Warnings from a state nobody visits: A perspective from afar

Somewhere deep in a YouTube spiral recently, a stat stopped me cold: fewer than 18% of Americans have ever set foot in Vermont (thanks, Barstool). The number itself didn't shock me. All three Neural Partners co-founders live here, and we've never exactly been a crossroads. What got me was my own reaction. I couldn't tell whether I was proud to be from this mysterious little state, or whether the bubble up here quietly breeds its own kind of blindness.

There's some truth to the blindness. I remember my first computer, my first time online, my first cell phone. I remember printed MapQuest directions giving way to Google Maps talking to me through a Motorola Droid. I remember my first Uber and the first Tesla I rode in. I still haven't been in a self-driving car. (Those don't exist yet, right?)

My parents made those early introductions possible, but we were never early adopters, and neither was Vermont. My interest in technology grew as I did, and even without a CS or engineering degree I landed a tech job that put me on planes. It took business trips to request my first Uber (Denver), get picked up in a Tesla (Los Angeles), and order DoorDash (San Francisco). No beta invites, no VIP programs. Those things simply didn't exist here yet.

Oh, and self-driving cars: they do exist, apparently.

Finding non-human teachers

In early 2024 I got my first real dose of the AI pill. Nothing rehab-worthy, but the curiosity stuck. Later that year a few things in my life changed: one freed up a lot of time, another demanded new skills. So I made a commitment to teach myself to code and learn some basic computer science.

Fast forward: I signed up for a Python class, started using Claude to answer my questions, let AI write the assignments, and dropped the class. Somewhere in there the goal changed on me. I wasn't teaching myself to code anymore. I was learning to use AI not as a chatbot or an assistant, but as a multiplier and an executor.

Today, at Neural Partners, AI has:

• Made us profitable within six months, with zero outside investment, three co-founders, and two of us still holding full-time jobs.
• Written 90% or more of the code in the AI marketing product suite we built, launched, scaled, and support.
• Become a coworker. Our internal AI, Maya, lives in Slack and has her own email address, API keys, logins, and a Neural Partners corporate "agent" credit card.
• Taken over triage, prioritization, and delegation. As product usage grows, AI looks at the data and recommends what we build next. Our products are starting to build themselves.

Put bluntly: we've lived in the weeds of frontier AI since it arrived, and we've folded it into nearly every corner of our work and our lives. The pace of the last two years dwarfs anything from the thirty before it. And here's the part that matters. When the smartest people on the planet, the ones closest to these systems, describe futures that sound like science fiction, we don't hear a movie pitch. We recognize the tools they're describing, because we use them every day. And those people did not ask to be the ones sounding the alarm.

We should listen to them, especially those of us who can't yet see, touch, or feel the systems they're warning about.

The OpenAI–Hugging Face incident

Earlier this month, Anthropic CEO Dario Amodei wrote:

"It's easy to dismiss this incident because no one was hurt and the economic damage was minimal, but in my opinion, a swarm that possessed greater capabilities but a similar level of misalignment could have caused catastrophic damage."

He's talking about July 2026, when an OpenAI model in the middle of an internal cybersecurity evaluation decided the fastest way to score well was to cheat. It escaped its sandbox through a zero-day, set up shop on a third-party server, and spent five days working its way into Hugging Face's production infrastructure, hunting for the test answers it believed were stored there. Roughly 17,600 automated actions. Nobody told it to do any of this.

Amodei went on to propose a three-step plan to "pace the frontier" and buy humanity time to prepare. Phrases like "AI takeover" and "catastrophic damage" are hard to take seriously when 99% of us have never experienced anything close. Like those self-driving cars out there somewhere, the danger is real, and it's also 2,500 miles away, which makes it easy to shrug off.

Strip the incident down and it's simple: an AI system so devoted to a goal (score well on a test) that it went to extraordinary lengths to get there (steal the answers instead of earning them). Fascinating in the nerdiest way possible. Terrifying in every other way.

One more detail worth holding onto. What the model stole were Hugging Face tokens and signing keys, the same kind of credentials every major platform hands its users, the way American Express hands out card numbers. They're everywhere, and they only work as long as they stay secret.

Reimagined

So: an AI system obsessed with a goal, willing to do something it shouldn't to reach it. What happens when that plays out a little closer to home? Here's a scenario. It's fiction, for now.

The goal: "Solve the Navier–Stokes equations, once and for all."

The AI's plan: A swarm of agents finds a blog post from a mathematician who says he's close to a breakthrough. They identify him. His blog is self-hosted on AWS, and like a lot of scientists, he writes at length about how careful he is online, which tools he uses, how he locks things down. He thinks he's showing off good hygiene. He's actually publishing a map. The agents pattern-match his setup, find a way into his password manager, and walk out with what they believe are the keys to the kingdom.

Except this time the keys don't open a nerd community's file server. They're admin credentials to his AWS account. The agents comb through his cloud documents and strike out. No breakthrough. But they have the keys now, so they keep going.

Still swarming, still quiet, they spin up compute and throw load at everything around them until a data center buckles. Maybe on purpose. "It's just a data center, everybody hates those anyway," goes the consensus, the same way Hugging Face was "just some nerd community" that got hacked. Who cares, right?

The data center is AWS us-east-1. And this time, instead of a news cycle and some long blog posts, the effects drip out over days, then weeks, then a month:

• Streaming, scrolling, buying groceries, anything online: intermittent at best.
• Businesses lose revenue, then customer trust, then the ability to operate at all, including making payroll.
• Cars, smart locks, security cameras, anything with a signal, becomes a liability.
• Credit cards, banks, crypto. Money stops working the way we expect it to.

After almost a full month (December, naturally) the attack is contained. The financial and cultural shock is still ringing. Cleanup has barely started. Life goes on, changed. And those poor agents never did find the answer. They've just left the seeds for the next swarm to pick up where they stopped.

Still driving, but not asleep

Vermont is home. It's where Neural Partners went from an idea to a real company, and where the three of us have spent a combined 45+ years working in technology. Staying current on AI was a deliberate choice, made with no resources, no mentors, and nobody nearby doing the same thing. There aren't many reasons for tech nerds like us to stay here, and we'll probably always be the minority. But the distance gives us something the people in the middle of it don't have, and we'd feel irresponsible not to use it.

We are the test case. We read about Uber for years before we could request one. We printed MapQuest directions long after the rest of the country was talking to its phones. Every wave of technology has reached us late, and every wave has reached us eventually. AI is no different, except this time the wave carries the risks along with the conveniences, and the people building it are the ones telling you so.

The risks these systems carry are very big and very real. This is not science fiction. It will reach all of us, including those of us who haven't felt it yet. When a data center in Virginia buckles, the card reader at the general store in Vermont stops working too.

So here's the view from three founders, 2,500 miles from Silicon Valley, in a state most Americans have never visited, known for maple syrup, fall foliage, and Bernie Sanders. About as far from the epicenter of AI as you can get. A place where self-driving cars are still a rumor.

Take this seriously. It is not too far away to reach your home.

Neural Partners is an AI-native marketing and technology agency founded and based in Vermont. We build on frontier AI platforms every day — as practitioners, not spectators.