Creating Gems in Gemini #2

Building High-Impact Gems with the First Principles Workflow

Avoid Zombie Mode

Most people use AI like zombies. They lurch around, throwing generic requests into a Large Language Model (LLM), and accept whatever "homogenized smoothie" of information comes out.

The problem? LLMs are trained on the entire internet, and a lot of the internet is inaccurate or low-quality. If you don't ground the AI, you get generic, hallucinated, or skewed results.

Our goal?
To create consistent, high-quality, brand-specific outputs by using Gems (custom AI personas) grounded in First Principles Thinking.

Part 1: The Two Types of Gems

To organize our work, we categorize our Gems into two distinct buckets:

1. Generic / Prototyping Gems (The Tools)
   • What they are: These are "Prompt Builders." They don't know about a specific client yet. They are designed to help us generate high-quality instructions.
   • Use case: A "Master Prompt Builder" that helps you structure a request for any client.
2. Brand-Specific Gems (The Workers)
   • What they are: These are tailored for specific clients (e.g., "Current Daily"). They are pre-loaded with the client's "brain."
   • What’s inside them:
     • Personas & Content Pillars.
     • Unique Value Propositions (UVP).
     • Audience Profiles.
     • Background Research.
     • Style Guides.

Part 2: The "First Principles" Workflow

We never just "write a prompt." We follow a specific sequence to ensure quality.

Step 1: The First Principles Analysis

Before creating a final output, we run the request through a "First Principles" Gem.

• The Goal: Strip away assumptions and get to the core truth of the project.
• The Process: Ask why are we doing this? What is this product at its absolute essence?
• The "Child Test": The output must be simple enough to explain to a six-year-old. If you can't explain it simply, you don't understand it (and neither will the AI).
• Why we do this: This acts as a filter to remove the "internet noise" and grounds the AI in reality before we start drafting marketing copy.

Step 2: The Master Prompt Build

Once we have the First Principles analysis (the clear, simple truth), we feed that information into a Master Prompt Builder.

• Input: The "First Principles" breakdown + specific project requirements (e.g., "Build a website brief").
• Output: A sophisticated, structured prompt that defines the role, objective, and success metrics.

Step 3: Execution (The Brand Gem)

We take that Master Prompt and finally apply it to the Brand-Specific Gem.

• This is where we add the client's specific context (keywords, previous newsletters, specific documents).
• Result: A highly tailored output that sounds exactly like the brand, but is structurally sound because it was built on First Principles.

Part 3: Organization & Storage

To stay efficient, we don't let prompts float around randomly. We organize Gems and prompts into specific functional folders:

• Foundation: Core brand elements (Research, Positioning, Voice).
• Operations: Day-to-day tasks (Job posts, SOWs).
• Search: SEO specific (Schema, Meta tagging, GEO playbooks).
• Content: Strategy for blogs, newsletters, and outreach.
• Web Development: Briefs and coding tasks.

Note: If a prompt is generic, it lives in the main folder. If it is for a specific client (like "Lambit"), it lives in that client’s specific marketing folder.

Summary

• Don't be a zombie: Stop blindly prompting. Ground your work in reality.
• First Principles is key: Break every problem down to its non-negotiable components. If you can't explain it to a child, start over.
• Structure leads to freedom: By using a rigid workflow (First Principles >> Master Prompt >> Brand Execution), we gain the freedom to create highly creative, accurate work.
• Centralize knowledge: Store specific client data (personas/pillars) in the Gem so you don't have to paste it in every time.