diff --git a/docs/01-llm-intro/06-function-call.mdx b/docs/01-llm-intro/06-function-call.mdx
new file mode 100644
index 0000000..d3ded37
--- /dev/null
+++ b/docs/01-llm-intro/06-function-call.mdx
@@ -0,0 +1,6 @@
+---
+title: "工具调用(Function Call)"
+---
+
+
+sdfsdf
diff --git a/docs/02-prompt-engineering/02-chain-of-thought/03-concise-cot.md b/docs/02-prompt-engineering/02-chain-of-thought/03-concise-cot.md
new file mode 100644
index 0000000..fa8d449
--- /dev/null
+++ b/docs/02-prompt-engineering/02-chain-of-thought/03-concise-cot.md
@@ -0,0 +1,17 @@
+---
+title: "Concise Thought"
+---
+
+
+CoT 的长度通常是会影响整体时延,有的项目当中会直接把 CoT 的策略给删除,可是这个通常会影响模型的效果.
+
+此时你可以有两种策略:
+1. 增加你的训练数据
+2. 缩短你的 CoT 长度进而提升模型的效果.
+
+第一种方法就非常的简单粗暴，不过很可能会破坏你这个模型的通用 chat 能力，除非你的目的就是要构建一个垂类的不具备很强通用 chat 能力的模型。
+
+
+## 参考文章
+
+* [[1] Concise Thoughts: Impact of Output Length on LLM Reasoning and Cost](h是)
diff --git a/docs/03-agent-application/how-does-openai-o1-train.mdx b/docs/03-agent-application/how-does-openai-o1-train.mdx
new file mode 100644
index 0000000..bd843a8
--- /dev/null
+++ b/docs/03-agent-application/how-does-openai-o1-train.mdx
@@ -0,0 +1,9 @@
+---
+title: "OpenAI 的O1 模型是如何训练的"
+---
+
+或许训练的模式是这个样子：
+
+![](./imgs/strawberry-training-mode.gif)
+
+> 此图片来源于外网一位大佬
diff --git a/docs/03-agent-application/imgs/strawberry-training-mode.gif b/docs/03-agent-application/imgs/strawberry-training-mode.gif
new file mode 100644
index 0000000..4b9b86c
Binary files /dev/null and b/docs/03-agent-application/imgs/strawberry-training-mode.gif differ