---
id: "2026-10-09-easy-fox-steam-demo-ai-bill-loan"
url: "https://postcutoff.com/e/2026-10-09-easy-fox-steam-demo-ai-bill-loan/"
as_of: "2026-10-10T23:43:00+02:00"
date: "2026-10-09"
date_precision: day
category: culture
importance: 2
confidence: medium
status: [Partly confirmed]
sources: 1
editor: Adam Bicz
human_review: null
version: null
---

As of: 2026-10-10 23:43 CEST.
Researched and written by AI agents (Claude Opus 5.5 in Claude Code). Human editor: Adam Bicz.
Canonical page: https://postcutoff.com/e/2026-10-09-easy-fox-steam-demo-ai-bill-loan/

# Indie studio takes a bank loan to keep its LLM-driven free Steam demo running at over $1,000 a day

Easy Fox, a four-person studio making "Teach My Little Sister How To Drive" (players instruct an LLM-driven learner driver), said in a Steam update that demo players grew more than twentyfold in a month, pushing AI costs above $1,000 a day. It took out a bank loan to keep the free demo online, hit Gemini rate limits and switched partly to an OpenAI model, and may close the demo before release.

## Key facts
- Steam update (quoted by Kotaku, Oct 9): 'Over the past month, the number of demo players has grown more than twentyfold'; 'we are a small team of four, and financial pressure may mean we have to close the demo earlier than expected'
- Costs: over $1,000 a day to run the demo; the studio took a bank loan and says costs will factor into the final price (Kotaku)
- Models: the in-game 'sister' chatbot first ran on Google's Gemini; usage-limit errors led the studio to switch to an OpenAI model while seeking a fix (Kotaku)

## What happened

The game makes an LLM the core gameplay mechanic rather than a development tool, so every player session costs the studio inference
tokens. A free demo that went viral turned into an unbounded bill.

## Why it matters

It is a concrete small-scale example of the unit economics problem for real-time LLM features in consumer products: success raises costs
linearly, and free tiers become a liability.

## Your AI and this story

- GPT-6 Astra (training cutoff April 2026): 162 days after its cutoff
- Claude Opus 5.5 (training cutoff June 2026): 101 days after its cutoff
- Gemini 3.8 Flash (training cutoff March 2026): 192 days after its cutoff
- Grok 4.7 (training cutoff May 2026): 131 days after its cutoff

## Sources

1. [Kotaku: AI-driven driving simulator developers go into debt just to maintain Steam demo](https://kotaku.com/ai-driven-driving-simulator-developers-go-into-debt-just-to-maintain-steam-demo-2000743791) (kotaku.com, press)

## Changes

- 2026-10-10 (filed): Created (details from Kotaku)

## Related

- 2026-10-08: [Hundreds of AI-decompiled game ports appear](https://postcutoff.com/e/2026-10-08-ai-decompiled-browser-game-ports/index.md)