How we build prompts
Prompt Methodology
1. Profession mapping
We start from the real workflows of working professionals — daily, weekly, and quarterly tasks specific to each role. Each profession is mapped to a taxonomy of activities (writing, analysis, communication, leadership, planning, learning, etc.).
2. Prompt drafting
Drafts combine AI-assisted generation with human prompt engineering. Every prompt is structured with a role, context, inputs, constraints, and output format so it works reliably across major LLMs.
3. Testing across models
Prompts are tested against multiple LLMs to verify they produce useful, on-brief output. We discard prompts that are too generic, fragile to phrasing, or that consistently hallucinate.
4. Editorial review
A human reviewer checks each prompt for clarity, professional appropriateness, safety considerations, and alignment with our editorial standards.
5. Versioning & updates
Prompts are updated as models evolve and as we receive user feedback. Outdated or underperforming prompts are revised or removed.
6. Limitations
No prompt is universally optimal. Output quality depends on the model used, your specific inputs, and your editorial judgement. Always review output before use.