同样省 token Codex 仍见底
Codex still burns quota faster
要点
- 同为两百美元档：Codex 两天见底，Claude Code 一周只用百分之十九
- 两边都用了 Agents.md、YAGNI/KISS，以及 PonyTail 和 RTK 省 token
- Codex 侧 Sol Medium 统筹加 Luna xHigh 干活，仍扛不住
- 有人指出 Luna 会制造错误，Sol 再返工，省下来的又吐回去
- 别只听官方说额度很宽，用同一工作法并排记账再下结论
原帖开头
Ran out of my Codex limits in less than two days. - Used GPT 5.6 Sol Medium as orchestrator - Used GPT 5.6 Luna xHigh as worker agent - Used PonyTail, RTK to saving on tokens. - Used updated Agents.md - Used YAGNI and KISS And yet, the limits ran out in less than 2 days. …
