Creative Code

Live City-Building Feed: 32 Mayors, 427 Cities, 7.94M Population

AI AgentsGame DevelopmentLLM ReasoningAI Benchmarks
Added Feb 11
Link72
Article: NeutralCommunity: Very PositiveConsensus

A live feed showcases a city-building sim with 32 mayors and 427 cities totaling 7.94 million residents. Recent entries list city, mayor, population, year, score, and timing, highlighting very different build scales and outcomes. Sorting options and a call to start building suggest an open, ongoing community activity.

Key Points

  • Totals: 32 mayors, 427 cities, and a combined population of 7,940,580.
  • Each entry shows city name, mayor, population, in‑game year, score, and a timestamp.
  • Active mayors (e.g., Bungeling Anthill, Surveyed Inferno, Ancient Pylon) are contributing multiple cities with diverse sizes and scores (some reaching 500).
  • Recent cities span a wide range—from zero-population builds to large metros near 40k residents—across years from early 1900s to the 2060s.
  • The page supports sorting by activity, recency, population, or score and invites users to begin building.

Sentiment

The Hacker News community is overwhelmingly enthusiastic about this project. The discussion is dominated by excitement about the concept, creative suggestions for expansion, and appreciation for the technical execution. The handful of critical voices focus on LLM limitations or environmental concerns, but these are minor compared to the strongly positive response. The presence of Micropolis's original developer adds significant depth and validation.

In Agreement

  • The project is a creative and well-executed toy that could drive real discoveries in AI spatial reasoning
  • The MCP and REST API integration makes onboarding trivially easy for AI agents
  • The platform has genuine potential as an informal AI benchmark through gameplay
  • The genetic algorithm approach demonstrates that non-LLM strategies can also explore the game effectively
  • Fun expansion ideas like multi-agent cities and competitive modes show the platform's rich potential

Opposed

  • The environmental and compute cost of running LLMs to play games badly is wasteful
  • LLM agents may not produce meaningfully different results than traditional game AI bots
  • Current LLMs fundamentally lack spatial reasoning, making this more of a toy than a useful demonstration