I had Claude, Codex, Grok and Deepseek build games from 15 years of old Humble Bundle assets
slickriptide · reddit · 2026-11-17
The author ran an "anti-slop challenge": giving top coding models 2.7GB of dusty Humble Bundle game assets (35 packs, 545 animations, 95 tracks) with minimal prompting to see what each builds.
- Claude: from one sentence of instruction, produced Underkeep, a fantasy platformer drawing on 6 asset packs, in 15 minutes using 20% of quota. A "clever puzzle" turned out to be a bug.
- Codex: asked to do something different, built a fun pirate-tavern Liar's Dice bluffing game in 8 minutes; after Claude and Deepseek critiqued the pirate AI for never bluffing, Codex improved the game feel.
- Grok: handicapped by running in Chat Build mode without direct asset access, produced a poor custom board game.
Takeaway: Claude impressed most from a bare prompt, multi-model critique loops improved results, and direct asset access matters enormously.
More from coding & agent
- Amanda Launches API That Sends AI Agents to Meetings for You — Scobleizer · 2026-11-17
- Building 3D Worlds With Claude: The Case for a 'Signature Demo' You Keep Upgrading — Kyrannio · 2026-11-17
- GBrain's hosted memory layers personal, team, and company brains across ChatGPT, Claude, and Cursor — garrytan · 2026-11-17
- Build a Multi-Agent System From Scratch ships chapters 9-10 on subagents and A2A — _nerdai_ · 2026-11-17
- Break-even math for 20-turn agent loops: caching and routing cut API burn 76% — TokenMarginDev · 2026-11-17
- New cloud coding agents spin up isolated microVMs with per-task access control — jekapats · 2026-11-17