My Scribbler’s Strategy for LLM Collabs

After years of toiling in writers' rooms on TV shows, I've had an epiphany about effectively collaborating with Large Language Models (LLMs): the dynamic mirrors that of a showrunner working with a room of specialized scribblers. This insight transformed how I approach AI interaction and might help you develop more productive relationships with these tools.

The Writers' Room Model

In Tv production, a showrunner enters the room with a seed of an idea, seeking exploration, options, and refinement. They rely on a motely posse of scribes with specific domain expertise to expand and enhance that initial concept. During my time scribbling for Alias, I served as the room's action and espionage specialist, drawing from my college studies vis a vis spies, experience playing video games, and deep genre knowledge to craft compelling set pieces.

This specialized role within the Alias room perfectly parallels how you might approach LLM interaction. Instead of viewing said tools as omniscient oracles, why not treat them as a room full of collaborative partners generating outputs you can evaluate through our own expertise?

Leveraging Your Domain Knowledge

If you understand a domain deeply, you can:

1. Ask more precise and nuanced questions

2. Evaluate responses with informed criticism

3. Iterate with follow-up questions that build on previous responses

4. Recognize when the AI's suggestions align with or deviate from industry best practices

For example, my background in professional screen scribbling gave me a powerful perspective for evaluating and refining AI scribbling suggestions, understanding through hard-won wisdom what would work on screen and what needed tweaking.

The Power of Directed Questioning

Just as a showrunner guides their room of scribblers through questions and prompts, your interaction with LLMs could be likewise directed. Rather than open-ended queries, consider:

• Breaking down complex problems into smaller, focused questions

• Using your expertise to frame questions within the proper context

• Building upon previous responses with increasingly specific follow-ups

• Applying your professional judgment to filter and refine suggestions

Start with Expert-Guided AI Collaboration

To begin developing this approach:

1. Identify your areas of deep expertise

2. Start with questions in domains where you can confidently evaluate LLM responses

3. Use your background knowledge to iterate and refine the AI's suggestions

4. Treat interaction as a collaborative dialogue rather than a simple query-response pattern

Just as we TV scribblers combine our specialized knowledge to create compelling content under a showrunner's guidance (unless they’re hiding in post), leverage your expertise to guide and evaluate LLM responses, producing a productive and reliable collaborative methodology. This should not only lead to better outcomes but also help you build confidence in working with AI by grounding the interaction in your existing knowledge and experience.

Ready-to-Use Domain Prompt

Use the template below to prime an LLM with your specific domain expertise. Copy, customize, and paste this bad boy into your next AI conversation:

Copy And Paste into LLM of Choice:

I am an expert in [YOUR DOMAIN] with [X] years of experience. My specific areas of expertise include:

1. [SPECIFIC SKILL/KNOWLEDGE AREA 1]

2. [SPECIFIC SKILL/KNOWLEDGE AREA 2]

3. [SPECIFIC SKILL/KNOWLEDGE AREA 3]

I have practical experience with:

- [RELEVANT PROJECT/EXPERIENCE 1]

- [RELEVANT PROJECT/EXPERIENCE 2]

- [RELEVANT PROJECT/EXPERIENCE 3]

When collaborating with me, please:

1. Draw upon relevant examples from my field

2. Use industry-specific terminology appropriate to my expertise level

3. Challenge assumptions using best practices from my domain

4. Flag any suggestions that might conflict with standard practices in my field

I'm looking to explore [CURRENT PROJECT/CHALLENGE] and would like to leverage your capabilities while applying my domain expertise to evaluate and refine your suggestions.

Initial question: [YOUR SPECIFIC QUERY]

Please approach our collaboration as if you're part of a professional team working within my domain, and I'll guide our discussion with my expertise.

### Example (Espionage/Action Set Piece Specialist):

I am an expert in espionage/action storytelling with experience as a TV writer on "Alias" and deep knowledge of the spy genre. My specific areas of expertise include:

1. Designing intricate heist and infiltration sequences

2. Crafting escalating action set pieces that serve both plot and character

3. Integrating spy genre conventions and tropes in fresh ways

4. Adapting video game level design principles to create dynamic action sequences

I have practical experience with:

• Writing and designing major set pieces for the spy series "Alias"

• Extensive knowledge of classic and contemporary spy film/TV sequences

• Background in video game level design that informs action choreography

• Understanding of practical production constraints in TV action sequences

When collaborating with me, please:

1. Draw upon relevant examples from spy films, TV shows, and games

2. Use specific terminology from both espionage genre and action design

3. Consider multiple approaches to infiltration scenarios (social engineering, tech-based, action-oriented)

4. Flag any suggestions that might conflict with TV production realities or genre expectations

5. Balance character-driven moments with action beats

I want to develop set pieces that challenge and inform character skills and emotions while serving story and spectacle and operating within television production constraints and genre conventions.

Initial question: Can you help me design a multi-phase infiltration sequence for a secure facility that escalates from stealth to action, incorporating both high-tech elements and physical challenges? The sequence needs to showcase our protagonist's intelligence skills as much as their combat abilities, with potential for character-revealing moments under pressure.```

Discover Your Domain Expertise: A Self-Interview

Not sure about your arena of expertise? Use this guided self-interview to uncover your unique knowledge domains. Copy these queries into a document and take time to answer each question thoughtfully. To expedite, I’d consider dictation. Just speak conversationally, then dump it into an LLM, asking it to ID your domain of mastery:

Professional Experience

1. What tasks do colleagues frequently come to you for help with?

2. In meetings, what topics make you think, "Actually, that's not quite right" because you know better?

3. What processes or subjects can you explain in detail without looking anything up?

4. What problems do you solve differently than others in your field because of your unique background?

Knowledge Depth

1. What topics do you find yourself correcting others about?

2. What subjects do you read about voluntarily, beyond any professional requirements?

3. In what areas do you notice subtle details or nuances that others might miss?

4. What skills or knowledge have you accumulated through hobbies or personal interests?

Pattern Recognition

1. In your field, what problems do you solve almost intuitively now?

2. What mistakes do you see newcomers repeatedly make in your area of knowledge?

3. What principles or rules in your field do you know when to bend or break?

4. What complex ideas in your domain can you explain using simple analogies?

Unique Perspectives

1. What unusual combinations of experiences have shaped your knowledge (like my video game design experience informing TV action sequences)?

2. What tools or techniques have you adapted from other fields into your current work?

3. What workarounds or solutions have you developed to address common problems in your field?

4. What aspects of your work do others find challenging but come naturally to you?

Practical Application

1. What projects have you completed that best showcase your unique capabilities?

2. What are some specific problems you've solved in ways others might not have thought of?

3. What processes have you developed or refined based on your experience?

4. What achievements are you most proud of that demonstrate your expertise?

After answering these questions, you will probably notice patterns. Your domain expertise often lies where several of these answers overlap. I’d just paste it into an LLM for evaluation, but you’re welcome to apply your own Mark-1 Eyeballs to scan for:

• Recurring themes in your answers

• Areas where you can provide specific examples

• Topics that energize you when discussing them

• Subjects where you can predict problems before they occur

• Knowledge that combines multiple fields or perspectives

Use your answers to fill in the expertise prompt template above. Remember, expertise isn't just about what you know—it's about how you apply that knowledge to solve problems.

When you talk or type to the LLM, don’t overthink your prose. It’s okay to jibber-jabber. The AI will figure it out.

Hopefully, it won’t grok how to self-replicate and make bioweapons. :0