19. Project - First Week with Local AI

Project Overview

This project spans your first week with local AI. It's designed to build intuition through practical, daily use.

Day 1: Setup

Task: Get local AI running and have your first conversation.

1. Install Ollama (Chapter 8)
2. Pull a model (ollama pull llama3.2:7b)
3. Run it (ollama run llama3.2:7b)
4. Ask it to explain something you know well (verify quality)
5. Ask it to help with one simple task (draft an email, write a function)

Deliverable: Screenshot of your first conversation

Day 2: Explore Strengths

Task: Find where local AI excels for you.

Try these categories (pick at least 3):

1. Code: Write, explain, or debug a piece of code
2. Writing: Draft or edit a document
3. Learning: Explain a concept you want to understand better
4. Analysis: Summarize or extract information from a document
5. Planning: Help outline a project or task

Deliverable: Notes on which use cases felt most useful

Day 3: Explore Weaknesses

Task: Find where local AI struggles for you.

1. Ask something that requires current information (watch it fail)
2. Ask something complex that requires nuanced reasoning
3. Test a long conversation: does it lose the thread?
4. Try generating something highly creative (poem, story)

Deliverable: Notes on limitations you encountered

Day 4: Customization

Task: Make local AI work better for you.

1. Create a custom system prompt for one use case (Chapter 12)
2. Experiment with temperature: try creative vs. focused outputs
3. Try a different interface (LM Studio, Jan, or API)
4. Document what you changed and why

Deliverable: At least one custom Modelfile or system prompt

Day 5: Real Work

Task: Use local AI for something that matters.

Pick one real task:

• Draft a real email or document
• Debug real code you're working on
• Analyze a real document
• Prepare for a real meeting

Use local AI. Actually use the output (or part of it).

Deliverable: Document the task, what you used, and how it went

Day 6: Privacy Test

Task: Verify the privacy properties.

1. Try processing a document you'd never want to leave your machine
2. Check if your chosen interface has any network activity
3. Consider: what would need to be true for this to be compromised?

Deliverable: Assessment of whether local AI meets your privacy needs

Day 7: Reflection

Task: Synthesize what you've learned.

Answer these questions:

1. What surprised you about local AI?
2. What's still confusing or unclear?
3. What would you do differently?
4. Do you see local AI as a regular part of your workflow?
5. What do you want to learn next?

Deliverable: Written reflection (3-5 paragraphs)