, and other standard libraries. Though I’m sure those will come soon. Vercel’s just-bash sports outstanding coverage, though people continue to find bugs.

This is the big takeaway from watching the last few weeks of SDD: agents and a pile of tests can get you really far, really fast, but for complex software they can’t get you over the line. Edge cases will generate new tests, truly hard problems will resist SDD implementation, and architectural issues will prohibit parallelism agents.

Vercel’s CTO and just-bash creator, Malte Ubl, sums it up best:

You can Ralph up a port or emulator in a weekend or two, but now you have to take care of it.

We coined a new term on the Oxide and Friends podcast last month (primary credit to Adam Leventhal) covering the sense of psychological ennui leading into existential dread that many software developers are feeling thanks to the encroachment of generative AI into their field of work.

We're calling it Deep Blue.

You can listen to it being coined in real time from 47:15 in the episode. I've included a transcript below.

Deep Blue is a very real issue.

Becoming a professional software engineer is hard. Getting good enough for people to pay you money to write software takes years of dedicated work. The rewards are significant: this is a well compensated career which opens up a lot of great opportunities.

It's also a career that's mostly free from gatekeepers and expensive prerequisites. You don't need an expensive degree or accreditation. A laptop, an internet connection and a lot of time and curiosity is enough to get you started.

And it rewards the nerds! Spending your teenage years tinkering with computers turned out to be a very smart investment in your future.

The idea that this could all be stripped away by a chatbot is deeply upsetting.

I've seen signs of Deep Blue in most of the online communities I spend time in. I've even faced accusations from my peers that I am actively harming their future careers through my work helping people understand how well AI-assisted programming can work.

I think this is an issue which is causing genuine mental anguish for a lot of people in our community. Giving it a name makes it easier for us to have conversations about it.

My experiences of Deep Blue

I distinctly remember my first experience of Deep Blue. For me it was triggered by ChatGPT Code Interpreter back in early 2023.

My primary project is Datasette, an ecosystem of open source tools for telling stories with data. I had dedicated myself to the challenge of helping people (initially focusing on journalists) clean up, analyze and find meaning in data, in all sorts of shapes and sizes.

I expected I would need to build a lot of software for this! It felt like a challenge that could keep me happily engaged for many years to come.

Then I tried uploading a CSV file of San Francisco Police Department Incident Reports - hundreds of thousands of rows - to ChatGPT Code Interpreter and... it did every piece of data cleanup and analysis I had on my napkin roadmap for the next few years with a couple of prompts.

It even converted the data into a neatly normalized SQLite database and let me download the result!

I remember having two competing thoughts in parallel.

On the one hand, as somebody who wants journalists to be able to do more with data, this felt like a huge breakthrough. Imagine giving every journalist in the world an on-demand analyst who could help them tackle any data question they could think of!

But on the other hand... what was I even for? My confidence in the value of my own projects took a painful hit. Was the path I'd chosen for myself suddenly a dead end?

I've had some further pangs of Deep Blue just in the past few weeks, thanks to the Claude Opus 4.5/4.6 and GPT-5.2/5.3 coding agent effect. As many other people are also observing, the latest generation of coding agents, given the right prompts, really can churn away for a few minutes to several hours and produce working, documented and fully tested software that exactly matches the criteria they were given.

"The code they write isn't any good" doesn't really cut it any more.

A lightly edited transcript

Bryan: I think that we're going to see a real problem with AI induced ennui where software engineers in particular get listless because the AI can do anything. Simon, what do you think about that?

Simon: Definitely. Anyone who's paying close attention to coding agents is feeling some of that already. There's an extent where you sort of get over it when you realize that you're still useful, even though your ability to memorize the syntax of program languages is completely irrelevant now.

Something I see a lot of is people out there who are having existential crises and are very, very unhappy because they're like, "I dedicated my career to learning this thing and now it just does it. What am I even for?". I will very happily try and convince those people that they are for a whole bunch of things and that none of that experience they've accumulated has gone to waste, but psychologically it's a difficult time for software engineers.

[...]

Bryan: Okay, so I'm going to predict that we name that. Whatever that is, we have a name for that kind of feeling and that kind of, whether you want to call it a blueness or a loss of purpose, and that we're kind of trying to address it collectively in a directed way.

Adam: Okay, this is your big moment. Pick the name. If you call your shot from here, this is you pointing to the stands. You know, I – Like deep blue, you know.

Bryan: Yeah, deep blue. I like that. I like deep blue. Deep blue. Oh, did you walk me into that, you bastard? You just blew out the candles on my birthday cake.

It wasn't my big moment at all. That was your big moment. No, that is, Adam, that is very good. That is deep blue.

Simon: All of the chess players and the Go players went through this a decade ago and they have come out stronger.

Turns out it was more than a decade ago: Deep Blue defeated Garry Kasparov in 1997.

You are only seeing the long-form articles from my blog. Subscribe to /atom/everything/ to get all of my posts, or take a look at my other subscription options.

How Generative and Agentic AI Shift Concern from Technical Debt to Cognitive Debt

Cognitive debt, a term gaining traction recently, instead communicates the notion that the debt compounded from going fast lives in the brains of the developers and affects their lived experiences and abilities to “go fast” or to make changes. Even if AI agents produce code that could be easy to understand, the humans involved may have simply lost the plot and may not understand what the program is supposed to do, how their intentions were implemented, or how to possibly change it.

Margaret-Anne expands on this further with an anecdote about a student team she coached:

But by weeks 7 or 8, one team hit a wall. They could no longer make even simple changes without breaking something unexpected. When I met with them, the team initially blamed technical debt: messy code, poor architecture, hurried implementations. But as we dug deeper, the real problem emerged: no one on the team could explain why certain design decisions had been made or how different parts of the system were supposed to work together. The code might have been messy, but the bigger issue was that the theory of the system, their shared understanding, had fragmented or disappeared entirely. They had accumulated cognitive debt faster than technical debt, and it paralyzed them.

I've experienced this myself on some of my more ambitious vibe-code-adjacent projects. I've been experimenting with prompting entire new features into existence without reviewing their implementations and, while it works surprisingly well, I've found myself getting lost in my own projects.

I no longer have a firm mental model of what they can do and how they work, which means each additional feature becomes harder to reason about, eventually leading me to lose the ability to make confident decisions about where to go next.

Via Martin Fowler

Tags: definitions, ai, generative-ai, llms, ai-assisted-programming, vibe-coding, cognitive-debt