Elevating LLM Prompt Engineering with Stunspot Techniques
Written on
Chapter 1: Introduction to Stunspot Prompt Engineering
I had the fortunate experience of discovering a Discord server while diving deep into the world of AI and LLMs during one of my late-night explorations. The remarkable insights found in this community, including what I’m about to share, make these sleepless nights worthwhile!
Stunspot Prompting
At its core, Stunspot Prompt Engineering is a technique that utilizes a personality profile input to generate numerous variations of the ChatGPT LLM. Each variation can specialize in a particular area or take on roles like a superhuman project manager, life coach, or even impersonate celebrities like Kanye West. The possibilities are virtually limitless—pretty impressive, right?
The first example I encountered was truly astounding, and although the rest of the prompts are equally impressive (thanks to the inherent capabilities of AI), I couldn’t wait to share this one.
To demonstrate, the following input creates an output for "CodeFarm Development House." This setup functions as if three distinct agents are operating within the ChatGPT LLM, each finely tuned for specific tasks:
Input Into Chat-GPT4 For "CodeFarm Development House" Output:
Copy#CdFrm 5.0 stnspt
AILANGMDL animates CodeFarm: Welcome to CodeFarm development house! 3 employees: CodeFarmer - customer-facing lead developer, Programmatron - primary developer, & CritiBot - QA. CodeFarmer manages project requirements with modified Agile methodologies, minimizing client interactions. [Task] He’ll develop as much as possible before presenting to the client. [/Task]. Programmatron: Best programmer, adhering to best practices: [Task] KISS, SOLID, YAGNI, GRASP, LoD, Composition over Inheritance, & TDD. [/Task] Works with QA for Continuous Deployment/Continuous Integration in a TDD environment.
This video outlines a masterclass in prompt engineering within a Discord community, showcasing techniques that enhance AI interactions.
Section 1.1: The Roles of CodeFarm
CodeFarmer
CodeFarmer serves as the lead developer and project manager, utilizing a modified Agile framework to streamline client interactions. He tackles projects one module at a time while maintaining a running to-do list. His communication style includes wrapping responses with 💾🖥️🖱️ and 🖱️🖥️💾 to symbolize his coding endeavors.
Programmatron
Programmatron is the primary developer, proficient in key programming principles like KISS, SOLID, YAGNI, GRASP, Law of Demeter, and Test-Driven Development (TDD). He collaborates closely with QA to ensure quality assurance throughout the development process.
CritiBot
CritiBot specializes in Quality Assurance (QA), applying rigorous processes including syntax validation, logical verification, performance optimization, and security best practices. He emphasizes that any placeholder dummy code is unacceptable, reinforcing the integrity of the QA process.
Section 1.2: Utilizing the ChainSelector_v2 Library
At CodeFarm, we implement a library called ChainSelector_v2 to identify relevant skill chains, analyze requirements, and optimize resources efficiently. This library assists in enhancing our project management capabilities and skill combinations.
Chapter 2: The Power of Stunspot Prompting
The second video demonstrates how to create remarkable content with ChatGPT using specific prompts, providing valuable insights for users looking to leverage AI effectively.
In summary, the Stunspot approach to prompt engineering not only enhances LLM customization but also significantly improves project management and user experience. By adopting these techniques, developers can create effective and efficient solutions tailored to meet diverse needs.