Code-Generation with LLM

• Generates code to do task
  • Writes test cases
  • Debugs itself by using errors as prompt for next iteration
  • Asks for user approval
  • Generate applet name/filename based on prompt
• Virtual environment creation with dependencies/whatever
• Prompting user for more information if needed

App-ify

• Save generated code to a folder so that it can be run again at any time
• For every "applet" in the folder, make a fucking button
  • When you press the button, it runs the thing that you made

Sharing

• If you create an applet that you like and think other people would find useful, you can upload to a "social media" platform
• Others can download it and use it for their own purposes

• The bottleneck is LLM capability
  • GoRILLA API LLM
• Because generated code is saved locally, it's personal and secure
• Voyager skill-building type shit
• For sharing: make a GH account for the application, every applet gets its own repository

Prompt

• Text prompt for some automatable task
• "Write a program to set an alarm and text Zara 'yo what did i miss in class'"
• THE ONLY LIMIT IS YOUR IMAGINATION

graph LR
prompt --> gpt
gpt[GPT] --> ec[Error Correction] 
ec --> gpt 
ec --> proto[Skill prototype] 
proto --> human[User feedback] 
human --> gpt 
human --> skill[Finalized skill] 
skill --> lib[Library] 
lib --> GUI