Day 3 · Friday3 hours
Privacy, Self-Hosting & Capstone
Protect yourself, own your AI, and ship your project.
By the end of this session you can
- Explain what an API key is and how leaks happen — and store secrets safely.
- Decide what you should never paste into a public chatbot, and why.
- Run a model fully offline on your own laptop with Ollama or LM Studio.
- Weigh local vs. cloud trade-offs and manage token cost.
- Ship and present your capstone, and give structured peer feedback.
Topic 1
The dangers of API-key and credential leaks
What a key is, how it leaks, and how to store secrets so it never happens to you.
- How leaks happen: screenshots, public repos, shared configs.
- The financial and security fallout of a leaked key.
- Safe storage: environment variables, secret managers, and never in chat.
Topic 2
The dangers of centralized AI data collection
Who sees your prompts, and what you should never paste into a public chatbot.
- How conversations may be used for training — and the settings that matter.
- Redaction habits and safe defaults for sensitive data.
Topic 3
How to self-host AI models
Running a capable model on your own hardware, fully offline.
- Ollama and LM Studio as easy local runners.
- GPU vs. CPU, and quantization for smaller memory.
- Choosing a model your hardware can actually run.
Topic 4
Private and decentralized AI
The movement toward user-owned, private inference.
- Morpheus AI and decentralized inference.
- Local vs. cloud trade-offs: privacy and control vs. power and convenience.
Topic 5
Cost management and staying current
Keeping the bill predictable and your skills fresh after the class.
- Token pricing, subscriptions vs. pay-as-you-go, and avoiding surprise bills.
- Where AI is heading — agents, on-device AI, open-weight parity — and how to keep learning.
Practice
Friday is protect-yourself-then-ship day. Run a model fully offline, do your personal AI security audit, then finish and present your capstone.
Hands-on lab
Offline model + AI security audit
- 1Run a model fully offline on your own laptop and chat with it — no cloud.
- 2Do a personal AI security audit: find where your keys/secrets live.
- 3Lock them down and set the privacy toggles on your accounts.
Peer exercise
Capstone peer evaluation
- 1Score each capstone against the shared rubric (1–5 per criterion).
- 2Give one written strength and one written suggestion per project.
- 3Close with a feedback circle: one breakthrough, one struggle, one next step.
Essay· One page, post-class
My personal AI playbook
- 1Write the tools, habits, and privacy rules you are keeping after the class.