Post

"LLMs can help non-programmers with no technical background get a foothold in doing something that used to require a LOT of technical knowledge to get started in, putting them in a better position to learn the technical stuff over time by providing a less steep learning curve" feels uncontroversial?

I think people who learn gamedev via LLMs will naturally gravitate towards using more technical/code-oriented tools. LLMs can't read minds, a user who knows what they want will either need to get much better at technical writing (in natural language) or just learn a programming language.

I've found limited use for LLMs in game *design* because code is ultimately a more concise and easy to edit representation of my project's rules than anything I could write in english. They're great for debugging though.

A side note- Board games and TCGs also run on code, it's just code for humans! The role of a game designer has always been about formally defining rules and instructions for a game, but that's never had to require a computer. It's cool we now have tech that can easily bridge that gap.

The fact that they fail to do the right thing if given vague instructions because they can't read your mind is a feature: it means that if you want to get what you want out, you have to improve your technical writing skills and have a clearer idea of what you want.

Working with an LLM might enable people to build up these skills while, say, iterating on a game. Instead of while trying to do exercises to learn C#. This is a lot more motivating. And when just using an LLM starts to break down, that is an on-ramp to picking up more technical knowledge.

If it turns out that things are easier to express in code at this point, so much the better. The point is not to avoid learning code: The point is to keep code from turning the learning curve of making games into a cliff.