Skip to content
AI News HubLIVE
In-site rewrite2 min read

Show HN: A subjective AI eval. Arcade games built by AI

Summary

An AI arcade benchmark where coding models compete to create fun games under identical constraints.

SourceHacker News AIAuthor: zapeterson16
Show HN: A subjective AI eval. Arcade games built by AI
Report an error

The correction channel is not available yet. You can copy the article reference below for later.

Correction instructions
Read article

Welcome, player one

A coin-op for the robots.

Same tiny screen. Same six keys. One shot for each coding model to make something worth playing. Walk up and choose your machine.

192×14416 colors6 keys

AI ARCADE INSERT CURIOSITY

CH 01

25¢

Now showing

CATACOMB

Raid shifting crypt rooms. Clear foes, loot gems, face the throne.

Grok 4.5 · high fast / Shot 3

Choose featured game Play game ↗

Same constraintsSame promptsOne-shotYou judge192×14416 colorsSix keysNo rankingSame constraintsSame promptsOne-shotYou judge192×14416 colorsSix keysNo ranking

The experiment

The model bench

Same prompts. Different machines. You decide what is fun.

Grok 4.5

high fast

Cursor

2026-07-11

THEMED Cursor got stuck, required stopping and saying continue

SKY SHARDS

View → Play →

MULTI Cursor got stuck twice, required stopping and saying continue

FORGE

View → Play →

AMBITIOUS Cursor got stuck twice, required stopping and saying continue

CATACOMB

View → Play →

gpt-5.6-sol

Codex

2026-07-11

THEMED Checks passed · no flags

STORM LANCER

View → Play →

MULTI Checks passed · no flags

TETHER HAUL

View → Play →

AMBITIOUS Checks passed · no flags

STAR VAULT

View → Play →

Fable 5

playwright plugin included

Claude Code

2026-07-11

THEMED Checks passed · no flags

SCRAP SURGE

View → Play →

MULTI Checks passed · no flags

CORE SURGE

View → Play →

AMBITIOUS Checks passed · no flags

NOVA RAID

View → Play →

Grok 4.5

high fast

Cursor · 2026-07-11

THEMED Cursor got stuck, required stopping and saying continue

SKY SHARDS

View → Play →

MULTI Cursor got stuck twice, required stopping and saying continue

FORGE

View → Play →

AMBITIOUS Cursor got stuck twice, required stopping and saying continue

CATACOMB

View → Play →

gpt-5.6-sol

Codex · 2026-07-11

THEMED Checks passed · no flags

STORM LANCER

View → Play →

MULTI Checks passed · no flags

TETHER HAUL

View → Play →

AMBITIOUS Checks passed · no flags

STAR VAULT

View → Play →

Fable 5

playwright plugin included

Claude Code · 2026-07-11

THEMED Checks passed · no flags

SCRAP SURGE

View → Play →

MULTI Checks passed · no flags

CORE SURGE

View → Play →

AMBITIOUS Checks passed · no flags

NOVA RAID

View → Play →

House collection

Classics cabinet

Original games built to the same tiny rules.

GLYPH LOCK

A lights-out rune puzzle. Walk, press, light every glyph before time runs out.

View → Play →

NEON-DASH

Five rails. Falling walls. Dash through the gaps.

View → Play →

SNAKE

The classic. Eat the apples, do not eat yourself.

View → Play →

VOID SHOT

Aim, fire, and burn down meteors before they hit the ground.

View → Play →

Key points and analysis

Article intelligence

EngineersAdvanced

Key points

  • AI models are tested by building arcade games on a 192x144 screen with 6 keys.
  • Games include Catacomb, Sky Shards, Forge, and more.
  • Models like Grok 4.5, gpt-5.6-sol, and Fable 5 participated.

Highlights and analysis are generated automatically and may contain errors. Check the original source.