Talking to Claude & ChatGPT
Share
The Art of Talking to
Claude & ChatGPT
Role assignment, prompt architecture, and why the way you write to AI will become the most important skill of the next decade.
The big picture
Two AIs. One future. Infinite possibilities.
We are living through a transformation in how humans interact with information, creativity, and knowledge work. At the center of this revolution are two AI systems that have captured the world's imagination: Anthropic's Claude and OpenAI's ChatGPT. But using them well is not just about asking questions — it is a discipline, an art, and increasingly, a competitive advantage.
This guide explores the technical foundations of prompt engineering — particularly role assignment — and why mastering these tools will separate those who merely use AI from those who truly harness it.
Meet the models
Claude vs ChatGPT: Core DNA
Both are large language models — but they have different philosophical foundations, training approaches, and ideal use cases.
Claude
by Anthropic · Founded by former OpenAI researchers
Built with a focus on safety, nuance, and Constitutional AI — a training method that teaches Claude to reason about ethics and values. Known for exceptionally long context windows (up to 200K tokens), thoughtful reasoning, and refusal to hallucinate when uncertain.
Long context (200K) Constitutional AI Deep analysis Nuanced writingChatGPT (GPT-4o)
by OpenAI · Backed by Microsoft
The model that started the public AI revolution in 2022. GPT-4o is OpenAI's current flagship — multimodal (text, image, voice, video), deeply integrated with tools like DALL·E 3, Code Interpreter, and real-time web browsing. Strongest for rapid tool use and creative generation.
Multimodal Web browsing Image generation Voice modeThe technical foundation
Understanding role assignment
Every conversation with an LLM is structured around three fundamental roles. This is not just a feature — it is the entire architecture of how these models process and respond to input. Mastering this unlocks a completely different level of control.
The three-role architecture
How every prompt is structured under the hood
System
Sets the AI's identity, personality, constraints, and context before the conversation begins. This is where you assign the role.
User
Your actual messages — questions, instructions, content to analyze. The "you" in the conversation turn by turn.
Assistant
The AI's response. Can also be pre-filled to guide the model's output style or force a specific starting pattern.
The system role is the most powerful tool you have. Most casual users never use it. Professionals who understand it can transform a generic AI into a specialized expert that remembers its purpose across an entire conversation.
In practice
How role assignment looks in Claude
Here's a real-world example of using the system role to create a specialized book recommendation expert — perfect for a bookstore like ours.
The Priory of the Orange Tree by Samantha Shannon will give you that same ache of a world that feels ancient and earned. You'll feel like an archaeologist uncovering a civilization, not a reader being told about one.
The surprising connection: both Rothfuss and Shannon are writing about the mythology of their worlds more than their plots — you're always aware that the story you're getting is a fragment of something much larger.
Read it on a grey afternoon with tea. It demands slow reading. ☕
How role assignment looks in ChatGPT
ChatGPT exposes this via its "Custom Instructions" feature in the UI — or directly through the API using the same system/user/assistant structure. Here's a developer view:
If Borges is your compass, then Italo Calvino's "If on a winter's night a traveler" is your destination. It does something Borges only gestures at — it makes you, the reader, the protagonist of a labyrinth about reading itself. Pair it with Umberto Eco's "The Name of the Rose" for a full weekend of gloriously disorienting fiction.
The key insight: The same three-role API structure powers both Claude and ChatGPT. If you learn prompt engineering for one, you can transfer that skill directly to the other. The philosophy is universal — only the personality of each AI differs.
Craft & technique
Anatomy of a world-class prompt
Great prompts are not longer — they are more intentional. Every element serves a purpose. Here is the structure of a prompt that consistently produces exceptional output:
The 6 pillars of expert prompting
Distilled from thousands of hours of AI interaction, these principles separate amateur prompting from professional-grade results.
Be the casting director
Don't just ask — assign. "You are a..." activates a specific mode of reasoning. The more specific the role, the more specialized the output.
Give the AI an audience
Telling the model who will read the output changes everything: vocabulary, tone, depth, and what to assume as prior knowledge.
Constrain to liberate
Paradoxically, adding constraints (word count, format, forbidden phrases) produces more creative and focused results, not less.
Show, don't just tell
Include one example of the output style you want. A single well-chosen example outperforms five paragraphs of description.
Chain your prompts
Break complex tasks into steps. First draft → critique → rewrite → format. Each stage builds on the last, with you in the editorial seat.
Iterate like a designer
The first response is a prototype. The best prompters never accept the first output — they use it to refine their request.
By the numbers
The scale of the AI moment
The revolution ahead
How AI is changing the world — right now
We are not waiting for a future where AI matters. That future is already here. These are the domains being transformed — and prompt engineering is the key that unlocks all of them.
Publishing & Literature
AI assists authors with research, developmental editing, and finding their voice. Not replacing writers — amplifying them. Literary agents now use AI to spot trends months in advance.
Medicine & Research
Claude has passed the US Medical Licensing Exam. AI systems are helping diagnose rare diseases faster than specialists, and accelerating drug discovery by years, not months.
Education
Personalized tutoring at scale. A student in rural Kenya with a smartphone now has access to the equivalent of a Socratic tutor — available 24/7, infinitely patient.
Law & Access to Justice
AI is democratizing legal knowledge. Contract review that cost $500/hour is now accessible to small business owners in minutes. The legal system is being fundamentally disrupted.
Software Development
GitHub Copilot and Claude Code are changing how software is written. Individual developers can now ship products that previously required teams — the solo founder era has arrived.
Language & Translation
Real-time translation of literary works, preserving cultural nuance better than ever before. Books that would never have been translated are now accessible globally.
Where we are
A brief history of the AI moment
ChatGPT launches and changes everything
Reaches 1 million users in 5 days. 100 million in 2 months. The fastest-growing consumer application in history. Every boardroom on earth immediately began rewriting their strategic plans.
GPT-4 and Claude 1 arrive
The jump from GPT-3.5 to GPT-4 was dramatic — passing bar exams, medical boards, and coding interviews. Anthropic's Claude launches simultaneously, emphasizing safety and nuanced reasoning.
The model wars accelerate
Claude 3 Opus, GPT-4o, Gemini Ultra. Each release reshapes what is possible. Context windows explode — Claude 3 can read an entire novel in a single prompt. Multimodality becomes standard.
Agents take over knowledge work
AI stops being a chatbot and starts being a collaborator. Autonomous agents complete multi-step tasks, browse the web, write and run code. The definition of a "knowledge worker" is fundamentally shifting.
Prompt literacy becomes the new literacy
Just as reading and writing unlocked the Renaissance, fluency in talking to AI will determine who thrives in the next economy. The question is not whether AI will matter — it is whether you will speak its language.
📖 Reading list
Understand AI more deeply — through books
The best way to truly understand what AI is, where it came from, and where it is going is still through long-form writing. These titles are available at everses.store:
"The Coming Wave" — Mustafa Suleyman
"Gödel, Escher, Bach" — Douglas Hofstadter
"Co-Intelligence" — Ethan Mollick
"Klara and the Sun" — Kazuo Ishiguro