FIXEDb350rel/studios-leadership3 min read
House House estimates LLMs involved in well under 0.1% of Big Walk code
House House says LLMs touched well under 0.1% of Big Walk's code during final debugging. The FAQ disclosure follows an Independent Games Festival dispute over AI submissions.

Patch notes
House House estimates well under 0.1% of Big Walk's code had anything to do with LLM use.
The cited example involved about 10 lines of code used to fix a problem with help from a Unity engineer.
House House used LLMs only during the final stretch of development, the studio says.
House House had no internal AI usage policy during development and calls that lack a failure.
Steam does not request disclosure of code or chatbot usage, according to House House.
By Elena Vasquez
House House estimates that well under 0.1% of Big Walk's code had anything to do with large language model (LLM) use during the game's final development stretch, according to a statement it added to the FAQ.
The estimate limits the reported use to a small fraction of the finished code. It does not resolve the debate that followed House House's decision to speak publicly.
How did LLM code reach Big Walk?
House House wrote: "In the final stretch of Big Walk's development, LLMs were used sparingly by House House to help work through difficult technical challenges and bugs." The statement says some model responses were copied as starting points, then modified into game code.
The studio cites about 10 lines of code used to fix a problem its team could not resolve manually or through direct contact with a Unity engineer. That places the reported use in debugging, not another part of game production. House House says generative AI had no role in any other context in making the game.
Why did the disclosure become an awards issue?
The sequence began with a statement from jurors participating in the Independent Games Festival awards. It noted a significant increase in games submitted with generative AI tools. The statement said the submissions had set off some fairly intense discussion and disagreement across the judge and jury pools.
The jurors said AI-generated code was not an exception. Their statement added: "Programming is a creative discipline whose labour deserves the same respect we extend to any part of making a game." It also addressed past work used to train models without the creators' consent.
The statement did not name a game. Aftermath reported that Big Walk drove much of the debate after House House's AI use became known. That overlap gave the discussion a specific target, even though the jurors did not identify one.
Big Walk's prominence amplified the revelation. The cooperative puzzler became a major hit this year and a strong contender in a relatively quiet year for video games.
House House had also kept the game off Xbox consoles in support of the pro-Palestinian Boycott, Divestment and Sanctions movement. That stance made its use of AI tools surprising to some observers.
What does this change for studios and teams?
For studios, the change begins with governance. House House says Big Walk was developed without an internal AI usage policy. The studio calls the omission a failure and says it will work toward a more comprehensive policy going forward. It also says its statement does not endorse AI tools or argue for their use, however minimal.
No deadline accompanies that commitment. The FAQ disclosure provides no industry threshold, no standard for measuring AI-assisted code and no review process for generated snippets.
Steam, by contrast, does not request disclosure of code or chatbot usage, according to House House. The developer says it posted its statement because the issue had become a matter of public interest. For other teams, that leaves voluntary disclosure as the practical trigger.
The mechanism is straightforward. Absent a platform request or a studio policy, teams can use AI-assisted code without listing that use on Steam. A written internal policy would give developers a separate point of control before any public statement.
House House's response does offer a workable sequence for other teams. It states the scope of tool use, estimates the affected code, identifies the debugging context and promises a policy. That sequence separates the operational fact of assistance from the governance decision about when to disclose it.
Percentages alone will not settle the debate. Even well under 0.1% represents production code moving through a development workflow. House House's own conclusion focuses on governance rather than its estimate.
House House has not set a deadline for the new policy. Its terms, and any further disclosure connected to the Independent Games Festival submissions, will show whether this remains a one-game exception or becomes a broader studio practice.
via pcgamer.com (Source)
More from Elena Vasquez
Show full bio
Staff writer covering regulation, policy and their impact on video game industry at Game Dev Wire.
248 articles