The Forge — Armory Forge Systems — Signet Article #041

If you read the tech news this month, you'd be forgiven for thinking the end is nigh.

On September 8th, a 27-year-old researcher named Jacob Coxon quit Anthropic after three years of training AI models — first at OpenAI, then at Anthropic. His resignation post said the people building this technology "earnestly believe it could kill us all by the end of the decade," and that the companies are "racing straight to self-improving superintelligence and gambling with our lives." It has been viewed more than 100 million times.

Around the same time, Anthropic's alignment lead, Evan Hubinger, put a number on it in public: more than a 10% chance that AI kills all humans within the next decade.

On September 6th, OpenAI's own chief scientist, Jakub Pachocki, published an essay titled "An Alien Mind." His conclusion: no lab, including his own, has solved alignment and monitoring well enough to keep scaling at maximum speed. He called this "a time that calls for extreme caution," and said he expects and hopes voluntary slowdowns become common.

By the weekend, the New York Times was reporting on doomsday conversations inside AI companies, Al Jazeera was covering lawmakers pushing safety laws in response, and TechCrunch was asking the reasonable follow-up: what's actually behind this wave of doom warnings?

Reasonable question. And an important one, because if you run a small business, none of that coverage was written for you.

Two questions wearing the same headline

There are two completely different questions being asked right now, and the news keeps mashing them together.

Question one: could a machine that can rewrite its own code, pursue its own goals, and acquire its own resources end up beyond human control? That's the question the doomers are asking. It's serious, it isn't new, and the people most worried about it are the ones closest to the frontier.

Question two: should a machine answer your phone at 7 PM on a Saturday, while you're under a sink with a wrench in your hand?

Both get filed under "AI." Only one of them has anything to do with your business.

What actually makes a machine dangerous

Listen closely to the doom argument and you'll notice it always involves the same ingredients: generality (the machine can do anything), autonomy (it sets its own goals), recursion (it improves itself), and resources (it can acquire power and money in pursuit of those goals).

That is a very specific recipe. And it is not the recipe for the software you'd use to book an appointment.

An AI receptionist that answers your calls and puts jobs on your calendar has a job description, not an ambition. It doesn't want anything. It can't pay for anything. It doesn't decide what to do tomorrow. Its worst day is mishearing a name and putting the wrong job in the wrong slot — a mistake you catch in ten seconds, not a civilizational event.

Even the doomers' own framing points here: this is a risk that lives at the top of the ladder — the biggest models, the biggest labs, the biggest ambitions. Which is exactly why, for the work a small business actually needs done, you should be sitting on the bottom rungs on purpose.

The boring truth the headlines skip

While the industry was arguing about extinction, here's what was quietly happening on Main Street.

In June, the U.S. Chamber of Commerce Foundation and Ipsos released the first Main Street AI Monitor. Half of all workers at small businesses already use AI at work, and the overwhelming majority use it to get more done: 64% for drafting, summarizing, and brainstorming; 26% for recurring tasks. Only 6% said they use it to automate a workflow with minimal human involvement.

Read that last number again. The revolution in small business right now is augmentation, not automation. Nobody is replacing their staff with an agent. They're handing the tedious hour to software so the human hour can go to the customer.

And the gains aren't theoretical. Industry research in 2026 puts the average small business at roughly 5.6 hours saved per week per person, with owners saving more than seven. An NFIB survey found the three most common uses among small businesses that use AI are thoroughly unglamorous: marketing content (68%), customer communication (52%), administrative work (47%). Every one of those is a job that currently eats your evenings.

Which brings us to the value list nobody writes headlines about:

Not one of those requires a machine that thinks for itself. Not one of them gets better because the model is smarter. They get better because something finally does them — every time, without being asked twice.

Why the doom news is actually useful

Here's the part I'll defend: the doom debate is the strongest argument for how a business should deploy AI. It's just not an argument for avoiding it.

If even the labs themselves say nobody has solved control, then the correct response isn't "avoid AI," and it certainly isn't "hand an agent your passwords and hope." It's what good engineers have always done: bound the scope, narrow the permissions, keep a human at the decision points, and keep your data under your control.

And that happens to be exactly what the useful deployments look like.

The doom debate is about whether some future system gets the keys to everything. Your question is whether one small system gets exactly one job — the phone, the inbox, the schedule, the books — and does it well enough to give you back ten hours a week.

Different machine. Different ladder. Different risk.

We're not building Skynet

Nobody needs to create a machine that thinks for itself to get value out of this technology. That was always the marketing department's version of AI, not the actual product.

The actual product is a worker that answers, books, follows up, and remembers — one job, done every time, at a price that makes sense for a business with eight employees.

The frontier labs can keep arguing about the end of the world. Good. Somebody should.

In the meantime, the phone is still ringing.

Armory Forge Systems builds AI workers that run real departments for small businesses. If you'd like to stop paying for the calls you never answered, let's talk.