Hovercrafts, AI Agents, and the Age of Perfectly Scheduled Nonsense

A personal story about Fife, feedback loops, and workflows no one asked for.

The next few years are going to be very silly.

Not just fast. Not just futuristic. Silly.

We’ll look back on this era the same way we remember 3D TV or Segways in shopping centres. A brief moment in history when we mistook novelty for progress and raced ahead without asking what we were actually trying to fix.

A few months ago, I asked a reasoning model I’ve been working with, OpenAI’s o3, a question I’d always been curious about:

“Could the hovercraft between Kirkcaldy and Edinburgh be revived?”

I know – niche. But if, like me, you grew up in Fife, the 2007 hovercraft trial was thrilling. For two glorious weeks, over 32,000 people crossed the Forth faster than by train and cheaper than by car. It made national news. Then the 2008 crash hit. Budgets vanished. The idea was shelved. Right idea. Wrong time.

Still, I was curious. And o3 took it seriously. It found feasibility studies. Modelled demand. Calculated operating costs. Recommended hull coatings. It even offered to help me prepare a grant application to Transport Scotland, though I’d still have to write the cover email and hit send. The response was thoughtful, well-researched, and fast. Like talking to a transport consultancy that never slept or sent me an invoice.

And importantly, I was still in the loop. I asked. It answered. I evaluated. That loop is already breaking. Because this next phase of AI doesn’t wait for you to ask.

We’re entering the age of agents. Tools that don’t just assist, but act. They’ll book meetings, send emails, write proposals, apply for grants, and drop calendar invites into inboxes you’ve never seen.

Microsoft is baking them into Copilot. Google is embedding them in Workspace. And Idris Elba is now in a TV advert telling us that AI agents can run your entire day.

At first, it’ll feel like magic. Your agent schedules a call. Their agent accepts. Slides appear. Minutes are summarised. Tasks are auto-assigned.

And then it gets silly.

You join a Zoom call, and no one knows who arranged it. A proposal arrives from an agent you don’t remember briefing. You’re thanked for feedback you never gave, on a project you didn’t request, based on a task you didn’t authorise.

You didn’t forget. You just weren’t consulted.

We’re building systems that act without context. Tools that make decisions without judgment. Workflows that move fast but mean nothing.

And it’s not just about speed. It’s about people.

Every time we take someone out of the loop, we don’t just remove friction. We remove jobs. We remove ownership. We remove the quiet, thoughtful moments where someone might have said, “Wait, is this even a good idea?”

This isn’t about bad work. It’s about meaningless work. Tasks with no intent. Decisions with no authorship. Activity for the sake of it. When everything becomes frictionless, we lose the chance to ask: Should we even be doing this?

Some things are meant to be hard: Becoming Prime Minister. Passing your driving test. Cancelling Sky TV (I wish it wasn’t). Because effort is a filter, it forces you to stop. To weigh. To choose. Remove that filter, and we don’t just move faster. We skip past understanding. We let software decide what’s worth doing based on whether it can be done, not whether it should.

I still believe tools like o3 will shape how we think, plan, and solve. They’re brilliant when used in dialogue, when they extend human capability without replacing human agency. But the current wave of AI agents, pitched as miracle workers for workflows no one was struggling with, feels like a frictionless fix to a problem no one defined.

AI agents will one day transform how we work. But this chapter, where they act without reason, and we celebrate it, is pure theatre.

In its current form, it’s not the future. It’s a novelty ride to nowhere.

Like a Segway in a shopping centre. Or a hovercraft to Fife.

By David Jamieson