Saving Local News: Build the System First

The math of local news fits on a napkin. Reporters cost money. The money that paid them, print ads, classifieds, the subscription bundle, is gone, and what replaced it online pays a fraction as much and rewards volume. That leaves three options. Raise outside money and burn through it. Find a patron and pray the interest holds. Or accept that the old staffing model is dead and design around it.

I designed around it.

The rescue plan

Most plans to revive local news start in the same place: hire the reporters back. Find the money, rebuild the newsroom, restore what was lost. The instinct is generous. It's also a reenactment. That newsroom was built for an economy that no longer exists, and rebuilding it without that economy just stages the collapse a second time, with better intentions.

Etaoin shrdlu

Newspapers have been here before. For most of the 20th century, papers set their type on Linotype machines, with operators casting each line in hot lead. The keys were laid out by letter frequency, so the first two columns spelled "etaoin shrdlu." When an operator botched a line, the quick fix was to run a finger down those keys to fill it out and toss it later, and every so often the slug escaped into print. Readers got "etaoin shrdlu" in the middle of the news.

Then computer typesetting arrived, and over the next two decades the composing rooms emptied. The New York Times printed its last edition set in hot type in 1978. To get there, it had promised its printers jobs for life, and the last of them retired in 2016.

Nobody proposed saving newspapers by going back to hot lead. Everybody understood that the type was never the point. The reporting was.

That's what AI is to us. Most of the cost of local news was never the judgment about what to cover and why it matters. It was the hours it takes to turn that judgment into finished work: transcribing the meeting, reading the agenda packet, drafting the story, on deadline, every day. AI takes on a large share of those hours. The judgment stays where it was.

The platform underneath

So when I started designing Mercury Local, the question wasn't whether to use AI. It was what kind of system could produce steady local journalism at a cost that doesn't depend on somebody else's money.

The answer was to build the system first and the publications second. Mercury Local is one codebase, one database and one editorial architecture, running independent papers that each have their own name, their own priorities and their own readers. Most local news startups build one publication and spend years trying to make it pay for itself. Build the platform first, and a new town mostly needs local knowledge: who governs it, when they meet, what the people there care about. The pipeline, the standards and the software are already built.

That's how The Farmington Mercury covers a Connecticut town of fewer than 30,000 people with the kind of steady attention that used to take a dedicated newsroom, and how three papers run on one platform without a round of fundraising.

1978

The local news crisis usually gets told as a loss, something taken away that needs to be given back. To me it looks more like 1978. The old machinery is going. The work it served isn't.

Every town that lost its paper needs this kind of architecture. I didn't build it because AI is interesting. I built it because the math works.

Related coverage

• News Deserts Don't Need Saving. They Need Infrastructure.

• One Article Is Easy. An Operation Is the Work.