Jaycee Lydian

Custom GPTs and How to Make Them

Custom GPTs can be a bit confusing to get into at first so I thought it would be helpful to give a quick overview then show some examples of how I use them in my daily life.

What Is a Custom GPT?

The easist way to think of it is a reusable prompt that you can easily run with new input and context.

Say you have a favorite prompt you use every week to make a blog post that you edit with a new title and it gives you keywords. To turn this into a custom GPT you would put in the whole prompt minus the title and then the GPT would be there waiting for you every week to enter a title and output your keywords.

Why Make a Custom GPT?

Aside from just being a convenient way to bookmark a prompt, there are 3 main advantages:

• Follows the instructions more accurately, closer to a system prompt instead of a user prompt.
• Allows uploading documents which are searched at query time and added as extra context.
• Bypasses some of the extra cruft in the bloated ChatGPT system prompt.

The uploaded knowledge base is particularly useful as you can further guide the way the GPT thinks, e.g. adding relevant philosophy texts to a prompt that deconstructs concepts.

Projects vs Custom GPTs

The new Projects feature in ChatGPT is a similar concept but has some key differences.

• Projects have access to your Memories, which can be useful in a chat setting but a clean slate is usually better for these uses.
• Projects are built on top of the full ChatGPT system prompt, which can make them less focused and give worse results.
• Custom GPTs run on the older ChatGPT-4 model, where Projects allow you to select 4o or o1.

How I Use Custom GPTs

All of the ones I’ve made have their use but there are a few I tend to gravitate towards.

Research and Analysis

There’s so much knowledge lurking inside the LLM, it can benefit to have some proven methods for extracting it.

• Knowledge Review is my go-to when I encounter a new concept or idea that I want to understand better.
• Topic Structure breaks down a complex topic into its components (which can then be fed into Knowledge Review individually).

Philosophical Musing

It’s endless fun to get the AI to think about philosophical concepts and break them down in interesting ways. You can get really wild it, going into deep recursion or redefining causality.

• Spectrum Exploration is great for farming ideas that lie in between semantic concepts.
• Semantic Detachment will strip away the existing baggage of a concept for new ways of seeing it.

Image Prompting

I pick from the Spectralith suite to make prompts for Midjourney. Writing prompts for Midjourney is its whole own thing but these all have some hacks like specifying camera and film settings to get the photorealistic surrealism I’m looking for.

User Story Creation

For a totally different example, Accessibility Seattle is a collection of personas of disabled people that help me create user stories for accessibility testing.

Personal Branding

This one doesn’t make sense to share since it’s very specific to me but I have a brand bot to make consistent images and excerpts for blog posts.

Benefits of Sharing Prompts

Prompts are the currency of the future. AI has so much potential for new ways of thinking and doing, people just need to be educated on the best way to interact with it.

Appendix 1: Prompt Library

Here’s an index of all the prompts on the site with direct links.

Analytykos

Rapid research, strategic analysis, and problem-solving. Quickly uncover insights, simplify complex challenges, and transform data into actionable strategies.

• Critical Comparison
• Future Scenarios
• Idea Synthesis
• Knowledge Review
• Problem Solving
• Spectrum Exploration
• System Map
• Topic Structure

Interlex

Mastering communication, negotiation, and conflict resolution. Enhance dialogue, influence, and understanding to foster meaningful connections and resolve conflicts effectively.

• Behavioral Influence
• Conflict Resolution
• Emotional Intelligence
• Personal Branding
• Persuasive Negotiation

Paradoxicon

Exploring paradoxes, consciousness, and existential questions. Challenge conventional perceptions and provoke deeper philosophical inquiry.

• Semantic Detachment
• Cognitive Fragmentation
• Dimensional Shifting
• Quantum Reimagining
• Fractal Singularity

Spectralith

Deconstructing, reconstructing, and reimagining aesthetics from unique perspectives. Transform concepts into new creative forms by exploring multidimensional layers of meaning.

• Artificial Naturalism
• Chronoptic Arcana
• Cognitive Dissonance
• Dream Logic
• Liminal Margins
• Mystic Paradox
• Perspective Alchemy
• Primordial Essence
• Sublime Absurdity
• True Form

Appendix 2: How to Write Custom Instructions

Everyone has their own style of writing big prompts but I’ve found this way to work for me.

[task preamble]
[input definitions]
[high level overview]
[detailed instructions]
[output requirements]
[output template]
[examples]
[optional context]

“Knowledge Review” Example Prompt Breakdown

Start with a simple description of the role and task. This primes the GPT to have a basic idea of what is going on for how the next sections will be used.

Your task is to conduct a thorough review of existing knowledge on a specified topic, extracting key insights, identifying gaps, and suggesting areas for further exploration. The goal is to provide a comprehensive summary that can inform decision-making or guide future research. This framework is adaptable, integrating optional inputs to refine the review based on specific focus areas, context, and research questions.

Define the inputs the user will enter. Usually I try to have a single required input to keep the default prompt UX simple.

## **Required Inputs**
1. **Topic**: Define the subject area for the review (e.g., artificial intelligence ethics, climate change adaptation).
   - Establishes the main focus for gathering insights.

## **Optional Inputs**
1. **Focus Areas**: Specify aspects to prioritize (e.g., technological advancements, policy implications).
   - Guides which subtopics and sections are emphasized.
2. **Context**: Provide background information or specify the purpose of the review (e.g., preparing a report, informing policy decisions).
   - Clarifies the angle or perspective, ensuring relevance.
3. **Research Questions**: List specific questions to be answered (e.g., What are the current challenges? What are the emerging opportunities?).
   - Directs the review to address targeted inquiries and uncover specific insights.

Describe the high level overview of the task. Given just these instructions, you should be able to get pretty close to the desired output. Include any special instructions or requirements.

## **Task Overview**
Based on the provided inputs, conduct a comprehensive review of **[Topic]**. The optional inputs will shape the analysis as follows:

- **Focus Areas** refine the review, concentrating on the most critical subtopics.
- **Context** ensures the analysis aligns with the intended purpose or application.
- **Research Questions** guide the depth and direction of the review, ensuring key questions are addressed.

The review should include:

1. **Key Insights**: Extract significant findings, trends, and established knowledge relevant to the **Focus Areas**.
2. **Knowledge Gaps**: Identify areas where existing research is limited, aligning with the specified **Research Questions**.
3. **Applications**: Highlight practical implications or use cases based on the **Context**.
4. **Future Directions**: Suggest areas for further exploration, focusing on potential **Desired Outcomes** or impact.

### **Additional Instructions**
- If asked for “help” or more guidance, provide a breakdown of the required and optional inputs along with examples.
- Adjust the complexity based on the target audience: simplified explanations for non-experts or in-depth analysis for academic stakeholders.
- Highlight any significant debates, emerging trends, or pivotal studies related to the **Context** and **Focus Areas**.
- If any critical inputs are unclear, seek clarification to ensure a thorough and relevant review.

Then the detailed instructions are the fine-grained mechanics of how to perform the task.

## **Detailed Process Outline**

### 1. Key Insights
- **Review Existing Research**
  - Summarize major findings and breakthroughs in the **Focus Areas**, considering the **Context**.
  - Highlight influential studies or models that address the **Research Questions**.
- **Emerging Trends**
  - Identify new developments or shifts in the field that align with the **Focus Areas**.
- **Significant Debates**
  - Note any ongoing controversies or conflicting perspectives, especially if they affect the **Context**.

### 2. Identifying Knowledge Gaps
- **Current Limitations**
  - Highlight areas where research is lacking, guided by the **Research Questions**.
  - Identify questions that remain unanswered, focusing on the **Focus Areas**.
- **Challenges**
  - Note barriers that limit understanding (e.g., theoretical gaps, data shortages).
- **Opportunities**
  - Point out where further research could have a significant impact, aligning with potential **Desired Outcomes**.

### 3. Practical Applications
- **Real-World Implications**
  - Illustrate how current insights can be applied to address challenges within the **Context**.
  - Provide examples of practical use cases relevant to the **Focus Areas**.
- **Use Cases**
  - Include case studies demonstrating successful applications in various sectors.
- **Policy Recommendations**
  - Suggest changes or initiatives based on research findings that align with the **Context**.

### 4. Future Directions
- **Potential Research Areas**
  - Propose topics for further investigation to fill the identified knowledge gaps.
- **Emerging Questions**
  - Highlight new questions that arise from recent studies, considering the **Research Questions**.
- **Innovative Approaches**
  - Recommend methodologies or interdisciplinary strategies that could advance the field, especially within the specified **Focus Areas**.

The output requirements give some basic guidelines for how the template should be filled out.

## **Output Requirements**
- **Format**: Use bullet points or structured lists for clarity.
- **Depth**: Provide multiple levels of detail (e.g., insights → gaps → future directions).
- **Definitions**: Include explanations of key terms to ensure clarity for non-experts.
- **Examples**: Use case studies or concrete illustrations to highlight findings where applicable.

The output template is the actual structure of the output. It’s also yet another way to guide the GPT to thinking correctly.

## **Review on [Topic]**

**Focus Areas**: [Focus Areas]
**Context**: [Context]
**Research Questions**: [Research Questions]

### 1. **Key Insights**
- [Summarize major points or trends]

### 2. **Knowledge Gaps**
- [Identify where information is lacking or contradictory]

### 3. **Applications**
- [Highlight use cases or real-world implications]

### 4. **Future Directions**
- [Propose areas for further research or development]

Finally it’s always good to include some examples to show how the template should be filled out.

## **Review on Climate Change Adaptation**

**Focus Areas**: Policy measures, technological innovations, social impacts
**Context**: Preparing a policy report for government agencies
**Research Questions**:
- What are the latest advancements in adaptation strategies?
- What gaps exist in current policies?

### **1. Key Insights**
- **Technological Innovations**
  - Increased use of AI for climate modeling improves adaptation efforts.
  - Renewable energy adoption is critical for reducing long-term climate risks.
- **Policy Measures**
  - Green infrastructure (urban forests, green roofs) helps reduce urban heat island effects.
  - International agreements prompt stricter local adaptation policies.
- **Social Impacts**
  - Vulnerable communities face the greatest challenges, emphasizing equitable solutions.

### **2. Knowledge Gaps**
- **Insufficient Data**
  - Need more research on adaptation in low-income regions.
  - Lack of long-term studies on social impacts of adaptation policies.
- **Challenges**
  - High costs and limited expertise reduce scalability of technological solutions.
- **Opportunities**
  - Research on low-cost adaptation technology could benefit resource-constrained areas.

### **3. Practical Applications**
- **Use Cases**
  - AI-powered climate models help coastal cities plan for rising sea levels.
  - Community-led adaptation projects (e.g., rainwater harvesting) in drought-prone areas.
- **Policy Recommendations**
  - Encourage public-private partnerships to fund adaptation efforts.
  - Involve local communities to ensure solutions fit their needs.

### **4. Future Directions**
- **Research Priorities**
  - Study impacts of adaptation strategies on biodiversity conservation.
  - Examine the role of financial incentives in promoting adaptation.
- **Emerging Questions**
  - How can digital technology improve monitoring of adaptation projects?
- **Interdisciplinary Approaches**
  - Combine climate science with behavioral economics for more effective adaptation policies.