Core · Module 1
Surfaces & Mental Models
Course outcome 1
Grok Bot gives agents four surfaces to work on. Learn which one fits which job, then try matching a few real examples.
Big question What should stay in chat, and what needs to live somewhere that lasts?
By the end of this module, you’ll be able to:
- Match each of Grok Bot’s four surfaces (agents, channels, computer, routines) to at least one enablement use case.
- Distinguish chat, which is quick and temporary, from the four surfaces, and know when work needs to be saved somewhere that lasts. (explained, not checked)
- Identify why choosing the right surface affects handoffs and reuse. (practiced in an optional reflection)
How you’ll show it: the matching lab (5 of 6 to pass), plus an optional practice quiz and reflection. Course outcome 1 of 6
Fast-track
Your Fast-track summary
You tested out of this module. Here’s the short version:
- A reviewer that works the same way every time → Agents
- A sprint review thread everyone can see → Channels
- Running scripts and editing local files → Computer
- A nightly check nobody has to start → Routines
Chat is great for quick back-and-forth, but it isn’t one of the four surfaces. Anything left in chat scrolls away.
Note: M3 is still required. The full lesson and practice are below if you want them.
The four surfaces
Chat is for quick back-and-forth. When work needs a specialist, a shared conversation, a place to run, or a schedule, use one of these four surfaces instead.
- Agents: the specialists. Each one holds a job and a persona.
- Channels: the conversation, where people and agents coordinate in the open.
- Computer: the workbench, where agents run scripts and edit files.
- Routines: the schedule. Recurring work runs without anyone starting it.
Where work lives
| Place | Good for | Watch out |
|---|---|---|
| Chat | Quick back-and-forth with one agent. | It scrolls away. Nothing here is a record. |
| Channel | Live coordination that people and agents can all see. | The thread moves on. Save decisions in a file or doc. |
| Agent memory | Helping one agent remember your preferences. | It’s private to that agent. Others can’t see or edit it. |
| A file or doc | The lasting record: schemas, inventories, briefs, saved outside chat. | Name it, and give every field an owner so it stays current. |
Try it: match each use case
Pick the best surface for each example. Get 5 of 6 right to pass. Two surfaces are each the answer twice, and chat is never the answer here.
Explain your thinking (optional reflection)
Saying why in your own words helps the idea stick. This isn’t scored.
War room: the work is a live conversation, where people and agents react to each other in real time. A one-on-one chat only includes you and one agent, and it scrolls away. Save the decisions somewhere that lasts, like a file or doc.
Nightly check: it has to happen every night, whether or not anyone remembers. A routine runs on its schedule by itself; asking an agent depends on a person starting it each time.
Quick quiz (just for practice)
Four questions with instant feedback. Your score here doesn’t affect your progress.