‘Big Walk’ and the Question Nobody Can Answer About AI

House House likely did not expect to become the center of one of the year’s most revealing discussions about artificial intelligence in video game design. Yet after the studio disclosed that Big Walk, their smash 2026 multiplayer hit, utilized a large language model (LLM) during the ending stretch of development, the reaction quickly expanded beyond House House’s own press release. Instead, the discussion, both in the Epilogue Gaming Discord as well as the comments section of Aftermath’s reporting on this AI news, shifted into much broader arguments about the morally concerning aspects of AI usage: stolen labor, lack of transparency, blatant environmental consequences – and, in this case, the purpose of cultural institutions like the Independent Games Festival. In both forums, the conversation arrives at the inevitable: games, and the people who play them, are overdue for a naked reckoning with the fact that AI is, in some ways, here to stay.

Before going any further, I am not a proponent of AI. Like so many of my articles this year, the reason I am compelled to opine on this issue is that I feel like I’m watching people speak past each other because there’s no conversational referees in online discourse to point out the inconsistencies in the moral reasoning that often occurs when AI is brought up.

The facts, at least, are relatively straightforward in the case with House House. They state that generative AI was used late in development while attempting to solve a difficult technical problem within Unity’s rendering pipeline. According to the studio, “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.” House House estimates that “well under 0.1%” of Big Walk’s code was connected to this process and emphasizes that “generative AI was not used in any other context when making the game.”

House House, developers of my beloved Untitled Goose Game, has earned enormous good will within the gaming community, and it pains me to see them scrutinized so harshly in a moment of transparency. Though, I understand why. Ordinarily, discussions about AI in games revolve around relatively clear ethical cases, like blatant artistic theft, the devaluation of creative labor – cases where the concerns are immediately visible. When the final creative work itself is generated by AI, questions about authorship, artistic integrity, and so forth become impossible to avoid, and rightly so. With Big Walk, the news is heartbreaking because of how cherished this multiplayer experience has been for so many people.

In retrospect, even that framing may give Big Walk too much importance. One of the most illuminating interventions in the discussion came from IMMORTALITY creator Sam Barlow, one of the signatories to the IGF juror statement that helped ignite the controversy. As Barlow later clarified, “Big Walk was not the impetus for this discussion. The impetus was a large number of games using gen-AI, and some in much more substantial ways.” What made Big Walk significant, he argued, was something else entirely. “When then trying to take a stance, Big Walk became the tricky edge case.”

We’ve been lucky enough that most notable examples of AI usage in games have been obvious and relatively easy to unite around. Neither developers nor players struggle to articulate objections to a project that uses AI-generated assets or attempts to replace creative workers altogether. The issue, Barlow highlights, emerges when an example appears that doesn’t comfortably fit within those preestablished patterns. 

The Big Walk controversy is unusual because it’s directly situated at the point where I notice many people’s ethical intuitions begin to diverge. In reading through the discussion, I’m not convinced that there is any clear consensus on how to wrestle with the news from House House. Almost everyone I’ve seen engage in the discourse agrees that AI poses some serious problems – the issue becomes the dramatically different conclusions we reach about, in this case, what House House did wrong.

Even House House’s statement feels aware of this tension. After contextualizing the circumstances surrounding the aforementioned bug fix, the studio made sure to emphasize that the disclosure was “not an endorsement of, or argument for, the use of AI tools, however minimal” and expressed concern that voluntary disclosures might contribute to “the normalisation of this technology, or its treatment as ‘inevitable.’” It’s strategically interesting to watch House House simultaneously defend its actions and distance itself from them, acknowledging that clearly some players will regard any use of an LLM as unacceptable, meanwhile others would argue that the specific circumstances mitigate any condemnation. In many ways, this contradiction seems to define the conversations that followed.

The primary concern, for many, is the issue of labor displacement. That is, generative AI is objectionable because it replaces human creativity, devalues artistic effort, and provides corporations with an easy justification to reduce headcount. House House’s case, however, appears comparatively narrow in this regard, if their statement is to be believed at face value. No artist was replaced, no actor lost work, and no writer was displaced, according to their statement. In this schema, the studio’s actions might be seen as disappointing, as I’ve repeatedly read, but they also do not resemble the AI adoption stories that typically generate more universal, widespread outrage across the games industry – like when Epic announced it was converting development entirely to AI.

At the same time, however, others believe the problem was never solely about labor. Another prominent objection involves participation itself. That is, if generative AI systems are unethical because they exploit training practices, environmental costs, copyright violations, and other corporate behavior that any decent person would consider reprehensible, then the distinction between generating an open world and generating a bug fix appear vastly different in scope and scale. From this perspective, the distinction loses importance, because the amount of AI usage matters less than the fact that the technology was used at all. I find this conclusion troubling.

Beneath Aftermath’s article, this disagreement repeatedly surfaced, with one commenter summarizing the stricter position succinctly: “genAI uses stolen work. Is it ok for a game to use only ‘a little bit’ of stolen work? Is it acceptable to use stolen work ‘only in the final stretch’? genAI should be banned. No excuse is acceptable for the use of genAI.” Regardless of whether I agree with this statement, the view espoused is internally consistent and honest, leaving little room for edge cases because it rejects the underlying technology wholesale.

And yet, many people involved in these discussions about House House, who broadly oppose AI, nevertheless arrived at strikingly different conclusions about their AI usage. One commenter wrote that “if they specifically used programming questions, I think that’s the least bad version of AI usage.” Another argued that “using AI as a tool to help you problem solve and figure out how to fix something and have asked others with equal or greater experience and nobody can figure it out, yeah I think that’s a reasonable use of AI.” Elsewhere, a commenter observed that the studio appeared to have “genuinely tried to do everything ‘the right way’” before turning to an LLM. Each of these commenters imply that they are already on board with the notion that AI is concerning, or at least not inherently good. Their concerns are more compelling to me than an outright ban against AI, but as you can see, the internal logic driving these people to each conclusion is slightly different – so my curiosity continued.

Similar arguments appeared throughout social media. Some argued that programmers aren’t replaced by AI, they are accelerated by it, describing AI tools primarily as systems that reduce logistical overhead and allow developers to focus on technical design rather than repetitive implementation work. According to this perspective, asking a chatbot for help fixing a rendering bug is categorically different from generating artwork because one augments human expertise while the other attempts to replace it. The anti-AI jurors behind the IGF statement explicitly reject this distinction.

The IGF statement does not merely oppose AI-generated art. It argues that “these harms do not depend on what is generated, be it music or writing, art or VO, marketing assets or code,” before directly stating that “AI-generated code is not an exception to these harms.” The statement grounds this claim in a view of programming as creative labor – a distinction celebrated by many people I follow online – that deserves the same consideration afforded to any other discipline belonging to game development. Under this framework, the issue becomes whether usefulness is the appropriate criterion by which technology should be judged at all.

Taken together, the discourse reveals that many people who consider themselves anti-AI are not evaluating the same thing. Or, at least, some are judging the specific decision House House made, while others judge the broader system that made the decision possible. One side is asking whether House House acted reasonably under the constraints of their difficult and already labor-intensive process. The other is asking whether participation can ever be justified, regardless of circumstance.

The peculiar status of coding itself is perhaps the most intellectually interesting thread running through the discussion of AI in game development. Defenders of House House repeatedly spoke of code generation or code assistance as qualitatively different from, let’s say, image generation. An Aftermath commentor, Josh Clow, questioned why programming received an “ethical pass,” stating, “programming is a creative endeavor.” One of the members of the Epilogue Discord said something similar, “problem solving is an inherently creative act.” Clow continues, “The LLMs people use to do it are still taking data that people put online in good faith for free and monetizing it.” This criticism strikes at the heart of an assumption that I wouldn’t have examined if not for House House’s statement: if AI image generation is objectionable because it’s trained on the labor of visual artists, why should AI-assisted coding be treated differently if it relies upon the labor of programmers?

The fact that many people instinctively feel there is a difference suggests that we are speaking from cultural intuition rather than based on a shared set of self-evident facts. From my limited understanding of modern software development – at least according to my friends who work in the field – this dependence upon reuse, adaptation, and consultation with pre-established workflow is nothing new. Developers routinely borrow techniques from online resources, and most of the IT people I know routinely integrate some degree of AI into their workflow at this point. One commentor, to this notion, explicitly framed House House’s use of an LLM in this case as a continuation of that progress, arguing that this was “exactly like what programmers have always done,” except that “in the past it used to be from Stack Overflow or existing patterns instead of an LLM.”

One might argue that this comparison oversimplifies what actually happens when a developer learns from another developer. Returning to Clow’s comments, he notes that the value of resources like Stack Overflow is not merely the existence of code snippets but the process of engaging critically with them. “The labor involved,” as several commenters effectively argued, is not the typing; it is the thinking. The concern, therefore, is not merely where the code comes from but whether the technology changes the relationship between developers and the intellectual process of problem-solving itself.

Along these lines, the earlier suggestion that problem-solving was inherently a creative act points to the fact that, if the developers had not turned to an LLM, they “might have realized something cool about their code more generally” or developed skills they otherwise would not have acquired. The implication here is that creative labor occurs not only in the final output but in the struggle itself. If one accepts that premise, then the ethical distinction between code generation and creative work becomes far murkier.

There is, however, another dimension to this discussion that I think is frequently overlooked: the increasingly difficult economic realities of making games in the first place. The conversation surrounding AI often assumes that developers have the luxury of choosing between human labor and automated alternatives without considering the financial circumstances that inform those decisions. In reality, game development has become an increasingly expensive undertaking, particularly for independent studios operating without the financial security of larger publishers. Rising production costs, inflation, increasingly ambitious player expectations, and the pressure to remain competitive have created an environment in which maintaining the same creative workforce is not always financially sustainable. For smaller studios like House House, the issue is not necessarily whether they would prefer to hire another specialist to solve a technical problem, but whether they can realistically justify that expense in the first place.

This is where I think some of the criticism directed at House House becomes particularly difficult to reconcile with the realities of development. If a studio encounters an exceptionally difficult technical problem that ultimately requires a relatively small intervention, is hiring an additional employee or specialist genuinely a reasonable expectation? That person still needs to be compensated, the project still needs to account for their time, and those expenses ultimately have to be absorbed somewhere. They might mean a more expensive game, a delayed release, or resources being taken away from another part of development. None of this makes the ethical concerns surrounding generative AI disappear, but it does introduce a consideration that discussions about responsible AI usage rarely seem willing to entertain: sometimes, developers are not choosing between an ethical option and an unethical one so much as navigating an economic environment that increasingly limits their alternatives.

This pressure extends well beyond the games industry. In fields like web development, for example, clients increasingly expect increasingly complex projects to be delivered at prices and speeds that make traditional, entirely human-driven workflows difficult to sustain. Developers who refuse to incorporate AI-assisted efficiencies can find themselves competing against freelancers and businesses that do. The result is a particularly uncomfortable economic incentive: even people who fundamentally disagree with the technology may feel compelled to adopt it simply to remain competitive. In that sense, AI adoption is not always a reflection of enthusiasm for the technology itself. Sometimes, it is a response to an industry that has already begun reorganizing its expectations around what AI makes possible.

I think this distinction matters because it changes where we direct our criticism. If the underlying problem is that companies can exploit unlicensed creative work, displace workers, and introduce technologies with significant social and environmental consequences without adequate accountability, then focusing exclusively on individual developers risks missing the larger structural issue. We can reasonably criticize studios for their decisions while simultaneously acknowledging that the economic incentives shaping those decisions are not entirely of their own making. There is a meaningful difference between holding developers accountable for participating in an unethical system and pretending that they have unlimited financial freedom to opt out of it.

Whereas much of the conversation has focused on what programmers can actually accomplish through these AI tools, i.e. practical value, the IGF jurors, by contrast, are articulating collective concerns as opposed to principally evaluating utility. Their statement identifies “ecological harms,” “economic harms,” “cognitive and psychological harms,” harm to human labor, and the broader information ecosystem. Obviously AI coding is at least somewhat effective in its current state, which begs the question of whether widespread adoption contributes to a future we wish to encourage. 

The jurors explicitly write that “we know firsthand the value of pushing back on what an industry has deemed ‘inevitable.’” Barlow’s later clarification sharpens the point still further: “99.9% of videogames shipped before and without Gen-AI assistance, so this isn’t inevitable.” In other words, the existence of the AI solution opted by House House does not automatically justify its use, even though I’m sure House House was relieved to avoid hiring yet another person to get the job done.

Josh Sawyer’s writeup on Clockwork Worlds addresses this concern directly, repeatedly returning not to capability but to agency. Throughout the article, he argues that the question facing IGF jurors is different from the question facing ordinary developers navigating economic realities. Awards, he writes, represent one of the rare spaces in which participants possess unusually strong collective agency. “This is not a place where you’ll be fired for choosing to draw a line,” he argues, before describing AI voting decisions as opportunities to resist normalization rather than merely adapting to it. For Sawyer, institutions like IGF have an ethical obligation to use moments of cultural recognition to reinforce or resist emerging norms.

Many observers have interpreted the IGF statement as a blanket condemnation of developers who use AI. Yet, the statement itself is about awards, not solely developers. Barlow emphasized this explicitly when he later wrote that the jurors’ position was “not a call to boycott a certain game, or to economically hurt a team,” but rather an attempt to define what a specific award should represent.

One of the final concerns I saw repeated throughout the House House discourse concerned transparency. Many critics of this situation were less upset by the bug fix itself than by the timing of this disclosure. With several commenters repeatedly describing the situation as “lying by omission,” House House is now facing severe scrutiny for whether its undeniably popular multiplayer title will be barred from year-end awards categories due to this news. Some people argue that House House could have disclosed the information at launch and that its decision not to do so undermined later attempts at transparency. Another commenter insisted that because House House had apparently discussed disclosure internally, the lack of disclosure until this point represents an actual ethical choice rather than an unfortunate oversight.

I’ve seen defenses against this line of attack, though I find myself sympathetic towards the framing of House House’s decision as an ethical choice. The defenses here feel flimsy, i.e. Steam, for example, does not currently require disclosure of chatbot-assisted coding, so why would we expect House House to go above and beyond here? This information emerged through the IGF process, after which the studio publicly explained exactly what occurred. Indeed, much of the backlash stems from the fact that House House chose to be unusually specific, not just generically saying that AI had been used, but describing precisely how: with ten lines of code.

For me, this situation raises a difficult question that anti-AI communities are going to have to confront eventually, which is the question of what incentives exist for disclosure. In other words, why would a company simply not “lie by omission” if the backlash to the news that they used AI for a bug fix would disappoint so many potential players?

My instinctive answer is that transparency is good regardless of consequences, and we want to incentivize more of that. Yet, transparency functions differently in practice than in theory, as I’m reminded when reading through the comments on this issue — to the point where I’m getting flashbacks to an article I wrote earlier this year. Arguably, in order to preserve nuance within this conversation, we should emphasize that we want people to be open and honest about their AI usage. Along these lines, another Epilogue member raised a concern that appeared repeatedly throughout this conversation: “If we’re going nuclear on everyone who uses AI for anything at all then a lot of people will just hide their AI usage.” This concern lies at the heart of my interest surrounding the House House case.

Whether one agrees with this assessment or not, the concern is difficult to dismiss entirely, as critics often argue, reasonably, that players deserve information that helps them make informed decisions. But if every disclosed use of AI is treated as morally equivalent regardless of scale or context, then the practical incentive shifts away from disclosure and toward concealment. House House only became a case study because House House disclosed; their (dis)honesty is the thread of their own unravelling. It is impossible to ignore the possibility that future developers will look at this controversy, therefore, and conclude that honesty carries significant reputational risk.

Obviously, transparency does not exempt one from criticism or consequence. You don’t avoid jail time because you confess to a murder, so to speak. Rather, transparency here means communities must eventually decide how they wish transparency to function – what the value of it truly is, and how much trust we can afford to place in the creators of any art we consume.

The deeper problem that remains is that discussions about AI frequently collapse distinct concerns like we’ve explored above into a single condemnatory category. Labor issues become synonyms intertwined with copyright concerns, which become twins with environmental concerns, and so on. Our anxieties about technological dependence, corporate consolidation, and cultural homogenization are entirely justified. Full stop. Our concerns here often point in the same direction, and because of that tendency, there is rarely a need to disentangle them. A case like Big Walk forces that disentanglement, if we are being intellectually honest with ourselves.

Overall, I think the overwhelming gaming discourse is currently exhausted by the ubiquity of AI. Or, to borrow from another Epilogue user, “I’m fucking tired of AI everything.” Yet, this same Discord member ultimately concludes that House House’s bug-fix scenario might be one of the situations where AI’s benefits outweigh the costs. Another participant in this conversation reached a similar conclusion, arguing that they supported disqualifying the game from certain awards if that was the criteria, while simultaneously stating that they had “a hard time being too upset with the devs on a personal level” given the circumstances and the studio’s honesty.

These are not contradictions so much as reminders that people often hold several values simultaneously – and that’s okay. The difficulty emerges when those values point in different directions. I think that you can simultaneously oppose AI companies, gen-AI, support labor protections and environmental sustainability, etc., as well as conclude that House House’s specific actions here were understandable, even reasonable. Likewise, someone can support indie developers, appreciate Big Walk, welcome House House’s previous political stances (supporting Palestine), and still conclude that any use of generative AI crosses a line.

A curiously positive note in the swirling controversy surrounding Big Walk is developer Natalie Spitzel proposing and then later creating a public network where developers can volunteer their expertise and make themselves available to answer technical inquiries from other developers. The idea’s premise is surprisingly simple: if the concern is that studios increasingly turn toward AI when confronted with difficult technical problems, perhaps communities should invest more heavily in strengthening human alternatives. Say what you want about AI usage, but I think this project is beautiful, one of the rare moments when social media acts a true force for good.

This proposal shifts the discussion away from individual moral judgment and toward institution-building. Since the debate around AI is primarily adopted into a binary choice between AI assistance and unsolved problems, the creation of a community-driven support network implicitly rejects that framing. Rather than asking whether House House should have used a chatbot, it asks what structures might make the chatbot unnecessary in the first place. 

The future of AI in game development will not be determined solely by abstract ethical principles. It will also be shaped by whether communities invest in alternatives. If people genuinely want to reduce dependence on AI systems, then declarations alone are unlikely to be sufficient. Human infrastructures must be strong enough to compete with the convenience AI currently provides.

Ultimately, I do not think this controversy is really about ten lines – or “well under 0.1%” – of code in Unity. Rather, we are all working from slightly different moral frameworks, and when we end up discussing something as amorphous, broad, and slippery as AI, we often end up missing each other through the discourse.

Unlike obvious asset-flip examples that we’ve been seeing in the wake of generative AI game development, Big Walk presents something far more uncomfortable: a beloved independent studio, celebrated for its humanity and community-oriented design philosophy, admitting to a narrow and highly specific use of a technology that many people view as fundamentally harmful. It collides our competing definitions of what the AI problem actually is.

When House House warned against treating AI as “inevitable,” they were speaking to one anxiety; when critics accused them of “lying by omission,” they were speaking to another. When commenters debated whether coding should receive an ethical exception unavailable to artists, that’s a different conversation entirely. Each position explored above reflects a different understanding of what is at stake, and each produces a different answer to the same set of facts. And I find that fascinating.

Since writing this, the controversy has developed another dimension, as Big Walk has been hit with a wave of negative Steam reviews following House House’s disclosure. The shift was particularly stark on November 10, when the game received 366 negative reviews alongside 426 positive ones, compared with just three negative reviews the previous day. The game has retained its overall “Very Positive” rating, but the review activity is still revealing. Whatever one thinks of House House’s decision, one has to reckon with distinguishing when criticism of a developer’s choices become an attempt to punish the work itself.

The most important question raised by Big Walk, for me, is whether the games industry – players, critics, and developers – can clearly explain where that ethical line with AI exists and why. The intensity of this debate suggests that many people assumed they already knew the answer, but what the reaction to Big Walk has revealed is that the answer is far less settled than anyone wants to admit.

Thank you for reading. Your Patreon support keeps our community entirely Ad free.