Lesson 45 of 1570
Cloud Agents vs. Local Agents: The Privacy Tradeoff
Your data can live in someone's data center or on your own laptop. Both are real options in 2026. Understand what you gain and lose with each.
Lesson map
What this lesson covers
Learning path
The main moves in order
- 1Two places an agent can live
- 2cloud agents
- 3local agents
- 4privacy
Concept cluster
Terms to connect while reading
Section 1
Two places an agent can live
When you use ChatGPT Agents, Claude Code, or Devin, the model that powers the agent lives in a data center owned by OpenAI, Anthropic, or Cognition. Your prompts, your files (if you share them), your commands — all go over the internet to be processed. That's cloud.
When you use Ollama, LM Studio, or OpenClaw on your own machine, the model runs on your hardware. Your data never leaves your laptop. That's local.
Compare the options
| Cloud agents | Local agents |
|---|---|
| Most powerful models (Opus 4.7, GPT-5). | Smaller models (Llama 4, Qwen 3.5, Gemma 4). |
| Per-token or per-month pricing. | Free after hardware + electricity. |
| Your data travels to a vendor. | Your data stays with you. |
| Always up-to-date. | You choose when to update. |
| Needs internet. | Works offline. |
| Strong tool ecosystems. | MCP support now native (llama.cpp, Ollama). |
When cloud is the right call
- You need the frontier model for hard tasks (coding, research).
- You don't want to manage hardware or updates.
- Your data isn't sensitive (public research, marketing copy).
- You're building something customers will use — you need the best.
When local is the right call
- You work with regulated data (health records, legal, finance).
- Company policy says 'no cloud AI on internal docs.'
- You want predictable cost — no per-token surprises.
- You care about running offline or controlling the model's version.
- You're building something personal and want full ownership.
Hardware reality check (April 2026)
Compare the options
| Model size | Works on | Rough quality |
|---|---|---|
| 3B parameters (Llama 4 3B, Gemma 4 3B) | Any M-series Mac, modest PC GPU. | Decent for chat, weak at multi-step. |
| 8B-12B (Qwen 3.5 8B, Llama 4 8B) | 16 GB RAM, mid-tier GPU. | Good at reasoning and tool use. |
| 30B-40B (Gemma 4 27B, Qwen 3.5 32B) | 32 GB+ unified memory (M3 Max, M4). | Near-frontier on focused tasks. |
| 70B+ (Llama 4 70B) | Dedicated workstation, Mac Studio M4 Ultra. | Close to cloud frontier a year behind. |
In the next two lessons, we'll look at OpenClaw — a specific open-source local orchestrator — and how Ollama makes self-hosting a model as easy as `ollama run`.
Key terms in this lesson
End-of-lesson quiz
Check what stuck
15 questions · Score saves to your progress.
Tutor
Curious about “Cloud Agents vs. Local Agents: The Privacy Tradeoff”?
Ask anything about this lesson. I’ll answer using just what you’re reading — short, friendly, grounded.
Progress saved locally in this browser. Sign in to sync across devices.
Related lessons
Keep going
Builders · 40 min
Builder Capstone: Design an Agent for Your Life
No code. Just design. Pick a real task you do every week and draft a complete agent spec — goal, tools, loop, stop, approvals, and what success looks like.
Builders · 28 min
Chat AI vs. Agent AI: The Real Difference
A chatbot answers. An agent does. Learn the line between a model that talks and a model that acts — and why crossing it changes everything about how you work with AI.
Builders · 30 min
Why Agents Fail (and How to Notice)
Agents fail in weird, quiet, expensive ways. Learn the six failure modes, the warning signs, and the simple habits that catch problems before they compound.