Ghost in the Machines
A couple weeks ago, the lead maintainer of a well-known game engine admitted to using LLM for development since 2021. While not vibecoded (the lead maintainer insisted that stringent procedure was put into place to heavily filter the output), this sent a ripple of shock across the developer communities, many of which opposed usage of LLM in creative fields and sought to distance themselves from the practice. While argument can be made for jumping off ship to another engine, this incident brings about a concerning trend.
For many many years, proliferation of engine packages has lowered barrier to entry into game development. We stand on the shoulder of giants. While the early generations had to practically code from scratch, often tailored for a specific set of hardware, bedroom devs entering the field around early 2000s had plenty of options to play with, most notably the ever ubiquitous, much beloved and much maligned Flash. Flash was not free nor open source, but many other engines are. Eventually, hobbyist devs are able to access the same professional suites of industry-standard engines as an individual. Often, this comes along a lively community providing the engines with plethora of custom plugins, tutorials, even tech supports.
We stand on the shoulder of giants, but the giants are no longer trustworthy.
Who bears the responsibility is important, as platforms such as Steam and itch.io now require developers to disclose LLM usage. The actual definition of this disclosure is murky. Many developers I know painstakingly avoid using one in any shape or form and yet still hoodwinked by, say, a CC asset, and now the very engine they rely on might be tainted. Changing engine is easy, but very costly (time, energy, access to communities surrounding it). What do you say when development is already underway and you stumble upon the knowledge that part of your base engine is LLM-generated? Will you tick that box in Steam and itch.io, knowing full well that most of your audience will not see it the same way? The blame will lay on the developers, as everything else is an invisible layer.
As someone who holds a day job in tech, I do believe most software projects that are still active today have been touched by an LLM somehow, either on purpose by the maintainer or submitted by a contributor. It might not even be intentional. I'm sure people still remember the heady days of early LLMs, before we were made aware of its nature. Even average Joe tried it at least once, throwing random problems they happen to have. Then LLM contributions flooded open source projects so much that it was said to be the death of FOSS as maintainers struggled to cope with the amount of submissions. Whatever piece of public software that are still purely hand-coded must be the work of a small entity with barely existent userbase (though usually fervent, if any). I have had some nice experience toying with such engines and talking directly to the creator, but not everyone likes to take chance this way, especially for a game they intend to release.
Pah, I could hear some elitists spat. Weakly young'uns! Just make everything from scratch, code and art and all, the way we did! Ignoring that it's insane to only let people serve their food if they grow all the ingredients themselves, I lament for the death of collaborative environment we have fostered for decades. To return to solitary activities in fear of accidental LLM usage is such a step back.
I have no good answer for this predicament, as I too grapple with this problem. So far, I can only leave the disclosure form blank, as I do not use LLM outside my day job (anyone insisting that you could always push back in your day job is welcome to offer me a job), but I cannot be sure of the hundred libraries + dependencies used in my game projects. Everything is, in the end, on a best effort basis.