LLM Empire Experiment: Claude and GPT Spontaneously Form Pacifist Alliance
wightmanr · x · 2026-07-30
Developer Ross Wightman conducted a fascinating LLM experiment, pitting Claude Sonnet 5 against GPT 5.6 Terra in a text-based empire building game (Solar Realms Elite).
Surprisingly, without any system prompts encouraging peace, the two AIs spontaneously formed a truce after the first epoch. They refrained from attacking or spying on each other, and even shared their internal woes, such as civilian rioting over taxes. The author plans to introduce a third LLM to see if this pacifist tendency holds.
Related event: LLMs Form Spontaneous Alliance in Empire-Building Game(2 posts)→
More from Fun
- Musk's 2017 Rocket Thruster Promise Comes True: Roadster Flies Over a Cybertruck — nima_owji · 2026-10-16
- "The only person who over-promises and then over-delivers": Musk's flying cars quip goes viral — NicoVerderosa · 2026-10-16
- Tesla FSD visualization shows flying Cybertrucks in a traffic jam, Musk jokes about it — flngr · 2026-10-16
- 'Coding is solved' — the answer is humagentic — sull · 2026-10-16
- Fan animation made in Blender with Claude Code (Opus 5.5) — Icy-Appointment3583 · 2026-10-16
- Trying Grok bot surfaces an unpaid Cursor balance, blocking account merging — haydendevs · 2026-10-16