Lesson 2A — Building a Research Workflow (Research OS)

1.0 Why This Matters¶

Modern analytical work is increasingly workflow-driven.

In practice, research and analytics involve:

• collecting data,

• testing ideas,

• revising prompts,

• debugging code,

• interpreting results,

• documenting decisions,

• and revisiting earlier work.

Without organization, students often:

• lose useful prompts,

• forget what worked,

• repeat mistakes,

• restart workflows,

• and struggle to reproduce results later.

2.0 AI Changes Workflow¶

AI tools can:

• generate code,

• explain concepts,

• summarize information,

• and accelerate experimentation.

However, AI also increases the importance of:

• verification,

• documentation,

• and organization.

3.0 The Research OS Idea¶

In this course, we gradually build a lightweight:

3.1 “Research Operating System” (Research OS)¶

The goal is not perfection.

The goal is to create:

• continuity,

• organization,

• and reflective workflow habits.

The Research OS acts as:

• a memory system,

• a workflow system,

• and a research notebook.

3.2 An Example Directory Structure¶

Create a folder called:

ResearchAI/

Suggested structure:

ResearchAI/

├── researchAI_brain.md
├── personal_log.md

├── prompts/
│   ├── useful_prompts.md
│   └── failed_prompts.md

├── datasets/
│   ├── raw/
│   └── cleaned/

├── notebooks/
│   ├── class01/
│   ├── class02/
│   └── class03/

├── exercises/
│   ├── class01/
│   ├── class02/
│   └── class03/

├── projects/

│   └── capstone_project/
│       ├── project_brain.md
│       ├── data/
│       ├── notebooks/
│       ├── outputs/
│       ├── references/
│       └── draft/

└── archive/

We will start slow and build up later on a need basis.

4.0 The Two Core Files¶

4.1 researchAI_brain.md¶

This is your:

• evolving knowledge system,

• workflow notebook,

• and idea repository.

It may include:

• useful prompts,

• code snippets,

• workflow ideas,

• debugging notes,

• AI limitations,

• and important lessons.

Think of this as:

“What do I want my future self to remember?”

Example: researchAI_brain.md¶

# Current Focus

# Important Lessons

# Useful Workflows

# AI Limitations Observed

# Good Prompt Patterns

# Mistakes to Avoid

# Interesting Datasets

# Research Ideas

# Questions to Explore

# Weekly Reflection

4.22 personal_log.md¶

This is your:

• reflection journal,

• experiment log,

• and learning diary.

You can record:

• what you learned,

• what confused you,

• what worked,

• what failed,

• and what you want to explore later.

Example: personal_log.md¶

# Lesson 2 Reflection (2026-05-13)

## What I learned
- how to inspect missing values,
- and why EDA matters before modeling.

## What confused me
- interpreting skewed distributions.

## What tools I explored
- DataMiner
- Orange

## Prompt experiments
- “Suggest 5 EDA checks for customer transaction data.”

## Things I want to explore later
- interactive plots,
- customer segmentation.

## Challenges

## Reflection

(OPTIONAL)

4.3 Example: useful_prompts.md¶

Useful EDA Prompt¶

Suggest:

• useful summary statistics,

• possible missing value checks,

• and appropriate visualizations

for a dataset containing:

• customer transactions,

• dates,

• categories,

• and revenue.

5.0 Prompting as Structured Reasoning¶

Good prompts usually include:

• the task,

• relevant context,

• desired output,

• and constraints.

5.1 AI Workflow Example¶

A typical workflow in this course might look like:

question
→ prompt
→ AI-generated code
→ run code
→ inspect output
→ debug/revise
→ interpretation
→ reflection
→ save useful workflow

This process is iterative.

Researchers constantly revise:

• prompts,

• assumptions,

• code,

• and interpretations.

6.0 Versioning and Iteration¶

Good analytical work evolves over time.

Do not expect:

• first prompts,

• first code,

• or first interpretations

to be perfect.

7.0 Responsible AI Use¶

AI can accelerate:

• coding,

• explanation,

• brainstorming,

• and experimentation.

However:

• AI can hallucinate,

• generate incorrect code,

• misinterpret outputs,

• or produce misleading explanations.

Humans remain responsible for:

• verification,

• interpretation,

• and judgment.

8.0 Reflection Questions¶

1. Why does workflow organization matter more in the AI era?

2. What kinds of things should be saved in a prompt log?

3. Why is reflection useful in analytical work?

4. What might happen if analytical workflows are poorly documented?

9.0 Homework¶

1. Create your:

   • ResearchAI/ folder,

   • researchAI_brain.md,

   • and personal_log.md.

2. Add:

   • one useful prompt,

   • one reflection,

   • and one workflow insight.

3. Write a short reflection:

What kind of workflow system do you think would help you most this semester?