LAB 001
測試中ModelPilot
公開測試版:貼上一個 Codex 任務,即時得出建議使用的 model 與 reasoning effort。
- 問題
- 使用者往往不知道哪個 Codex model 與 reasoning effort 適合眼前任務。
- 目前假設
- 一次貼上、並說明理由的透明建議,足以在投入自動路由之前先驗證需求。
- 下一個決定
- 量度真實使用者會否採納建議,並再次回來使用這個工具。
01開始任務
02任務 context
03候選建議
04人手決定
用作測試在工作開始時降低選擇阻力的研究地圖: 開始任務 → 任務 context → 候選建議 → 人手決定