<!-- mobian-agent-page publisher="dailydev" canonical="https://daily.dev/posts/pok-mon-red-s-most-brutal-obstacle-isn-t-a-gym-leader-it-s-a-doorway-that-broke-my-ai-afmadb4bs" -->

---
title: Pokémon Red's most brutal obstacle isn't a gym leader —...
description: An experiment running NousResearch's Pokemon-agent harness — a headless Game Boy emulator with a REST API — using both a local LLM (Qwen3.6-27B) and a cloud...
canonical: https://daily.dev/posts/pok-mon-red-s-most-brutal-obstacle-isn-t-a-gym-leader-it-s-a-doorway-that-broke-my-ai-afmadb4bs
twitter:card: summary_large_image
twitter:site: @dailydotdev
og:type: website
og:site_name: daily.dev
og:title: Pokémon Red's most brutal obstacle isn't a gym leader — it's a doorway that broke my AI | daily.dev
og:description: An experiment running NousResearch's Pokemon-agent harness — a headless Game Boy emulator with a REST API — using both a local LLM (Qwen3.6-27B) and a cloud...
og:url: https://daily.dev/posts/pok-mon-red-s-most-brutal-obstacle-isn-t-a-gym-leader-it-s-a-doorway-that-broke-my-ai-afmadb4bs
og:image: https://api.daily.dev/og/posts/AfmadB4Bs.png
og:image:alt: Pokémon Red's most brutal obstacle isn't a gym leader — it's a doorway that broke my AI
og:image:width: 1200
og:image:height: 630
og:locale: en
---

> ## Documentation Index
> Fetch the complete documentation index at: https://daily.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# Pokémon Red's most brutal obstacle isn't a gym leader — it's a doorway that broke my AI

**[XDA Developers](https://daily.dev/sources/xda-developers)** · 7 min read · 0 upvotes · 0 comments

## Summary

An experiment running NousResearch's Pokemon-agent harness — a headless Game Boy emulator with a REST API — using both a local LLM (Qwen3.6-27B) and a cloud model (GPT-5.6 Sol via Codex CLI) to play Pokémon Red. The local model hallucinated its location and got stuck staring at an in-game SNES for 20 minutes across 371 button presses. The cloud model performed better, picking up a starter Pokémon and reaching Viridian City, but then looped endlessly at a Pokémon Center door for 2,072 total actions. Both models failed at the same scripted transitions — doors, stairs, cutscenes — which turned out to be a bug in the agent harness: the collision map incorrectly reported door tiles as impassable walls. The local model's vision capability actually helped it bypass this bug once by trusting what it could see over the faulty map data. The conclusion: the models themselves weren't the bottleneck; the scaffolding was.

## Full article

daily.dev links to this article rather than hosting it. Read it at the original source: <https://www.xda-developers.com/let-my-local-llm-play-pokemon>

## Similar posts on daily.dev

- [Insights into Claude Opus 4.5 from Pokémon — LessWrong](https://daily.dev/posts/insights-into-claude-opus-4-5-from-pok-mon-lesswrong-v6f8uhfdx) · Hacker News · 4 upvotes · 0 comments
- [I Built a Pokémon Trainer AI With LangChain](https://daily.dev/posts/i-built-a-pok-mon-trainer-ai-with-langchain-6iazswiz1) · Atomic Spin · 1 upvotes · 0 comments
- [Someone wired up a digital fly's brain in a Pi 5 to play Pokémon, and you can help it progress](https://daily.dev/posts/someone-wired-up-a-digital-fly-s-brain-in-a-pi-5-to-play-pok-mon-and-you-can-help-it-progress-rdqqak8o5) · XDA Developers · 0 upvotes · 0 comments

---

Tags: [#llm](https://daily.dev/tags/llm), [#ai-agents](https://daily.dev/tags/ai-agents), [#local-ai](https://daily.dev/tags/local-ai), [#qwen](https://daily.dev/tags/qwen)

[View this post on daily.dev](https://daily.dev/posts/pok-mon-red-s-most-brutal-obstacle-isn-t-a-gym-leader-it-s-a-doorway-that-broke-my-ai-afmadb4bs)

```json
{"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]}
{"@context":"https://schema.org","@type":"TechArticle","headline":"Pokémon Red's most brutal obstacle isn't a gym leader — it's a doorway that broke my AI","url":"https://daily.dev/posts/pok-mon-red-s-most-brutal-obstacle-isn-t-a-gym-leader-it-s-a-doorway-that-broke-my-ai-afmadb4bs","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/pok-mon-red-s-most-brutal-obstacle-isn-t-a-gym-leader-it-s-a-doorway-that-broke-my-ai-afmadb4bs"},"datePublished":"2026-07-31T21:04:41.881Z","dateModified":"2026-07-31T22:54:46.561Z","description":"An experiment running NousResearch's Pokemon-agent harness — a headless Game Boy emulator with a REST API — using both a local LLM (Qwen3.6-27B) and a cloud...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/9a31a761370942181130aabda6537ff6?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/9a31a761370942181130aabda6537ff6?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"XDA Developers","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"XDA Developers","logo":"https://media.daily.dev/image/upload/s--kCg6yyAP--/f_auto,q_auto/v1774964407/logos/xda-developers?_a=BAMAMiWQ0","url":"https://daily.dev/sources/xda-developers"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/pok-mon-red-s-most-brutal-obstacle-isn-t-a-gym-leader-it-s-a-doorway-that-broke-my-ai-afmadb4bs","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"llm,ai-agents,local-ai,qwen","timeRequired":"PT7M"}
{"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"XDA Developers","item":"https://daily.dev/sources/xda-developers"},{"@type":"ListItem","position":3,"name":"Pokémon Red's most brutal obstacle isn't a gym leader — it's a doorway that broke my AI"}]}
```

