This Dev Scheduled ChatGPT to Auto-Update His Agent Configs With the Best Model Deals

JnBrymn · x · 2026-09-15

JnBrymn shares a practical setup: ChatGPT runs on a schedule, queries the Artificial Analysis API for model data, builds Pareto plots of intelligence vs. price per frontier lab, and then auto-updates his pi and omp configs.

The payoff: when codex hits a usage window limit, he immediately jumps to the best-value models in the upper-left of the plot. The skill is open-sourced as update-ai-model-frontier in arcturus-labs/agent-skills, with PRs welcome for other agent harnesses.

Original post →

More from coding & agent

coding & agent channel →