Flying Fries and AI

Introduction

Artificial intelligence is increasingly present in our lives, in the news, and across all publishers, small and large.

Debates continue about the responsible use of these new tools, their relevance in a professional environment, and the quality of the content these LLMs manage to generate.

The most obvious cases are the images and videos generated by Sora or similar models, which are generally amusing, sometimes impressive, but rarely interesting. And it's still fair to say that an image or video created by a real human artist, with real sensitivity, will always have more value than this generic stuff.

The idea of this article is not to be hypocritical about the situation—simply blaming AI and claiming that everything is AI's fault through a quick statement. Since this is a technology that isn't going to disappear anytime soon, and since not everything is black and white, that's where we need to dig deeper: where do we want to stand on this?

2023–2024: Tentative First Steps

At Flying Fries, the story with AI started early — and modestly. Back in 2023 and 2024, our use of AI was essentially limited to texture work with MidJourney: grunge effects, dirt layers, and wear-and-tear patterns that we needed as raw material to create our final textures in Photoshop and Painter. Nothing groundbreaking. Nothing that replaced actual craft.

2025: A Shift

2025 marked a turning point in the tools we used, if not yet in the depth of our usage. We moved progressively from MidJourney toward Artlist.io, where generators like Nano Banana offered more control and more interesting results for illustrations — including some of the posters and visuals you can find inside the Quasar. (We also used the same technology for new illustrations in the recent XF-11 update.)

It's also through Artlist.io that we explored text-to-speech more seriously, which gave birth to K.A.R.A. — the Quasar's synthetic voice, with her perfectly calibrated sarcastic tone. That one was fun.

Still, all of this remained fairly marginal. Useful, but not transformative.

The Real Turning Point: Claude Code

The shift happened when a friend, Pascal, introduced me to agentic development with Claude Code.

In less than two days, Claude Code allowed me to completely rework and secure the technical layer of the Flying Fries website — with zero downtime and no visual changes whatsoever. What followed was just as significant: working with Claude Chat to produce complete front-end and back-end designs, then bringing them to life step by step in my VSCode environment over the course of a few days.

I wouldn't have had the time — or the skills, if I'm being totally honest — to produce a website that clean and that complete on my own.

Since then, AI — and Claude in particular — has become a regular part of how I work at Flying Fries. Not to cut costs (as a reminder: Flying Fries has no employees, and I'm still working full time for another client alongside this). But to raise the bar on what I can actually deliver on my own.

Which brings me to what this means in practice...

More Ambition, Not Less Work

Thanks to AI, you now have a new website that's far more user-friendly and more secure than the previous one. You also have — if you haven't found them yet — some genuinely useful Windows tools in the Resources section: a Flight Simulator Starter for your main simulator and all its satellite applications, a translator, a launcher on steroids for VSCode, and an evolution of a Blender plugin I had started years ago, now polished enough to share. All free, as always.

There will almost certainly be AI-assisted elements in future add-ons too — whether that's code, tools, or custom-crafted lessons to help me tackle things that felt inaccessible before, like WASM.

The Hind's weapons (even if they're only for show) are an area where WASM could let us do some cool things. The same goes for finally bringing physical believability to the XF-11's drop tanks and the Scrapyard Monster's ballast.

Working With More Humans, Not Fewer

Here's something I want to be clear about, because it matters to me: using AI more has not meant working with fewer people. If anything, the opposite is true.

Steiny is still my flight model master — that relationship hasn't changed. Captain Kenobi and I have near-daily exchanges, and that collaboration is as strong as ever. I still very much look forward to working more with Pseud; life on both sides has made synchronization difficult these past few months, but that's temporary. What we built together on the Quasar was something special, and I can't wait to see what Pseud brings to the Gabriel — and to the Hind.

Beyond that, I'm also investing significantly in overhauling the sound packs across all our aircraft — free updates for everyone — carried out by Grégoire, a talented freelancer who absolutely deserves to be paid for his craft.

And there's a new project in the pipeline, which have been announced by the time you're reading this: a scenery pack covering airports from around the world, with the ambition of having different artists contribute to different airports. It's early days — we'll see how many people end up involved — but the direction is clear:

Each aircraft Flying Fries has released has meant working with more people than the one before. That's not a coincidence. It's one of the greatest things this adventure has given me.

For this airport pack, in order to get something unique with different and strong vibes at each location, I'm also using image generation (Chat GPT, Nano Banana) to create imperfect but interesting concept art. It brings up ideas and provides quick visualization of what a location could look like and which direction to follow or avoid. It's very powerful for getting you on the right track quickly, but it doesn't replace the real work that needs to be done afterwards, and most creative ideas still come from my sick brain anyway. It's just one more tool and a real boost for independent studios like Flying Fries.

Staying in Control — Keeping It Human

The question I ask myself every day is: How do I make sure it's still me? (Scary thought, right?)

The answer is: I check and validate everything. Every piece of text that goes through an AI tool gets reviewed. When articles are translated or reorganized, I go through them to make sure my tone is preserved. I make manual corrections and feed those corrections back to the model so it can calibrate better over time. It's a symbiosis, not a delegation.

No text goes live without me confirming it actually sounds like me. That's a line I won't cross. And actually, there is still A LOT of content out there that is directly, unfiltered, and uncorrected (whoops!) pure me!

Conclusion

I'll be honest: I resisted for a long time. I genuinely believed that only human expertise is truly interesting. But I've come to realize that human expertise can only endure if we use the right tools of our time. That's been true since the railways, since the internet, since every wave of technology that people initially feared would erase or replace them.

AI is ultimately a very advanced, very seductive, and potentially dangerous tool. Which is exactly why it needs to be learned, managed, and kept in check.

The goal at Flying Fries is not to use AI as a profit multiplier or a shortcut. It's to use it where it's relevant, while making sure that everything it touches still carries our human texture.

That part is not negotiable.