Studio News
I'm Not Afraid of AI. Twenty Years of Maintenance Taught Me Why.
September 2026
Most of what gets written about AI and privacy is written from fear: fear of what the models can do, fear of who's watching, fear of what gets sold. Some of it is fair. But fear doesn't help you decide anything, and it certainly doesn't help you build anything.
I spent twenty years as a maintenance technician, working on engines and then aircraft in the UK. In that world, being afraid of powerful machinery isn't a strategy. You learn how it works, you follow the procedure, you use the right tool, and you check your work. That's the mindset I bring to software now, and it's the reason this studio exists.
What maintenance taught me about powerful tools
Three habits from the hangar carry straight over to building with AI:
- A tool speeds up the job. It doesn't share the responsibility. A torque wrench makes the work more accurate, but it never signs the work off. A person does, and that person had better understand what they're signing.
- The faster the tool, the more the checks matter. An unchecked quick fix is still a shortcut, whether it's a fastener or a function.
- Understand the system before you trust it. You don't need to be afraid of a machine to want to know exactly where it can fail.
That's how I work with AI. I build these apps with AI tools, and several of them run AI models entirely on the iPhone. OffgridScribe, for example, transcribes speech with a model that lives on your device. I review what the tools produce, test it, and publish what I find, including the network audit we ran on our own apps.
Privacy is what good design gives you
This is where I think the usual privacy conversation gets it backwards. The answer to powerful technology isn't to avoid it. It's to use it in a way that keeps you in control. When the model runs on your phone, the recording never has to go anywhere. You get the benefit of the technology without handing over the data. That isn't fear. It's design.
The uncomfortable part: who says it matters
I heard a podcast recently that made a point I haven't been able to shake. Imagine two people publish the exact same article, both written with the help of AI. One is a well-known name. The other is a stranger. You already know how it goes: the known name gets far more reach, more likes, more trust. It's a thought experiment rather than a study, but I don't think many people would bet on the stranger.
AI has made writing cheap. Anyone can produce a clean, confident article in minutes. What it can't produce is a reason to believe the person behind it. That's the scarce part now, and it's why I'd rather put my name and my background next to this work than hide behind a logo. It isn't about sounding important. I run a small studio with no big audience, so the only way to earn attention is to show the work. If I get something wrong, there's a real person you can ask.
What to expect from here
- Original testing, with the methods and the limits written down
- How the apps are actually built, including where on-device AI falls short
- Tech news and analysis, looked at through one question: who's in control?
- Corrections, when I get things wrong
Technology isn't the thing to be afraid of. Not knowing how to use it is. That's what I'm trying to get better at, in public.
More from Offgrid Studio
Also read: About Offgrid Studio and We Ran Our Own Network Audit. Here's What We Found.