AI Future Fridays: Neighborhood Tech and the Network You Own Together

Welcome to AI Future Fridays. Every week this column takes one force shaping the near future, tells you the truth about it, and hands you something to build before Monday. This week: neighborhood technology, the quiet comeback of infrastructure that ordinary people own and operate themselves.

Here's what's different about this moment. Community networks have existed for twenty years, and for nineteen of them the blocker wasn't the radios. It was the human overhead: who tracks the node that stopped answering, who notices the boil-water notice, who writes the message nobody has time to write. That overhead is exactly the kind of work agents are good at now, which means the volunteer who used to quit after eight months might stay.

You're not a bystander here. You're a rooftop.

The Neighborhood Mesh Network Signal From a Rooftop in Brooklyn

The clearest signal isn't a product launch. It's a volunteer network that kept going long enough to get boring, and then got funded.

NYC Mesh started in 2014 with a website and a handful of antennas borrowed from the Catalan network Guifi.net. It now runs over 2,000 active member nodes across the five boroughs, routing through a supernode at 375 Pearl Street in Lower Manhattan, where a volunteer-run nonprofit peers with Apple and Google like any other carrier. In July 2026 the group announced an $898,035 ConnectALL award from New York State, earmarked partly for its first full-time staff member after twelve years of nobody being paid.

Twelve years of volunteers, and the state eventually showed up with a check. That's the shape of a thing that worked.

Then there's the other half of the signal, which costs almost nothing. IEEE Spectrum has walked readers through the LoRa mesh gear that lets a handful of battery-powered radios pass text across a neighborhood with no towers, no fiber, no license and no grid. One of those costs less than a month of broadband.

The Backstory of Community-Run Internet and Ham Radio

Every future has a lineage, and this one's is a century old. People have always built their own communications when the official version was too expensive, too slow, or absent.

Amateur radio operators have relayed storm reports since the 1930s. The free-network movement of the 2000s produced Freifunk in Germany, Ninux in Italy, Guifi.net in Catalonia, and Red Hook Wi-Fi in Brooklyn, which famously carried traffic after Hurricane Sandy drowned the neighborhood's infrastructure. The Internet Society documented how these groups organize: flat, licensed under a commons agreement, with a rule that anyone who joins must extend the network to the next person on the same terms.

The LoRa chapter is newer and has a name attached to it. Kevin Hester built the first version of Meshtastic in early 2020 because he wanted to stay in touch with friends while out of cell range, and he open-sourced it. Hundreds of contributors later, it's the default answer when somebody asks how to text across a valley with the power out. Volunteers put nodes on rooftops in North Carolina and Tennessee after Hurricane Helene took out tower after tower, and encrypted messages hopped across ridgelines that had defeated the fuel trucks.

Here's the myth worth killing: that community networks are a protest against telecoms. Mostly they're not. They're a hobby that turns into infrastructure, built by people who like knowing how things work. The timeline rhymes:

• 1935 to 1999: amateur operators relay emergency traffic when commercial systems fail, establishing the social norm that neighbors handle their own comms.
• 2004 to 2014: European free-network projects prove rooftop Wi-Fi meshes can scale to thousands of households under a commons license.
• 2017: after the FCC repeals its net neutrality rules, NYC Mesh signup requests jump from roughly 20 a month to more than 400.
• 2020 to 2026: LoRa drops the cost of a working off-grid node to pocket money, and agents arrive to carry the coordination that burned out every previous generation of volunteers.

What a Neighborhood Can Run on Its Own Hardware

Two very different networks get called "mesh," and confusing them is how people end up disappointed. One moves real bandwidth. The other moves sentences. You want both, for different days. Worth deciding which one you're building before you buy anything, the same way you'd write the spec before the workflow.

That third row is where community tech usually dies, and it's the least discussed. Researchers studying mutual aid groups published in the ACM Digital Library found that groups end up running on whatever free tools are lying around, and those tools quietly concentrate control in whoever happens to hold the login. The network is democratic. The spreadsheet is a monarchy.

The Countermove: Build a Neighborhood Alert Digest

The street finds its own uses for things. If the hard part of neighborhood tech is the attention it demands, then the countermove is to automate the attention and leave the judgment to people.

Do the boring, load-bearing things first. None of them need a model:

• Find your local mesh before you build one: most US metros already have a LoRa group and a community broadband project, and joining beats founding.
• Put one node high and leave it alone: elevation beats everything else in radio, and one always-on rooftop node makes every pocket node in range useful.
• Write down who knows what: a single shared page of who has roof access, who has a generator, who checks on whom. Paper counts.
• Pick a channel everyone can reach: one group thread people already read, not a fourth app. If that's Discord, fine, and automating the Discord side is straightforward.

Then build the digest. The agent watches public, boring, non-human sources on a schedule: your utility's outage map, the water authority's advisories, the city's road closure and permit feeds, an air quality endpoint, the storm track when there is one, and the status of your own mesh nodes. It compares against yesterday, drops anything unchanged, and sends one short message to the group thread each morning. Two lines when nothing's happening. Six when something is.

That last part is the whole trick, and it's the same discipline behind a standup summary that people read: a digest that cries wolf gets muted in a week, and a muted digest is worse than none. Suppress, don't summarize. The reason this changes the power balance is unglamorous: the neighbor who was doing this by hand gets their mornings back, so they're still around in a year.

One rule, and it isn't optional. Point it at conditions, never at people. No cameras, no plate readers, no tracking who came and went, no pipeline to anybody's enforcement database. A block that watches its infrastructure gets safer. A block that watches its residents has built a cage and called it a newsletter. If you want to see how we draw that line elsewhere, the agent workflow directory is full of monitors pointed at systems rather than at humans.

The Forecast for Community Mesh Networks Through 2027

The honest read for the next 18 months: the hardware curve keeps falling, the coordination problem mostly gets solved, and the social problem stays hard. Two of those three are going your way.

The social number is the uncomfortable one. Pew Research Center found that 26% of US adults say they know all or most of their neighbors, down from 31% in 2018, and the expectation gap is wider still: 76% say they'd bring in a neighbor's mail while they were away, but only 52% believe a neighbor would do it for them. We're more willing than we think each other is. Community networks are built out of exactly that willingness, and the first job of any neighborhood tool is to make it visible.

Now the limits, plainly, because a forecast that only flatters you is useless. LoRa mesh moves a trickle of data, needs roughly clear line of sight, and congests when crowds pile on. The Meshtastic project had to ship special event firmware to keep 2,000-plus nodes from drowning each other at conferences like DEF CON, and Wired has been blunt that none of this replaces 911, a satellite messenger, or a working phone. Typical range is a couple of kilometers, not the 331-kilometer community record you'll see quoted. Community broadband has a harder ceiling: it needs buildings, landlords, and volunteers who show up on install day.

The resilient neighborhood of 2027 won't be the one with the best gear. It'll be the one where four people know how the gear works.

The coordination side is the part that improves fastest, because it's ordinary recurring reporting work wearing a civic hat. Expect more municipalities to quietly fund mesh as backup emergency comms, more grant money following the NYC Mesh precedent, and a wave of off-the-shelf nodes aimed at people who have no interest in flashing firmware. Expect at least one vendor to sell a "neighborhood safety" product that's surveillance in a friendly font. You'll know it when you see it.

The Horizon for Neighborhoods That Own Their Network

Here's the future worth building, and the order of priorities matters. A street with its own radios is not more important than ending the wars grinding on right now, stopping the atrocities we can all see, or pulling the climate back from the line. It's a piece of the third one, and a small piece: a neighborhood that can coordinate through a heat dome or a flood is climate adaptation at the only scale most people can personally touch.

Picture the ordinary version. It's the second night of an outage. Nobody's phone has service. The thread on the mesh is dull in the best way: the pharmacy on the corner has power and will hold insulin in its cooler, the hydrant on the east side is flowing, somebody's charging a neighbor's oxygen concentrator off a car inverter, and the node on the tallest roof is at 61% battery and will make it to morning. No command center. No one watching anybody. Just a street that can still hear itself.

That's the whole dream, and notice what isn't in it. No cameras. No one's movements logged. Nothing a landlord or a police department could subpoena, because nothing about people was ever collected. The tools stay pointed at pipes and poles and weather, which is where tools belong. We've made the same argument about keeping the model on hardware you own, and it's the same principle one layer down: sovereignty isn't a feature you're granted, it's a thing you hold.

A hundred years ago a neighbor with a radio set was the most useful person on the block during a storm. Give that neighbor an agent to carry the paperwork, and they'll be that person again, for longer, without burning out.

Before Monday: Your Neighborhood Mesh Checklist

You've got a weekend. None of this takes all of it.

1. Search for your metro's name plus "mesh" and find out whether a LoRa group and a community broadband project already exist near you. Most cities have both.
2. Order one LoRa node, or borrow one. Two radios on one street is enough to learn whether elevation or trees win where you live.
3. Write the one-page sheet: who has roof access, who has a generator, who should get checked on first, and two phone numbers that don't depend on an app.
4. Pick the group thread people already read, and get permission to post a daily line in it. Skip this and the digest talks to nobody.
5. Build the digest with the prompt above, pointed only at outage, water, road and air quality sources. Run it for three mornings and delete whatever you skimmed past.
6. Walk it over to one neighbor in person and show them. The tech is the easy half.

Frequently Asked Questions

• What is a neighborhood mesh network?+
  It's a network your neighbors own and run instead of renting. Two flavors dominate in 2026. Community broadband groups like NYC Mesh put rooftop radios on volunteer buildings and share real internet access. LoRa mesh projects like Meshtastic use cheap low-power radios to pass short text messages with no towers, no fiber, no license and no grid power. The first gives you bandwidth. The second gives you a channel that survives when everything else fails.
• How much does it cost to start a neighborhood mesh network?+
  A single LoRa mesh node runs roughly the price of a nice dinner, and two of them are enough to prove the idea on your own street. Community broadband is a bigger lift because it needs roof access, a willing building and volunteer installers, so most people start by joining an existing group rather than founding one. The expensive part was never the hardware. It's the coordination, which is the part an agent workflow can carry for you.
• Is a mesh network a replacement for 911 or cell service?+
  No, and anyone who tells you otherwise is selling something. LoRa mesh carries short text and telemetry at a trickle, needs rough line of sight, and congests fast when a lot of people talk at once. Treat it as a second channel that keeps working when the first one dies, not as a substitute for emergency services or a working phone.
• Can a neighborhood alert digest turn into neighborhood surveillance?+
  It can, which is why the design matters more than the code. Point the agent at infrastructure and hazards: outages, water advisories, road closures, air quality, storm tracks, permit filings. Never at people. A digest that watches conditions makes a block safer. A digest that watches residents builds a cage with extra steps, and that's a line this column does not cross. The same restraint shows up in how we point monitoring tools at systems rather than at neighbors.