polish

Author: longxudou

content/blog/workflow/index.md

@@ -51,16 +51,21 @@ Unlike traditional REPL (Read-Execute-Print Loop) environments that focus solely
 </figure>
 
 
-This blog focus on **workflow** and **evaluation**. We detail the **on-policy annotation** and **SFT training** in https://terminal-agent.github.io/blog/annotation/.
 
-## What Makes Reptile Workflow Special?
+## What Makes Reptile Special?
 
-Compared with other CLI agents (e.g., Claude Code and Mini SWE-Agent), Reptile stands out because of:
+Compared with other CLI agents (e.g., Claude Code and Mini SWE-Agent), Reptile stands out for the following reasons:
 
 - **Terminal-only beyond Bash-only**: Simple and stateful execution, which is more efficient than bash-only (you don't need to specify the environment in every command). It doesn't require the complicated MCP protocol—just a naive bash tool under the REPL protocol.
-  > See [TTY-use Post](https://terminal-agent.github.io/blog/tool/) for more details on making terminal backend work.
+
 - **Human-in-the-Loop Learning**: Users can inspect every step and provide prompt feedback, i.e., give feedback under the USER role or edit the LLM generation under the ASSISTANT role.
-  > See [On-policy Annotation Post](https://terminal-agent.github.io/blog/annotation/) for more details on annotation and naive SFT training.
+
+
+This blog focus on **workflow** and **benchmarking**.
+
+> See [TTY-use blog](https://terminal-agent.github.io/blog/tool/) for technical details on how to make terminal backend work.
+>
+> See [on-policy annotation blog](https://terminal-agent.github.io/blog/annotation/) for annotation details on SWE tasks.
 
 
 ## Our Insights in Building General Agents
@@ -198,6 +203,8 @@ Reptile is open source and we welcome contributions! Whether you're interested i
 
 Visit our GitHub repository: [https://github.com/terminal-agent/reptile](https://github.com/terminal-agent/reptile)
 
+We are inspired by excellent community work such as [terminal-bench](https://www.tbench.ai/) and [mini-SWE-agent](https://github.com/SWE-agent/mini-swe-agent). We thank the community for their efforts and valuable insights!
+
 ## Conclusion
 
 The terminal has been humanity's interface to computers for 50 years. With Reptile, it becomes the interface between humans and AI agents.
@@ -228,6 +235,8 @@ If you find Reptile useful in your research or applications, please cite:
 }
 ```
 
+---
+
 > **Fun fact**: The name "Reptile" has a dual meaning: it refers to the REPL (Read-Eval-Print-Learning Loop) workflow in terminal interactions, and also pays homage to OpenAI's Reptile meta-learning algorithm (2018), which pioneered few-shot adaptation. Like its namesake, our Reptile learns to quickly adapt to new tasks—but through human-in-the-loop collaboration rather than pure algorithmic optimization. Both share the same philosophy: learning efficiently from minimal examples to master diverse tasks.
 >
 > *Reference: [On First-Order Meta-Learning Algorithms](https://arxiv.org/abs/1803.02999)*