02. The Two Worlds - Cloud vs Local

Cloud AI: The Default Experience

When most people use AI today, they're using cloud-based services. ChatGPT, Claude, Gemini, and similar offerings share a common architecture:

You → Type prompt → Internet → Cloud server → Model processes → Response → Internet → You

What happens to your data:

1. Your prompt travels across the internet to the company's servers
2. The model processes it on hardware the company owns
3. The response travels back to you

The data policies vary: some services use your inputs for training by default, some don't, some let you opt out. But the fundamental architecture remains: your data goes somewhere else, gets processed on someone else's hardware, and the response comes back.

Advantages of cloud AI:

• No hardware requirements—you just need an internet connection
• Access to the largest, most capable models (hundreds of billions of parameters)
• Easy to use—no technical setup
• Consistent performance—servers are optimized for AI workloads

Disadvantages of cloud AI:

• Privacy: your data leaves your control
• Latency: round-trip time to server + processing time
• Dependency: service can change pricing, policies, or availability
• Cost: while some is free, capability often requires subscription
• Offline: requires internet connection

Local AI: Your Machine as the Server

Local AI changes the architecture. Instead of sending your data to a remote server, you run the model on your own hardware:

You → Type prompt → Local model (on your machine) → Response → You

The data never leaves your machine. There's no server, no internet transmission, no third-party policies.

Advantages of local AI:

• Privacy: your data never leaves your device
• Offline capability: works without internet
• No dependency: you own the model, it's yours indefinitely
• Cost predictability: one-time hardware purchase, then free use
• Customization: you can modify system prompts, parameters, and behavior

Disadvantages of local AI:

• Hardware requirements: you need a capable machine
• Model capability: limited by what your hardware can run
• Setup complexity: requires installation and configuration
• No "infinite" context: limited by your available RAM/VRAM

The Real Tradeoff

This isn't a binary choice. It's a tradeoff spectrum.

For casual questions, cloud is fine. For sensitive documents, local is the only option that guarantees privacy. The skill is knowing when to use which—and local AI gives you that choice.