The Answer Isn't Macro

There is a serious debate underway about whether AI will destroy jobs.

It is not a cartoon argument. Serious people are making evidence-based claims on both sides. One camp points to comparative advantage and elastic demand.

The other points to collapsing entry-level hiring and the disappearance of the bottom rung of the career ladder.

Both are observing real phenomena.

Neither is exactly wrong.

But something is missing from both sides.

They are reasoning about society from the outside. They look at labor markets the way a satellite looks at terrain , useful for spotting large features, blind to what determines whether a specific place actually works.

I live in a four-county region in upstate New York: Schoharie, Delaware, Ulster, and Greene. Rural. Aging. The kind of place that appears in national data as a footnote, if it appears at all.

I have been here for twenty years.

The economists debating AI and jobs are asking the right question in the wrong register.

Their question is: Will employment exist?

The question that determines whether ordinary people are actually okay is different: What holds a community together when the systems that were supposed to serve it stop working?

Those are not the same question.

You can have employment and still have a community that is dissolving.

You can have productivity gains and still notice that the hardware store is fielding different questions, the school sees families leaving without explanation, the diner is quieter at lunch.

These signals do not appear in dashboards.

They show up as disruptions in a familiar rhythm, detectable only by people who have paid attention long enough to know what normal feels like.

Large systems are not designed to notice disturbances like that.

They excel at tracking what has already been defined, once it crosses a threshold large enough to register.

By the time something becomes visible at a national scale, it has often been shaping local behavior for years.

This is not a failure of attention. It is a limitation of distance.

A satellite can see a forest fire, but it cannot smell the smoke in a neighbor's kitchen.

The macro AI debate also assumes that the relevant response must be macro: policy, redistribution, institutional design at scale.

Maybe. Eventually.

But policy has to be built before it is needed, and the people capable of building it are often the same ones being outpaced by the forces that make it necessary.

That is not pessimism.

It is a pattern.

What I have been thinking about instead is smaller and more specific.

What if the answer, or part of it, is not macro at all?

What if the right unit of response is the community, not as a philosophical preference but as a structural fact?

The Mountain Eagle is a regional newspaper serving four counties.

Its owner acquired it in 2016 and has spent nearly a decade absorbing failing weeklies, growing it into six regional brands.

The staff has many elders. They know their communities the way certain knowledge can only be known , through accumulated relationship, consequence, and time.

When I talk about what we have built together, I do not mean a better content management system.

I mean infrastructure.

That distinction matters.

A media company asks: What content should we create to capture attention?

Infrastructure asks: What conditions must exist so that real life can coordinate?

Those are not the same question.

At its best, the local newspaper was never primarily a media company.

It was a coordination layer.

It created a shared timeline.

It recorded what happened so memory did not evaporate.

It reduced rumor by making events legible.

It gave local businesses a trusted place to speak and residents a place to return to , not just to consume, but to orient.

When that function erodes, communities do not just lose information.

They lose orientation.

Everything feels new, even when it is not. Issues repeat without recognition.

Trust thins, not necessarily because people disagree, but because there is no longer a shared place to remember from.

What we are building is an attempt to restore that function and extend it.

The platform underneath the Mountain Eagle handles advertising operations, subscriber management, events, community calendars, local commerce, delivery routing, and back-office workflows.

It is designed around a simple principle: move load from humans to infrastructure.

For a while, improvisation looks like competence.

An email thread substitutes for a system.

A spreadsheet becomes a ledger.

Institutional memory lives in one person's head.

Work moves.

Nothing fails loudly.

Everyone feels a little tired, but quietly proud they are making it work.

Improvisation, however, has a shelf life.

The real cost of improvisation is not inefficiency.

It is moral.

It asks humans to do what systems are supposed to do , and then blames them when cracks inevitably appear.

It externalizes risk onto the most conscientious people in the room: the ones who notice when something is off, who stay late, who feel responsible even when responsibility was never formally assigned.

Our journalists are comfortable writing in Word.

They save to Google Drive.

That is what they know, and that is what they should keep doing.

A local AI agent scans their drafts, extracts structured content, builds web-ready articles, and places events on the community calendar.

A manager reviews and publishes.

The staff workflow does not change.

Their knowledge stays where it belongs, with them.

The technology serves the people, not the other way around.

The AI in this system is not general. It is local.

It knows the difference between Cobleskill and Catskill.

It understands the publication's voice and structure.

It handles the mechanical translation between analog workflow and digital platform , precisely the kind of repeatable work that frees human attention for judgment calls that require actual knowledge of actual people in actual places.

This is what the macro debate misses when it talks about AI and jobs.

The question is not whether AI will replace local journalists.

It will not replace a seventy-year-old reporter who has covered a county for three decades.

It cannot replicate the knowledge of which town supervisor actually makes decisions, which business is quietly struggling, which family history sits beneath a zoning dispute.

That knowledge does not exist in a dataset.

It lives in people, accumulated through proximity and consequence over time.

What AI can replace is the friction between that knowledge and the community it serves.

The infrastructure we have built is designed to be owned, not rented.

Two locally owned telecommunications companies provide the bandwidth.

The hardware is local. The data stays local. Subscriber relationships, advertiser relationships, institutional archives , none of it lives on a platform controlled by someone with no stake in the region.

This is not ideological. It is structural.

When infrastructure belongs to the place it serves, accountability becomes tangible. Failure is visible. Success is shared. Extraction becomes harder to disguise.

The model we are building toward is cooperative: publisher entities collectively owning the infrastructure, governed by member institutions rather than distant capital.

New publishers can join.

Publishers that stop serving their communities can lose their place in the network.

The cooperative owns the hardware.

Local publishers own their relationships.

Communities retain stewardship over their data.

This is not unprecedented.

Rural electric cooperatives operated on similar principles a century ago, serving communities that investor-owned utilities had no incentive to reach.

The structure works when it is composed of institutions rather than individuals, and when governance answers to consequence rather than quarterly performance.

If the template works here, it is replicable , not as a product, but as a pattern.

None of this happened quickly.

It took twenty years of living in a place.

Building relationships.

Watching what worked and what failed. Waiting for the tools to mature.

Finding the right collaborators.

Building in the right sequence.

Technology was almost the last piece to fall into place.

The macro debate about AI and jobs is not wrong to worry.

Entry-level displacement is real.

The concentration of surplus in the hands of infrastructure owners is a genuine risk.

But the answer to those concerns is not only macro.

It is also this: communities that own their infrastructure, support their local institutions, and invest in the unglamorous work of durability and accountability are building something that cannot easily be disrupted from the outside.

Not because they are insulated from change.

But because they belong to the place they serve.

And that , unlike scale , cannot be centralized.