Some time ago, I decided to revisit the idea of creating games using a framework instead of a game engine. The reason was quite simple: I enjoy tinkering with code, and I dislike mouse clicking.
Making a game from scratch in code makes me feel I understand what is going on, because if something is hidden, it is hidden behind a coded function. When doing the same in a game engine, you need to do a combination of mouse navigation and property filling. A lot is hidden under the user experience.
When I tried an engine, I felt overwhelmed by the number of options and my lack of knowledge of what lies beneath. That, and the fact that reproducing all that you do requires "muscle memory", which is not reusable like code. And I do prefer systems that don't rely on memory that much.
Engines require too much to remember for their own sake.
So I went for game frameworks.
But also, I went for AI. My company also went for AI, training everyone in AI in a developer’s job.
I recently finished an AI course and achieved a certified Vibe Coder status1. I played around with Cursor and Visual Studio Code, and I have to admit it's a very comfortable tool to use.
For example, I know how to make websites. I've been making them since the early 2000s. These days, you can use some fancy SaaS, or you can code it yourself. Or you can explain your website to a coding agent and let it get to work. Never have the problem of centering a div vertically! :D
If you know version control, know how to break down your tasks, how to write automatic tests for it, and how to vibe code all of it - YOU CAN BE UNSTOPPABLE!
And sometimes you need to learn.
There will be many impediments in your road.
One of them is communities. Before AI, developers relied on communities. Communities are great! But communities can be toxic. Sometimes you visit a forum to share a problem of yours, and all you'll hear is that you have a "skill issue". You want to learn, and all you'll learn is that you need to learn more. Thanks, Captain Obvious!
So you go to Chad Gippity, world-famous therapist, coach, and best friend2. Without judgment and toxicity, only helpful guidelines. Yeah, I'm glad AI will take their jobs.
Enter Bezy.
Among many weird decisions (installation fee), Unity actually did something good - an AI agent for game developers! And from what I see from Thomas Brush's YouTube clip, it resembles what an AI agent does in, e.g., Visual Studio Code.
I'm not saying game engines should implement AI agents...
Actually, this is what I'm saying.
DO IT!
PS. I am especially interested in Godot implementing this. As I aim at making 2D games right now, Unity and Unreal are a bit overkill for me. I know people are working on this at this moment, so my hopes are up.
It would be much easier to implement if engines had a code-only mode.
But we work with what we have!
No, I don't know for sure how much time I will brag about it
You know everything in the world is screwed when Chad can replace these people in most tasks
This is an interesting read, specially coming from someone who already makes games, I imagine the industry will be moving on that direction, actually that’s good having human validation on the whole process.
But people willing to learn or start their own games, I would say tools like Pico-8 (more visuals, but coding) or Love2D (pure coding) are the best to understand what’s under the hood and then be able to Jump into coding and AI.
Unreal Engine is pushing its blueprints pretty hard so it's actually harder to find examples and tutorials for doing things in C++ purely, but the bigger the game and the less Indie the developer the more of it ends up being in c++.