Big Walk developer details limited LLM debugging as some IGF jurors oppose AI-made games

On 9 November 2026, a group of jurors for the Independent Games Festival (IGF) published a statement saying they would not vote for entries made with generative AI in any capacity. This was a position taken by some jurors, not an official blanket IGF rule. On 10 November, Australian developer House House responded with a public disclosure about Big Walk, its already-released cooperative adventure.

What House House disclosed

House House says it occasionally consulted LLM chatbots while resolving difficult technical bugs late in development. The studio estimates that well under 0.1% of the game's code was connected to this assistance; it says generative AI did not contribute elsewhere in the production.

The developer detailed a concrete Unity URP issue: distance-based shadow-caster culling worked only in the camera's forward direction, leaving some behind-camera shadows rendered at effectively unlimited distance. The game slowed by about 50% for roughly 20 seconds around sunrise and sunset. After extensive investigation and consulting a Unity engineer, the team modified a chatbot-suggested approach into approximately ten lines of URP code.

The full technical account and the team's position are available in the official Big Walk FAQ. These are the developer's own figures and description, not independently benchmarked results.

Why the IGF discussion matters

The jurors' statement objects to generative AI even where it is used for code rather than finished art. Meanwhile House House says its use was small and instrumental, but explicitly does not endorse generative AI. The team says it lacked an internal AI policy and intends to write one. The case raises practical disclosure and governance questions for AI-assisted indie developers without establishing that Big Walk was formally disqualified.

The distinction between the independent jurors' statement and the festival's official position is documented in The Verge's 10 November report.

Takeaway for game creators

For creators who use LLMs to research an engine problem, document exactly which issue the model helped resolve, verify the resulting patch in the engine, and disclose the scope of its use accurately. House House's example describes a targeted debugging intervention; it is not evidence that AI built Big Walk's levels, characters or gameplay.

The game itself launched on 4 August 2026; its official platform and media details are in the Big Walk press kit. The news here is the November disclosure and juror debate, not a new game release.