How a supersuit compares
- 1Chat apps (ChatGPT, Claude Chat) stop at the answer: they cannot act on anything.
- 2AI built into an app (Copilot in Microsoft 365, Gemini in Google Workspace) stops at the edge of the one app it lives in.
- 3Industry agents (Harvey for law, Hebbia for finance) stop at one profession's documents and tools.
- 4Guided agents (Claude Cowork, ChatGPT Work) stop at limits tuned for everyone.
- 5Assistants you text (Instinct, Tab) stop at what their own phone and accounts can reach.
- 6Agents with a Linux computer of their own (OpenAI's Dots, Meta's Muse) stop at the computer the company runs for them.
- 7Do-it-yourself agents (OpenClaw) go as high as you can build and keep them yourself.
- 8Coding agents (Claude Code, Codex) have no ceiling on your computer, but start knowing nothing about you.
Its own category
A supersuit
A supersuit is built on a coding agent, the most capable category there is, and fitted around one person. It holds the whole context of their life and work, organized so it stays true, on a computer they own.
That is the difference that compounds. Every category above works from a slice of you, or from none of you. A supersuit works from all of you, and you can give it all of you because it is yours.
- Who holds your context
- You, in separate worlds: your personal context is yours, each company's is the company's
- Can you see your context
- Yes. Every piece is a file you can open, read and correct
- What it can do
- Anything you can do on a computer, across your whole life and work
Appendix: every category in detail
There are many AI products and only a handful of categories. Each category gives a different answer to three questions, and those answers decide how useful it can become.
| Category of tool | Who holds your context | Can you see your context | What it can do |
|---|---|---|---|
| Chat apps | The company | No | Answers |
| AI built into an app | Each app | Partly | Works in one app |
| Industry agents | The vendor | Partly | Deep work in one profession |
| Guided agents | Mostly the company | Partly | Guided tasks |
| Assistants you text | The company | No | Errands by text |
| Agents with a Linux computer of their own | The company | No | Tasks on its own cloud computer |
| Coding agents | You | Yes | Anything on your computer |
| Do-it-yourself agents | You | Yes | A lot, if you build it |
| A supersuit | You, in separate worlds | Yes, every file | Anything, across your whole life and work |
- Who holds your context?You, or the company that makes the tool.
- Can you see your context?Files you can open and correct, or memory and computers you never see.
- What can it do?Answer you, work inside one app or sandbox, or work across your own computer.
The categories of AI tool
Chat apps
ChatGPT, Claude Chat, Gemini, Meta AI, Microsoft Copilot, Le Chat
- Who holds your context
- The company that makes it
- Can you see your context
- No. Its memory of you is on their side
- What it can do
- Answers, writes and reasons with you
- Good for
- Thinking out loud, drafting, quick questions
AI built into an app
Copilot in Microsoft 365, Gemini in Google Workspace, Notion AI, Slack AI, Granola
- Who holds your context
- Each app holds its own slice
- Can you see your context
- Partly, inside that app
- What it can do
- Works inside the one app it lives in
- Good for
- Speeding up work you already do in that app
Industry agents
Harvey for law, Hebbia for finance, Abridge for medicine
- Who holds your context
- The vendor, with what your firm uploads
- Can you see your context
- Partly, inside their product
- What it can do
- Specialist work in one profession's documents and workflows
- Good for
- A firm's core specialist work, done to that field's standard
Guided agents
Claude Cowork, ChatGPT Work
- Who holds your context
- Mostly the company that makes it
- Can you see your context
- Partly
- What it can do
- Real tasks, inside careful limits tuned for everyone
- Good for
- A first taste of AI doing work instead of only talking
Assistants you text
Instinct, Tab, Poke, Caddy, Pally, Arlo, and Muse on WhatsApp
- Who holds your context
- The company that runs it
- Can you see your context
- No. It works on their servers
- What it can do
- Errands and answers from a message thread
- Good for
- Quick asks from your phone, wherever you are
Agents with a Linux computer of their own
OpenAI's Dots, Meta's Muse, xAI's Grok Bot, Manus
- Who holds your context
- The company, on a cloud Linux computer it runs for you
- Can you see your context
- No. The computer it works on is out of sight
- What it can do
- Real tasks, around the clock, on that Linux computer
- Good for
- Errands you hand off whole, that keep running while you are away
Coding agents
Claude Code, Codex, Gemini CLI, Cursor
- Who holds your context
- You, on your own computer
- Can you see your context
- Yes
- What it can do
- Anything you can do on a computer
- Good for
- Everything, once someone sets it up around you. Out of the box it is generic and forgets
Do-it-yourself agents
OpenClaw and other open-source agents you host yourself
- Who holds your context
- You
- Can you see your context
- Yes
- What it can do
- A great deal, if you are the engineer
- Good for
- Builders who want to assemble and maintain their own
Compared one on one
- Comparing Grok Bot and Supersuit
- Comparing ChatGPT Work and Supersuit
- Comparing OpenClaw and Supersuit
Where context held by someone else leads: Context prisons