The System Prompt Studio

Ravit Dotan
Sep 20
4 min read

The System Prompt Studio is a workspace that guides you through building system prompts, which are the foundation of custom chatbots and AI agents (read more about them here). The Studio is designed based on my approach for engaging with AI: prioritizing expanding your capabilities, supporting your agency and critical thinking, developing your skills, and incorporating AI ethics principles.

Go ahead and take it for a spin, or read more about how it works below. You have 10 free back and forths.

Inside the Studio

I think of the Studio as a place rather than an "assistant" or "companion". It is a space where you think and build.

The Studio contains five rooms, and each room offers specific activities that will help you develop your system prompt. Here is the floor plan:

Here are the activities the rooms offer:

Vision Room

The Vision Room is where every system prompt begins. Clarify what you want to create and why it matters to you. The activities in this room are:

Explore vision - Work through targeted questions that help you think deeper about your goals, motivations, and what success looks like
Choose name - Find a memorable name for your system prompt with guided suggestions and workshopping
Create a vision card - Build a structured template that captures your complete vision in an organized format you can reference throughout development

Building Room

The Building Room is where your vision becomes a system prompt. The activities in this room are:

Get Empty Template - Download a complete system prompt template you can fill out as you work through the building process
Select Scope - Decide whether to build your entire vision or focus on a specific part first - great for starting with manageable pieces
Draft Process Breakdown - Create the step-by-step workflow that becomes the heart of your system prompt - this is where your expertise gets captured
Draft Setup Definitions - Define the role, task, input format, and output format that frame how your prompt operates
Draft Behavior Guidelines - Establish responsibility guardrails, communication style, and safeguards that ensure ethical, professional operation

Testing Room

The Testing Room is where you evaluate the performance of your system prompt. The activities in this room are:

Run Basic Quality Check - Analyze your prompt for consistency, completeness, and clarity issues that could cause problems
Develop Test Cases - Create diverse scenarios that represent different ways people might use your system prompt, including edge cases
Execute Testing - Run live simulations where you can interact with your prompt step-by-step or see complete outputs in action

Revision Room

The Revision Room helps you find the root cause of problems in your system prompt and resolve them. The activities in this room are:

Get Revision Recommendations - Receive concrete suggestions for fixing specific problems you've identified, organized by priority and impact
Identify Problems and Solutions - Work through diagnostic questions when you're not sure what's causing issues - helps pinpoint root causes and develop solutions

Future Room

The Future Room helps you discover possible outcomes of using your system prompt, good or bad, by simulating future scenarios. You start by defining what aspects of your prompt's future impact you want to explore, then activate the Future Machine to hear from stakeholders across different possible futures. You can also process your insights into actionable improvements for your system prompt. The activities in this room are:

Set Intentions - Define what aspects of your prompt's future impact you want to explore - from responsibility concerns to alignment with your broader AI vision
Activate Future Machine - Experience immersive time-travel scenarios where you hear from multiple versions of stakeholders who have experience with your system prompt:
- Decide how far into the future you want to go (6 months, 2 years, etc.), who you want to talk to (users, colleagues, your future self, etc.), and what you want to ask them
- Hear what three versions of that person have to say—each coming from a different possible future with different experiences of your system prompt
- Follow up with any of the versions or move on to talk to other people
- Take notes throughout to capture insights for later processing
Turn Insights into Actions - Process what you learned from future scenarios and identify concrete improvements to make before deployment

Responsibility Guardrails

The System Prompt Studio incorporates several guardrails for responsible use, including:

No anthropomorphism - This chatbot presents itself as a studio with rooms rather than a type of person and encourages you to do the same in the system prompts you create. This practice Remind you that this technology is not a human-like entity and that we get to choose which metaphors we want to use to describe it.
Critical thinking - The Studio is designed to prevent over-reliance on AI tools and increase critical thinking. A key strategy is encouraging you to think first and prompt later. Activities typically begin by asking you to provide your own thoughts and then help you dig deeper. Another key strategy is using tentative language and provides multiple options for each decision. It uses phrases like "might" and "could be" rather than definitive statements, always asking "does this sound right?"

Bias - Imager is designed to avoid making assumptions about gender, race, age, location, or other characteristics, and instead prioritize representing human diversity thoughtfully

Privacy - You data remains yours. Imager runs on platforms that commit not to collect or train on user data.

About the Studio's creator

My name is Ravit Dotan. My approach to AI combines different parts of my background. Primarily these ones:

I'm a socially-oriented technologist. I build no-code AI tools to create the world I want to see, and I seek out people who want to do the same.
I'm an AI ethicist. I've worked with tech companies, investors, and local governments on responsible AI development, adoption, investment, and procurement. Practical responsibility is baked into everything I build and teach.
I'm a philosopher. I got my PhD from UC Berkeley, specializing in philosophy of machine learning, philosophy of science, and philosophy of knowledge. I've taught hundreds of people to think critically about worldviews and form their own.

All of this shapes how I think about and build AI. You can read more about me and follow my work on my blog and Linkedin page.