Four LLM mayors, one deterministic city: open-source loop-design experiment wins CoreWeave Hacks top prize
hackgoofer · x · 2026-09-16
Solo developer Bereket Belachew won the top prize (Best Loop Design, with Weights & Biases) at CoreWeave Hacks with four-mayors, a fully open-source experiment now on GitHub.
The setup:
- A deterministic city (seven state numbers, eight policy levers, a 20-year term) is governed by an LLM mayor.
- The inner loop (year by year: observe the city, act) is identical for every mayor. Only the outer loop — what carries across the term boundary — differs, isolating loop design as the variable.
- Same city, same model, same seed: watch which city grows.
- Ships with a paper page, three graders, full run traces, and figures computed from run files by script.
A clean controlled experiment on agent memory and outer-loop design, useful for anyone building long-horizon agents.
More from coding & agent
- free-claude-code: open-source project with 55k stars runs Claude Code on free models from 50 providers — goyalshaliniuk · 2026-09-16
- Agent Flywheel guide: exhaustive markdown plans, beads tasks and agent swarms to cut token waste — doodlestein · 2026-09-16
- A tiered approach to token budgeting when developing with agent swarms — doodlestein · 2026-09-16
- Doctor builds WebMCP-enabled DICOM viewer so Codex can read scans and write reports — iamrobotbear · 2026-09-16
- Designer's bet: the agent with the best computer-use model will be the 10x unlock — _AustinCalvert_ · 2026-09-16
- Meta's Muse agent security blueprint: freedom to act, not to rewrite permissions — TheTuringPost · 2026-09-16