Learning from ChatGPT

#4557 Domain Specific Language for Prompting https://www.jetbrains.com/mps/ #4557  Define a domain specific language for LLM prompting Designing a Domain-Specific Language (DSL) for LLM prompting—let’s call it PromptScript —is an excellent way to move away from messy, unstructured text and toward predictable, reproducible, and structured AI interactions. A great prompting DSL needs to handle variable injection , role definition , few-shot examples , and output constraints cleanly. Here is a blueprint for a declarative, human-readable DSL designed for LLM workflows. The Core Blueprint: PromptScript 1. Keywords and Syntax @agent : Defines the persona, system instructions, and temperature. @context : Injects background data, documents, or variables. @examples : Groups few-shot input/output pairs. @execute : The main prompt instruction or user query. @schema : Enforces a structured output format (e.g., JSON, YAML). {{variable}} : Standard syntax for dynamic data binding. Example 1: Cu...

#4102 DSL; +Knowledge Graph Creator DSL https://gemini.google.com/app/9a560135499f22c3 DSLs require a compiler, right? How do I develop a simple/extensible "Hello World!" DSL? Yes =::= Would you like me to show you how to expand this into a "file-based" DSL where we read commands from a .txt file instead? List and describe 10 DSL's that I could quickly implement and test, based on my background and interests I wrote a report for extra credit in 1985 college class suggesting these types of DSL Don't remember, but could have been something like game development =::= Since you've had this idea since college, which specific "automation" task from back then would you have loved to see scripted this way? I can help you write the specific function for it. An application/DSL for today would be a knowledge graph creator that takes input from a GUI or file =::= Would you like to try adding a "Conditional" to this? For example, making it so the ...