House House disclosed a 10-line AI-assisted fix — 19 IGF judges reject games using AI

Your game is slowing down because of an engine bug. The team is stuck, consulting a specialist has not helped, and a chatbot suggests an approach that works after some manual reworking. The fix takes ten lines. For the developer, the problem is solved. For some festival judges, that is enough to refuse to vote for the game.

That is precisely the situation facing House House, the creator of the cooperative game Big Walk. The story is useful for any team using AI in development: agreeing internally on how to use a tool is not enough. You also need to understand the requirements of platforms, competitions and the people who influence a product's promotion.

How ten lines sparked a dispute

According to House House, the team encountered a Unity bug in the final stages of development: in certain situations, the engine generated shadows of practically infinite length, slowing the game down. The studio says it consulted a Unity engineer, who considered a fix impossible.

The developers then asked a language model. It suggested a possible solution that required substantial reworking. Ten lines of code made it into the game. This is the studio's own account: it does not establish that the model independently diagnosed the problem and produced a ready-to-use, verified patch.

House House estimates that all code associated with such LLM consultations amounts to significantly less than 0.1% of the codebase. That figure describes the scale of use as estimated by the developers. It says nothing about the fix's quality, its reliability or the number of hours saved.

There is also an important technical caveat. The story does not prove that the bug could not have been fixed through conventional means. The source article also mentions an assessment to the contrary. The practical outcome is more modest: after consulting a chatbot and reworking its response themselves, the team found an approach that worked for them.

The storefront and the jury have different expectations

Nineteen members of the Independent Games Festival jury, which has more than fifty members, signed a statement refusing to vote for games that use generative AI in any way. They also called on the festival to introduce a corresponding ban in future years. Their stated reasons include the effects on labor, the economy and the environment.

This is the position of a group of judges. The statement itself does not mean the festival has already disqualified Big Walk or adopted the proposed blanket ban. That distinction matters when assessing the consequences: individual members refusing to vote and official exclusion from the competition are different decisions.

House House explained that it had not disclosed the use when releasing the game on Steam because Valve does not require disclosure of this kind of chatbot and code use. The studio declared the AI assistance when submitting the game to IGF, then made its explanation public. The dispute is detailed in this Ars Technica article.

This creates a problem for product teams: meeting one platform's requirements does not guarantee that other industry participants will consider the disclosure sufficient. Players may never see the model's output on screen, but the production process itself can still matter to the audience.

What teams should decide before their next release

The disagreement comes down to what counts as acceptable use. Some participants in the debate distinguish between technical advice and the large-scale generation of graphics or music. Others consider rejecting the technology itself a matter of principle, regardless of how much it contributed. The percentage of code does not resolve this dispute: the two sides are judging different things.

A simple AI usage log can help a team: what problem they were solving, what the model suggested, what went into the product and who checked the result. Such a document helps answer questions from a publisher or competition organizer accurately. Without it, even a good-faith disclosure can easily become a collection of vague recollections just before a deadline.

Before submitting, teams should separately check the wording covering code, content and development tools. If the rules are ambiguous, it is better to obtain written clarification in advance. Technical review of the patch remains the team's responsibility: public approval of a tool does not validate a fix on the developer's behalf.

The Big Walk case illustrates a specific risk: a small change can affect access to professional recognition. That means the decision to use a chatbot for a difficult bug affects both development and future communication about the product.