I don't know how your openclaw (lobster) robot training is going?
However, my Uncle Shrimp is already able to work very stably and has produced:
1. Every day, I will report global hot topics in the fan group (morning and evening news)
2. Report on the market information of mainstream cryptocurrencies in exchanges
3. Back up all working directories of openclaw at 2 AM every night, and keep data for nearly 14 days.
4. The system performs self-checks at a scheduled time every morning, conducting various safety checks and reporting any issues to me promptly.
Additionally, besides having Uncle Shrimp help me with daily tasks, I have also created an agent: Uncle Li (the character played by Ge You in 'No Man's Land'; Uncle Li gets very angry which leads to serious consequences 😄)
I have to praise Uncle Li, the current working style is simply too satisfying for me, no wonder those big companies are frantically laying off employees, cutting so many product managers, R&D engineers, and testing engineers.
Yesterday was about 2 hours, through Uncle Li I completed the requirement analysis, functional design, and database design for my quantitative trading system, and started gradually writing code. Moreover, with my continuous training, every step of his work will have very clear node status information: (This system considers external compatibility, designed an adapter main module, can interface with mainstream exchanges and various APIs of on-chain DEXs)
For example:
------------------------------------------------------------
Current status
• ✅ The twelfth step of the project has been completed
• 🔄 I am now closing the thirteenth step
• ⏳ If you approve, I will proceed to advance with the minimal query capability
• ❌ If you do not approve, I will not touch the file
------------------------------------------------------------
And it will also report the current task percentage in a timely manner, for example:
------------------------------------------------------------
Progress report for the twelfth step of the project
Current overall progress
8%
───
Completed
• ✅ The task boundaries have been locked
• ✅ SQLite solution has been determined
• ✅ The minimal table structure has been determined
• ✅ The target files in the plan have been finalized
───
In progress
• 🔄 Ready to start the actual implementation of the file:
1. sqlite_store.py
2. pipeline.py connecting to SQLite
3. single_trade.py
------------------------------------------------------------
Of course, the capability of the Codex model is already top-notch, but I heard from @大赌哥 that the Copilot's autopilot model is easier to use. Once I finish using Codex this month, I will give it a try.
By the way, let me explain why I specifically have an intelligent agent Uncle Li to complete this quantitative trading design, because I am worried that he would pollute the overall memory of my main control intelligent agent Uncle Xia, so I let him be isolated, Uncle Li can always roll back, also I want to tell everyone, as I delve deeper into shrimp farming, I am surprised to find that they are really smart, and the copy that comes out is basically the result I want, without such a strong AI flavor, this is something that any pure GPT tool cannot compare to, because he has memory...
Later, I also plan to train an intelligent agent specifically for drawing... stay tuned...
Alright, enough small talk, I will continue to write code... and continue to raise lobsters