Prompt Engineering For Science

Human-AI collaboration workflows for scientific tasks.

[Warning: under construction.]

Philosophy

• Be open: Describe what you need, not how to do it. Instead of "use Quantum ESPRESSO for DFT", ask "I want to simulate a hydrogen molecule—what are the approaches? Compare methods and suggest the best tool." Expect solutions beyond your assumptions.
• Manage context: Keep relevant files in your workspace—source code, examples, templates. Reference them with @file to ground AI responses.
• Make a demo: For batch tasks, polish one example first. Then ask AI to process the rest following that standard.

Flows

Category	Flows
Paper Writing	Literature → Plan → Draft → Polish
Coding	Build · Debug · Explain
Teaching	Slides · Quiz · Homework · Exam · Evaluation
MCP	Build AI tool servers

Universal Workflow

1. Human prepares → materials, context
2. AI generates → drafts, suggestions
3. Human reviews → quality control
4. Iterate → "fix", "simplify", "continue"