Overview
A series of experiments exploring how generative AI can become a genuine thinking partner—not just a tool that spits out text.
The Question
Everyone's using ChatGPT. But most people treat it like a search engine or a writing assistant. I wanted to explore: What happens when you treat AI as a collaborator in your actual workflow?
Experiments
Prompt Frameworks
I developed reusable prompt patterns for different thinking modes:
- Research mode — Systematic exploration of a topic with source tracking
- Devil's advocate — Force the AI to argue against my assumptions
- Synthesis mode — Combine multiple sources into coherent insights
- Decision support — Structured pros/cons with weighted criteria
Writing Workflows
Instead of asking AI to "write a blog post," I built multi-step workflows:
- Brainstorm angles (divergent)
- Select and refine (convergent)
- Outline structure
- Draft sections with specific voice guidelines
- Self-critique and revise
This produces writing that actually sounds like me, not generic AI slop.
Teaching Non-Technical Users
Helped teammates understand:
- What LLMs are good at (and bad at)
- How to prompt effectively
- When NOT to use AI
- Privacy and data considerations
Key Insights
The quality of AI output is directly proportional to the clarity of your own thinking.
AI doesn't replace thinking—it amplifies it. If your input is vague, you get vague output. If you're precise about what you want, AI becomes remarkably useful.
Mental Models That Helped
- AI as intern — Smart but needs clear direction and fact-checking
- AI as mirror — Reflects back the quality of your prompts
- AI as brainstorm partner — Best for generating options, not making decisions
Applications
This work influenced how I approach:
- Product research and competitive analysis
- Writing and content creation
- Career guidance tools (potential Jobcadu feature)
- Personal introspection and journaling
What's Next
Exploring how AI can support education and self-directed learning—helping people build mental models, not just find answers.
This is an ongoing experiment. The field moves fast, and so do my methods.