Florian Rival

I tried to explain my view but maybe that was not convincing, that's ok we can have different views! I do think there are like any technology bad usage of it (gen AI for art) and good usage of it, and with game engines I'm hoping we can find good usage of AI which is probably helping to code/no-code

For learning, I meant that the AI has a mode that doesn't build anything in the engine, and instead helps teach users by replying to their questions in any language and linking to documentation and tutorials.

I understand your position though, I just view AI generating code differently.

As AI is currently still not cheap enough to run on a device, it's indeed part of the premium plans.
I hope this is useful to clarify my long term vision (not trying to change your mind, even if I'm sad for game jams). I understand not wanting your brand to be associated with AI though.

The GDevelop vision is: reusable concepts (objects, behaviors, logic built on actions/conditions), easy to learn and assemble any game, with an open-source engine + strong ecosystem (free or commercial stuff). So I think a good AI could help new gamedevs and we must experiment with this now.

For example: I could imagine in 2-3 years that it's normal for everyone to use an extension that helps you autocomplete/build GDScript with a small LLM trained on it. And that would be great for all Godot developers! As long as you understand the code rather than blindly trusting the output.

Using an LLM to help coding is slowly become more normal among developers. I'm hoping there is a good chance we'll soon get open-source/local LLMs that will perform well (and with training data with no ethical/copyright issue) and with a small inference cost (could run on device).

I agree, it's a controversial subject. So I haven't integrated any form of AI generative art, because it's not helpful for our users, and has serious issues with copyright/stealing art. But on the "coding" side I have a different vision: I think it's safer and useful.

Anyway, even though I'm sad to lose your sponsorship here, I wanted to give my reasoning behind the core team spending time on this and my hope to find real open-source (not just open weight), running-on-device LLMs in the future. Hope we can work again on other things in the future :)

So this informed my decision to spend more time on the AI tools here (the "agent") because it seems genuinely helpful to beginners or even intermediate users, who can ask things in plain language and get from fairly ok to great results and learn/unblock themselves.

For the second point ("agent"/"Ask" AI), I've done a lot of user interviews in the past months with devs (GDev users or not) and everyone was using either a bit or a lot LLMs (for searching help or generating code). I've observed this to become common (sadly or not, depending on the point of view).

I'm sad to hear this but I can understand your choice! Just to be clear, is this related to gen AI for art (which we don't do, we pick always assets from the store - clarifying this point to be sure) or for the coding/"agent" aspect (the "Ask AI" button in GDevelop)?

GDevelop: Free, Fast, Easy Game Engine - No-code, Lightweight, Super Powerful | GDevelop GDevelop is a free, fast, easy, open-source game engine. Learn 2D and 3D game development with our tutorials. Publish everywhere: iOS, Android, Steam...

If that sounds interesting to you, or if you tried GDevelop in the past, give it a try: gdevelop.io
or come help us making the best open-source, super fast, lightweight, incredibly accessible game creation app in the world: github.com/4ian/GDevelop 🥰

We're glad to see not only hobbyists, but also indie developers and large game studios using GDevelop. And we're seeing an uptick in its adoption in education.
And we're about to launch a new, integrated, complete course to learn GDevelop directly from the app.

With both a strong commercial ecosystem and a strong open-source project and contributors, we can go very far, and build the foundation, go-to solution for creating games (of all kinds) or any interactive media.
Anyone can build something with GDevelop.

But that's not the only thing. A game engine can be good, but you need an ecosystem around it. That's why it's open-source and that's why the GDevelop company is building commercial offerings: cloud projects, leaderboards, multiplayer, cloud builds, education plans, ads support..

Building a great game editor is good. It's even better when it's made to be super accessible, like a design tool, with an intuitive system of behaviors that can be composed and "visual" events so that "coding" is fast, concise and does not need in-depth knowledge of a language.

Our goal with GDevelop is not to make the best open-source game engine in the world - though that would be a nice side effect! We're there to build on top of state-of-the-art, industry standard open-source technologies (think: PixiJS, Three.js, Jolt) to make them super accessible

Looks like my baby engine is now a grown-up!
Now with 3D, 3D physics, multiplayer, prefabs - all open-source - but also an entire ecosystem (asset store, cloud builds, leaderboards, player accounts, 100s of extensions, premium templates). And it's just the beginning ⬇️⬇️