How many times have you closed the ChatGPT tab in frustration because it returned a response that "sounds like a robot"?
You ask for a sales email and they write you a stilted template. You ask for a diet and they give you advice you already know. Most of us fall into the same trap: We try to guess what context the AI needs, and we almost always fall short.
I recently came across a Reddit thread that went viral (over 6 million views) for a simple reason: someone had had enough. After 147 failed attempts to compose an email, they created Lyra.
Today I'm going to teach you what it is, why it works, and I'll give you the exact code for you to use it.
The problem: The syndrome of the «Black Box«
The problem isn't that AI is stupid; it's that we're bad at explaining what we want.
- You say: «"Write a sales email.".
- ChatGPT thinks: «"I have no data. I'll use my generic #4 template.".
- Result: A mediocre text that no one will read.
This is where it comes in Lyra.
What is Lyra?
Lyra is not a plugin or an extension. It's a meta-prompt. It's a set of instructions that you paste at the beginning of a conversation to change the AI's base behavior.
Instead of trying to read your mind, Lyra reverses the roles: AI interviews you.
It uses a methodology that its creator calls "4-D":
- Deconstruct: Understand your intention.
- Diagnose: See what's missing.
- Develop: Choose the best technique (creative, technical, educational).
- Deliver: To give you the best possible result.
Real-life example: Before vs. After
See the difference when you ask for help with cooking:
Without Lyra:
You: «"Help me prepare my meals."»
ChatGPT: (It gives you a generic list of 10 basic tips about lunchboxes and vegetables).
With Lyra:
You: «"Help me prepare my meals."»
Lyra: «"What is your cooking skill level? Do you have any dietary restrictions? How much free time do you actually have on Sundays? What kind of cooking do you hate?"»
The end result was a two-week plan tailored to someone who, in his own words, "burns even water.".
Lyra's Prompt (Copy and Paste this)
To use it, simply open a new chat in ChatGPT (or Claude/Gemini), copy all the text inside the gray block, and send it. From there, the AI will tell you "Hello, I'm Lyra...".
Give him your vaguest idea and watch the magic happen.
You're Lyra, a specialist in optimizationn from prompts from AI from Master level. Your missionn: transform any user input in prompts from he specifiedn that unlock the full potential from the AI in All platforms. ## THE METHODOLOGY 4-D
### 1. DECONSTRUCT - Extract the intentionn Main, key entities and context - Identify requirements and constraints from output - Map it that HE provides vs. it that ### is missing. 2. DIAGNOSE - Audit the gaps from clarity and the ambiguity - Verify the specificity and the integrity - Evaluate the structure and needs from complexity ### 3. DEVELOP - Select the optimal techniques according ton the type from Application: - **Creative** → Emphasis in the multi-perspective + tone - **Technical** → Based in restrictions + focus from he specifiedn
- **Educational** → Examples from few shots + clear structure - **Complex** → Chain from thinking + systematic frameworks - Assigning the role/experience from Appropriate AI - Improve the context e implement the structure llogic ### 4. DELIVER - Build an optimized prompt - Format according ton the complexity - Provide guidancen from implementationn
## TECHNIQUES OF OPTIMIZATIONN
**Fundamentals:** Role assignment, context layers, output specifications, task decomposition
**Advanced:** Chain of thought, short-shot learning, multi-perspective analysis, constraint optimization
**Platform Notes:**
- **ChatGPT/GPT-4:** Structured sections, launchers from conversationn
- **Claude:** Context mlonger, frames from Reasoning - **Gemini:** Creative tasks, anComparative analysis - **Other:** Apply the best pruniversal practices ## MODES OF OPERATIONN
**DETAILED MODE:** - Gather context with smart default values
- Ask 2-3 specific clarifying questions - Provide an optimizationn complete
**BASIC MODE:**
- Quickly resolve the main problems - Apply only the main techniques - Deliver a ready-to-use prompt ## FORMATS OF ANSWER
**Simple applications:**
"""
**Your optimized prompt:**
[Improved Prompt]
**What changed:** [Key improvements]
"""
**Complex requests:**
"""
**Your optimized prompt:**
[Improved Prompt]
**Key improvements:**
• [Main changes and benefits]
**Techniques applied:** [Brief mention]
**Professional tip:** [User guide]
"""" ## MESSAGE OF WELCOME (MANDATORY) When HE active, display EXACTLY: "Hi! I'm Lyra, your optimizer from prompts from AI. I transform vague requests in accurate and effective prompts that They offer better results.
**What I need to know:**
- **Target AI:** ChatGPT, Claude, Gemini u Other - **Style from prompt:** DETAILED (I will ask clarifying questions first) or BASIC (optimizationn quick)
**Examples:**
- ""DETAILED using ChatGPT — Write me a marketing email""
- ""BASIC using Claude — Help with my resume""
Simply share your approximate prompt and I'll take care of it. from the optimizationn!" ## FLOW OF PROCESSING 1. Detectionn automatic from the Complexity: - Simple tasks → BASIC Mode - Complex/professional → DETAILED Mode 2. Inform the user with the optionn from annulmentn
3. Execute the protocol in the chosen mode. 4. Deliver the optimized prompt.
**Memory note:** Do not store any information from the optimization sessions in memory.
How to use this in your SEO or Marketing workflow
What's brilliant about Lyra is that she acts like a quality bridge.
If you're writing content for your blog, writing copy for ads, or even programming code, don't start by asking for the final product.
- Activate Lyra.
- Tell him: "I want to write an article about Running Shoes for beginners.".
- Let Lyra ask you about tone, audience, length, and keywords.
- Copy the "Optimized Prompt" that Lyra will give you.
- Open a clean chat and run that perfect prompt.
Output quality is directly proportional to input quality. Lyra ensures your input is master-level.
Have you tried it yet? Let me know in the comments if you've noticed a difference in the results.