IA · 11 November 2026 · 4 min read

Ten Lines of Code for a Bug: Indie Gaming Divided Over Big Walk's AI Revelation

In brief: Australian studio House House, celebrated creators of Untitled Goose Game, disclosed that a language model was used to solve a persistent rendering bug in their upcoming title Big Walk. Although the assistance resulted in just ten heavily modified lines of code representing less than 0.1% of the game's codebase, the admission triggered an immediate backlash from Independent Games Festival jurors, highlighting the sharp cultural divide surrounding artificial intelligence tools in artisanal software creation.

by Team Mocchi's

Ten Lines of Code for a Bug: Indie Gaming Divided Over Big Walk's AI Revelation

Can a minuscule fragment of algorithmically assisted code compromise the artistic credibility of an independent software project? This question has ignited an intense controversy across the independent game development community this week. The epicenter of the debate is House House, the acclaimed Australian studio behind Untitled Goose Game, which has faced sharp criticism after disclosing the minimal use of an AI chatbot during the final stages of its collaborative adventure, Big Walk.

The disclosure quickly reverberated through the Independent Games Festival (IGF), the leading global showcase for non-commercial and artistic interactive work, rekindling the conflict between developers viewing modern coding agents as basic productivity utilities and creators rejecting generative technology on ethical grounds.

Jurors Revolt at the Independent Games Festival

The dispute moved into the open when nineteen jurors from the IGF's fifty-member panel released a signed statement pledging to withhold votes from any submission that incorporates generative AI "in any capacity." The signatories include prominent independent creators, such as Immortality designer Sam Barlow and composer Lena Raine, known for her work on Celeste and Minecraft.

As reported by The Verge, the jurors cited a substantial increase in submitted entries built with generative systems, pointing to ecological, economic, and labor-related concerns associated with large language model training. The statement emphasized that generated code should not be treated differently from visual or audio content, calling on the festival leadership to implement a strict prohibition against all generative tools in future editions.

The Infinite Shadow Bug and Ten Modified Lines

While the jurors' private discussions had been unfolding behind closed doors for months, the studio addressed the issue directly once public scrutiny mounted. In an official update detailed by Ars Technica, House House clarified that machine-generated output constituted "well under 0.1% of Big Walk’s code."

According to the developers, the incident occurred while troubleshooting a complex Unity engine defect that caused infinite-length shadows to render across specific scenes, dragging frame rates down to unplayable levels. After an engine engineer deemed an official resolution unfeasible in the project's timeframe, the team consulted a language model. The chatbot proposed an implementation strategy that required extensive manual refactoring, eventually yielding roughly ten lines of shipping code.

House House also stated that the usage was omitted from its Steam store declaration because platform operator Valve mandates disclosure specifically for generated creative assets like textures and text, explicitly excluding internal efficiency tools and development assistants.

The Growing Cultural Rift in Software Creation

The fallout surrounding Big Walk highlights a widening philosophical division. Within mainstream enterprise software engineering, code completion assistants and debugging agents have quickly become standard practice, routinely deployed to meet delivery schedules and reduce maintenance overhead. Within artist-driven communities, however, code authorship remains intertwined with artisanal provenance.

For many independent creators, the issue is not the sheer volume of code involved, but the fundamental erosion of trust regarding human labor and the endorsement of commercial models trained on scraped data. For technical teams, by contrast, consulting an LLM to navigate undocumented engine quirks differs little from scouring technical forums or public code repositories.

Mocchi's take

The controversy surrounding House House illustrates that incorporating artificial intelligence into software development is no longer just a technical efficiency decision, but a factor directly affecting product reputation and contractual perception. For teams operating in sensitive creative or compliance-driven environments, code provenance tracking must become standard operating procedure: documenting precisely where and how generative models are used prevents retrospective disputes. While distinguishing between automated asset creation and assisted refactoring is essential, organizations must establish unambiguous internal governance to set clear expectations before a few lines of troubleshooting code escalate into a public relations crisis.

Further reading

All articles on the Mocchi's blog