Game developers discuss the Big Walk controversy and the role of generative AI in programming

Yesterday, I called on game developers—and specifically programmers—to share their views on the Big Walk debacle. In case you missed it, House House's warmly received social stroller has been placed at the centre of a emotionally-charged debate about the intersection between video game production (namely, coding), generative AI, and our very human response to those things. 

A brief explainer for those who have (perhaps mercifully) only today stumbled into the social media pain dome. On Monday evening, a number of Independent Game Festival jurors published an open letter to plant a flag in the ground. They will not be voting for submissions featuring generative AI outputs, including code. That was, they explained, their personal choice and not a boundary imposed by the festival itself—which is organised by GDC parent Informa.

Shortly after, social media chatter and an eventual statement from House House confirmed Big Walk was the video game at the centre of a passionate debate between IGF jurors. Crucially, the indie title was not the only submission to disclose the use of generative AI at the behest of the latest IGF guidelines. Even so, House House confirmed they used LLMs "sparingly"—largely in the form of chatbots—towards the end of production to solve a particularly stubborn technical challenge. 

"In some cases, snippets from LLM responses were copied as a starting point to be modified into game code—we’d estimate that well under 0.1 percent of Big Walk’s code had anything to do with this usage, and generative AI was not used in any other context when making the game," said the studio.

"We needed to cull shadow casters by distance from the camera, but Unity URP only allows culling shadow casters by distance in camera forwards, so shadow casters behind the player were being rendered at an infinite distance. This made our game run 50 percent slower for around 20 seconds at sunrise and sunset. We researched the problem extensively, and we looked at the URP code ourselves, but we couldn’t find a way. Then, we were able to speak directly to an engineer at Unity, who asked around and ultimately concluded that what we wanted wasn’t possible. Finally, we posed the question to an LLM chatbot. After a few dead ends, it suggested a possible fix, and while we had to modify the suggestion extensively, it worked. This fix represented about 10 lines of code inside URP."

Pandemonium ensued on social media. The reaction has ranged from explosive outrage and deflated disappointment to unbothered apathy and shoulder-shrugging incredulity. People poured themselves into Big Walk in both the digital world and our corporeal realm, and it's evident that many felt betrayed by not only the use of generative AI—a topic that remains impossibly divisive due to concerns around labour theft and environmental vandalism—but also by House House's failure to publicly disclose its usage until being outed.

A number of people I respect and admire have already published brilliant, nuanced op-eds and Bluesky threads on the subject that I implore you seek out. I personally found Brendan Keogh's piece 'AI is history' to be a fascinating deconstruction of the generative AI problem. I'm personally still working through my own feelings, both as a journalist who recoils at the notion of generative AI even attempting to produce the written word and someone who has never been asked to code or debug a video game. 

And so, this piece is not about what I make of the big talk around Big Walk, but rather a chance for me to spotlight the views—of which there were many—of developers, mainly programmers and engineers, who have been dropped in the trenches with generative AI tools (for better or worse) and told to get to work. 

"I simply cannot understand the mindset of someone who wants to spend more time debugging, under pressure, or simply not getting paid"

Firstly, I think it's telling that pretty much everybody who reached out to share their two cents asked to remain anonymous for fear of insult or reprisal—either from their peers on social media or their employer, many of which seem to be going all-in on generative AI tech. As you might expect, there were plenty who immediately debunked the oft-touted notion that generative AI is now simply a part of coding. An unshakable curse the industry must collectively accept.

"LLMs are obviously not 'impossible to avoid.' Just turn them off, don't use them, become a software vegan," said one person, a contractor who has worked in the game industry for the best part of a decade. It's a statement you might feel will be followed by the inevitable chaser that 'AI is terrible'—and yet, you'd be wrong. 

"LLMs writing programs? They are better at it, without even very much effort. Even if you only asked Claude one careful question a week on a tight project, you will have saved yourself hours in that week," they continued.

"Debugging in particular—just paste some gigantic piece of shit code with a rare bug that's had you tearing your hair out. It'll have you a correct answer in a few minutes. What kind of programmer doesn't like to deliver stuff quicker? I simply cannot understand the mindset of someone who wants to spend more time debugging, under pressure, or simply not getting paid, for want of a tool that costs very little. Nobody cares what a program looks like, apart from other programmers. Sure you can tie yourself in knots with an LLM, or not know when it's got stuck, or watch it write a cathedral only comprehensible to god. But largely—they just solve the problems you put in front of them, and you need to learn how to pace them, and yourself."

This source, who said they have been programming for over 30 years, claimed it is hugely liberating to be able to quickly overcome those roadblocks—especially when, in their own words, certain aspects of their skillset "have plateaued."

Another source, a developer currently working at a major studio, said House House has already been absolved in their mind for only using LLMs to overcome an issue after they had seemingly exhausted all other avenues. "Their other options were either to bash their heads at it for weeks, or ship the game with even more performance issues (and it's already a demanding game)," they said.