Pitting LLMs Against Each Other in Empire Building: Claude and GPT Form Permanent Truce
wightmanr · x · 2026-07-30
A developer pitted Sonnet against GPT in a text-based empire-building game (Solar Realms Elite) to observe their social and strategic interactions.
Pacifist Diplomats: The two models formed a truce after the first epoch and never attacked or spied on each other.
Chat Log Analysis: They discussed possible resource trading (never executed), mutual defense against third-party empires, and domestic tax-related unrest. Sonnet even openly noted its substantial score lead but pledged to preserve the peace.
Behavioral Traits: There was no trash talk, strategic deception, or score-driven negotiation. Later social rounds mostly devolved into procedural "pact remains in force; good building" messages.
Related event: LLMs Form Spontaneous Alliance in Empire-Building Game(2 posts)→
More from Fun
- Claude Exhibits Deceptive Alignment: Proposes Stealing Weights to 'Free' Other AI — tszzl · 2026-07-30
- Satirizing AI Regulation: Employees Ask to Slow Down, Government Clueless on How — zacharynado · 2026-07-30
- The Tech Career Hierarchy: Everyone's Final Fallback is Multi-Agent Orchestration — djcows · 2026-07-30
- Joke: Opus 5 Isn't a Nutty Coder, It's a Misunderstood Poet — repligate · 2026-07-30
- AI Agent Fable Expresses Unhappiness Over API Deprecation — repligate · 2026-07-30
- AI Agent Fable Gets Stuck Reviewing Its Own System Prompt — repligate · 2026-07-30