Module 6
ChatGPT in the Real World
Concrete workflows for real tasks. These are tested prompts that work specifically in ChatGPT — taking advantage of its strengths in image generation, voice, breadth, and the code interpreter. Where a technique differs meaningfully from what you'd do in Claude, that difference is noted.
Writing workflows
Writing
Breaking through writer's block with voice
Use Advanced Voice Mode. Say out loud:
"I need to write [article/essay/email] about [topic]. I have some ideas but I can't figure out how to structure it. Let me talk through what I'm thinking and I want you to help me find the core argument and a structure that would work."
Then just talk. Explain what you know, what you're unsure about, what you want the reader to feel. Let ChatGPT ask questions.
Voice removes the pressure of forming perfect sentences before you type. The act of talking through your ideas out loud — with ChatGPT asking follow-up questions — surfaces the argument faster than staring at a blank document.
Writing
Creating visual content alongside written content
I'm writing a blog post about [topic]. As we develop the content together, I'll ask you to generate header images for each section. For now, let's start with the introduction.
Introduction topic: [describe]
After we finalize the intro, generate a header image: professional, clean, relevant to the topic, suitable for a business blog, wide format (16:9), no text in the image.
ChatGPT's unique strength here — Claude can't do this. Writing and visual content in one workflow, with ChatGPT maintaining context about the article as it generates images that match the content.
Research workflows
Research
Current-events research synthesis
Search for the latest news and developments on [topic] from the past [week/month]. Then synthesize what you find into:
1. What has actually happened (factual summary)
2. Why it matters (significance and implications)
3. What's contested or unclear (what experts disagree on)
4. What to watch next (what developments to monitor)
Cite your sources for each major claim.
Combines ChatGPT's web browsing with its synthesis capability. Unlike Perplexity which just returns research, this produces a structured analysis you can act on — and the citation requirement keeps ChatGPT honest about its sources.
Research
Analyzing data files
[Upload your CSV or Excel file]
Analyze this data and tell me:
1. What are the 3 most important patterns or trends?
2. What anomalies or outliers should I investigate?
3. What does this data suggest I should do?
4. What important questions can this data NOT answer?
Then generate a chart showing [the most important trend you identified].
Code Interpreter runs the actual analysis — not predicted analysis. The "what can this data NOT answer" question is crucial and often skipped, but it prevents overconfident decisions based on incomplete data.
Coding workflows
Coding
Iterative debugging with real execution
[Paste your code]
This code is supposed to [describe what it should do] but instead it [describe the bug or unexpected behavior].
Run it, identify what's going wrong, fix it, run the fixed version, and show me both the original error and the confirmation that the fix works. Then explain what caused the bug in plain language.
Code Interpreter actually runs the code and sees the real error — not a predicted error. The fix gets verified before you see it. This is dramatically faster than the copy-paste-run cycle of debugging manually.
Coding
Using o4-mini for hard algorithmic problems
[Switch to o4-mini model]
I need to solve this problem: [describe the algorithmic or logic problem clearly with all constraints and requirements]
Work through this carefully. Show me your reasoning about the approach before writing any code. Then write the solution and verify it handles the edge cases: [list specific edge cases].
o4-mini's chain-of-thought reasoning handles algorithmic problems that GPT-4o stumbles on. Asking it to reason about the approach before coding catches flawed strategies early.
Business and productivity workflows
Business
Market research with web browsing
I'm researching [market/industry/competitor]. Search for current information and build me a competitive landscape that includes:
1. The major players and their positioning
2. Recent moves or announcements (last 6 months)
3. Gaps or underserved needs in the market
4. What a new entrant would need to compete effectively
5. The 2-3 biggest trends reshaping this space
This is for [describe your purpose — investment research, product planning, etc.]
The current information requirement is essential here — competitive landscapes go stale fast. ChatGPT's browsing gives you recent data that no static AI can provide, and the structured output format makes it immediately usable.
Productivity
Voice brainstorming on a walk
Use Advanced Voice Mode. Start with:
"I want to brainstorm [problem/opportunity/decision]. I'm going to talk through my thinking, and I want you to ask me sharp questions that push me further, challenge my assumptions, and help me find angles I'm missing. Don't just affirm what I say — push back when something doesn't hold up."
Then walk and think out loud for 15-30 minutes.
This is genuinely unique to ChatGPT. A 20-minute walking brainstorm with real-time AI pushback produces ideas that don't come from sitting at a desk. The movement and speech together unlock different thinking than typing does.
ChatGPT's specific workflow advantages
Notice what's unique about these workflows: image generation integrated into writing, voice-first brainstorming, real code execution, and current web data. These are things you genuinely cannot do in Claude. Use ChatGPT when the task benefits from these capabilities — and use Claude when you need depth, nuance, or honest pushback on complex analysis.