I let 6 ai models trade real money for 48 days straight and the results are not what anyone predicted
Moon dev: "i gave the exact same data to openai xai anthropic google moonshot and deepseek and one of them did something the others couldn't even come close to" "48 days on a
CLI has been way better for me in terms of productivity and token usage
Especially when using the free Zen models, DeepSeek V4.1 Flash, and GLM 5.3 Flash on Go. The one thing I still haven't figured out is the best way to organize my environment. I use OpenCode + Pi a lot,
This solves the problem of GPT-6 limits.
This setup uses OpenCode + Union Alpha with DeepSeek V4.1 Flash subagents to run cheaper AI workflows. Setup: 1.Install the union-crew repo. 2.Type /uno-crew <task>. 3.Start building.
You have reached the end of the archive
All of deepseek