Holy sh*t... this is insane. 🤯 Someone found a way to use DeepSeek V4 Flash as subagents inside Codex. The result? They replaced a $200/month Claude setup with a much cheaper AI workflow. The setup reportedly takes just 3 minutes: → Install the “model-router” repo → Enable subagent models → Select your models → Done Instead of relying on one expensive model, developers are now building AI workflows where different models handle different tasks. The future of AI might not be one model doing everything... It might be teams of specialized models working together. 👀 Follow futurewalt.ai for more AI tools and updates. #ai #codex #deepseek #aitools