Over the last thirty years, digital tools have changed how news media operates, gradually moving down the list. The first internet wave built websites, apps, and targeted marketing tools. Then, the widespread arrival of database APIs, meeting transcripts, and then open government and commercial databases transformed information gathering. Today, maintaining websites, fixing apps, and editing reporting is easing the editing and debugging of content. As a result, the most bespoke aspects of news production, the relationship building necessary to obtain access to the best sources (how do you get someone’s phone number?) and the piecing together and analysis of multiple, complex sources (such as investigative reporting), still remain outside the reach of digital tools.
When considering the cost of a news publication you can break out the costs from the more automated to more bespoke:
- Presentation (print, website, email newsletter, app, etc)
- Marketing/Advertising
- Information gathering
- Editing/Debugging
- Development and maintenance of information access
- Information analysis
The more your news publication relies on higher order work (human-driven analysis vs publishing machine-collected information), the more costs the publication incurs. Thus, the more valuable the data needs to be.
However, my recent experience with operating EnergyThink.ing suggests to me that we’re hitting an inflection point where the availability of compute, paired with the availability of machine-readable data is making only highly-gated and charity-supported reporting remain viable.
The Barbell Economy of News
We live in a “barbell economy” for news. At one end, information value is low, but production costs per “unit” of news is low. On the other end, information value is high while unit production costs are also high. Either you have lots of customers attracted to either high volumes of low-cost, low-value information, or a few customers attracted bespoke, expensive reporting.
Along that barbell, there are three ways of making money in news: The first is to acquire hard to obtain information and sell it at a high price. For example, a “tip sheet” or newsletter with industry-specific reporting. The second, is to take large volumes of machine-collected information and organize it in a way that many people can use for their specific needs. Sports apps and weather apps are like this.
The final way is to use new technology by applying machines to gather and organize the information – something that was once premium – and turn it into a less expensive stream for larger amounts of people. If you push the price down somewhat, but still charge a higher price for a larger market, you make more money over a larger, more stable audience. Real estate listings is the classic example, where Realtors banded together to collect and gatekeep their data, and sell it at a high price, limiting access to the kinds of customers they want (Zillow, but not random app developers).
The rules of the news barbell economy are unavoidable. Publications that exist in the middle of the barbell, those that need to spend moderate sums to gather information that is not valuable to large audiences, such as city hall or community news, tend to struggle with profitability. This is why so many metro and community dailies have closed or transitioned to charity-supported non-profits over the last thirty years. It costs a lot to maintain good reporters in courts, city halls, and neighborhoods. And even in large cities like Chicago and Los Angeles, only a few tens of thousands of people actively read that reporting, and will only pay a couple dollars a day to read it – far from enough to cover operational expenses.
On one end of the barbell, sports stats cost very little to collect (leagues are already doing it as part of their games), stats have a never-ending shelf life, and you can display them to millions of people for pennies each viewing.
Then, on the opposite end of the barbell, an insider tip sheet on oil market data impacts the actions of only a few thousand people around the world. But those people crave highly detailed information that gives them an edge. They will pay exceedingly high amounts for exclusivity.
Technology-Driven Changes On The Barbell
When I joined my first media endeavor, Chicagoist, in 2004, a website revolution was underway with free and cheap blog software like MovableType and WordPress. The software allowed non-technical writers to input their text, and then see it formatted so that readers didn’t have to hunt for the latest story. An incredible innovation for “regular people” with no tech background, thousands of news blogs bloomed, covering niches that never existed. For instance, Chicagoist was a highly snarky, news-ish site that targeted young adult Chicagoans, a brand new genre for the time. Our readership exploded, and local TV even did stories on us. Yet, by today’s standards, Chicagoist was somewhere in the middle of the barbell.
By 2010, when I launched a neighborhood news site, CenterSquareJournal.com, the new revolution was in digital advertising. Although we sold our own ads, businesses were increasingly buying ads “programmatically” through sites that would target readers by demography. Although anyone with a website could display ads, the revenue from those ads decreased every month as ad servers found more and more readers to target across more websites. For digital advertising (few people sold subscriptions then) everything became about volume and niche websites lost revenue. By 2013, my neighborhood news site was dead – and it was a national industry news story, a harbinger for sites being pushed away from the middle of the barbell.
When I launched The Daily Line in 2015, we moved down the technology line again: easing information gathering, but also to towards the niche end of the barbell. I started the publication on the back of running a newsletter during that year’s mayoral election, where I asked people I knew around the city to send me scans of political mail their received. I organized the images, then reported on when political mail was hitting various parts of the city, what the campaign message was, and reported it in an email to everyone who subscribed.
My tools were dozens of Chicagoans’ personal phone cameras and scanners, Google docs I used for organizing the scans, and a cheap email service. My newsletter was such a unique resource, by election day I had over 30,000 free email subscribers. From that base I launched a paid digital news publication that made heavy use of city and county government data for our reporting. Although we had live reporters in City Hall, a great deal of our work involved looking things up in government databases, then asking politicians for comment. It was great niche content and the publication still operates today under a different owner.
Today, we’re looking at the next revolution pushing us further to the barbell edges: AI provided tools for building and debugging practically any web app a person can think of, and AI acting as a kind of “editor” that can suggest new lines of inquiry, one of the most important and unheralded roles of a good editor. The hobby site I launched in April, EnergyThink.ing, was almost entirely “vibecoded”. It relies heavily on government websites and APIs to obtain information, and uses AI to edit and surmise the information into custom email newsletters. Every user gets content serving their customized niche, as far to the edge of the barbell you can get.
The Barbell Gets Smashed
There are three important things to know about EnergyThink.ing. The first is that it is targeted at a very niche group: People who want to follow state-level energy policy, potentially in multiple states at the same time. The second, is that although I know something about setting up websites and maintaining servers, I am not a programmer and that AI wrote all the code and gave me all the instructions I needed to build the site, requiring little technical expertise on my part.
The final thing to know is that EnergyThink.ing utilizes machine-run databases with free, or very low-cost tools to run the site that needs no human intervention. All together, everything costs me less than $150 a month to run, whereas a typical newsletter costs at least a few thousand dollars a month to operate, since you need to pay a human to edit and send it (although some people don’t pay themselves to run their hobby sites).
EnergyThink.ing pulls together the two ends of the barbell. It is a niche news site, with a potentially high value for a small group of people, operated at an extremely low cost. A major innovation! I should be raking in the dough, right?
Almost as soon as I launched EnergyThink.ing, I began to wonder: If I could do this, who else could?
Indeed, there are at least three other sites I know of that do similar things: ClimateHerald.org, a free site run by Michael Dexter, gathers climate news and policies from cities to track local progress. And then Hdata and Halcyon both ingest state public utility commission and electricity grid dockets to push them through AI, allowing natural language queries. Hdata ran a $12.5 million investment round in April 2025 and Halcyon took in $21 million last March, according to Pitchbook. All of it is likely going towards becoming the best high-end energy regulatory data service by ingesting as much data as possible with the clearest AI search system.
The Increasingly Niche Niches of Niche News
Hdata and Halcyon’s moats against invaders like EnergyThink.ing, I’m guessing, is to pull in as much data as they can from messy state regulator websites, and then become the best organizers of that data. I imagine they hope that people in high-end energy planning will happily pay five figure subscription fees for the convenience of well-organized and easily accessible data.
But, as I’m learning from my work with EnergyThink.ing, there’s a few inherent limitations: State governments are cranky and change their website structures with no notice and the value of the data is limited to the quality of the questions you ask it. In other words, even if you have a really good AI to manage queries, if the user doesn’t understand what’s in the data or have a good idea of how to ask the right kinds of questions, they won’t learn very much.
The first problem is manageable by adding more developers to the project (increasing overhead). The second issue is only manageable if there is enough incentive for users to want to learn – that the answers translate into user profits.
And so that raises a critical question about niche data and reporting: Is there enough value there?
Halcyon is trying to educate users by including (the very talented) climate reporter Nat Bullard on their team, who has helped generate some interesting questions answered by their data, like “Track planned U.S. substations and switching stations”. A good question for generation developers! But is it worthwhile to enough people to be a profitable subscription service for Halcyon? That’s also a good question!
The power of data collection, along with rapidly decreasing development costs is making it possible to go down narrower and narrower niche corridors, chasing higher fees. But if the data collection is getting easier, and development costs are too – if a programming rube like me is able to make EnergyThink.ing – what does the future hold for niche news?
What If Your Moat Dissolves?
Also, what happens if a bunch of state PUCs turn their state dockets into APIs? This exact thing happened to state legislatures about fifteen years ago. It used to be if you wanted to track state legislative data, you needed to pay thousands of dollars a month for bill tracking services, now you can get those same services for tens of dollars a month or even free.
It is possible to query any AI service: What battery legislation is pending in Illinois? and get a very good answer. When will PUC dockets be ingested into the major AI systems? For anyone creating an information service, you’re essentially in a race with the major AI – that you can only lose.
This seems to be the future for almost every public database anywhere. There’s an exception for private data collections: If you’re part of a trade group or cartel that collects specific data that lives nowhere other than on your servers, like real estate listings owned by Realtors.
Building on data or reporting you don’t control is a recipe for disaster – and public data is data you don’t control. If you are collecting and creating your own data and news, there could be enough value to cover the cost of operation, whether that be news collected by a reporter on the street, or a trade association collecting data from members.
In the end, we’re moving to a world where either data and news is gated, thus made artificially valuable, or paid for by charitable organizations. Yet, cheap AI compute is putting more and more pressure on both of those solutions.
Other Things I’ve been Thinking About
The Great Shift From Workers to Owners – Some economists think a K-shaped economy has become embedded in US’ operating system.
A new media ecosystem is taking root on the left, reshaping Democratic politics – The right watches Fox, the left reads a broad constellation of websites.
The Richest Country Is Pretty Mid Now – A pretty good video narrative of how big investors’ obsession with leverage has crushed small/mid-sized businesses and we’re all paying for it.
Degeneracy is a Symptom – A linkage of China’s laying flat, the US’s deaths of despair and the possibility that the whole world is experiencing a global market/government failure.
Subaru Socialists and the Great Disappointed – If you’re under 45 in the US, you’re probably feeling cheated and ready for political change.
36 Hours In Black Chicago – I’m from Hyde Park and have been to half these places. It’s hard to sell white out-of-towners on this stuff. Take Natalie Moore’s advice and go.
Your gas car works fine. Consider an EV anyway, scientists say. – As someone who made the leap, once you go, you’ll never want to go back.
‘Oh Fuck,’ Announces Belching, Groaning Bus Passenger Directly Behind Woman – The Onion provides the most accurate portrayal of a terrible city bus ride most everyone in a big city has experienced at least once.
Cinnamon Rolls – Made this last week. Gonna make it again. Add pecans. It’s unreal.