An AI Terminal Is an Assistant That Acts on Its Own
You've probably used a chat box: you ask, it answers, and that's it. An AI terminal is different—it sits inside your computer and can actually open files, run commands, and edit code.
Precisely because it really acts, the thing designers worry about most is one thing: how to keep you always able to see and control what it's doing.
Three Core Design Principles
Almost every reliable AI terminal revolves around these three principles:
1. Visible process: everything it wants to do at each step—which file it'll touch, what command it'll run—is laid out for you to see, never done secretly.
2. Stoppable anytime: you can hit stop at any moment; it won't barrel down one road to the bitter end.
3. Actions need authorization: before dangerous actions that touch the real environment (deleting files, pushing code), it asks you first by default.
Why It's Designed This Way
Because the AI makes mistakes, and what it touches is your real files, your real repo. Once it misunderstands you, the cost is real.
So these designs aren't meant to inconvenience you—they break "trust" into small, controllable grants, so you dare to use it and use it with peace of mind.
自测 · 学完检查一下
想真正动手做题、记进度、攒连胜？到互动课里练。
What's the key difference between an AI terminal and a plain chat box?
答案：It can act on its own in your real environment (edit files, run commands)
A chat box can only "talk"; an AI terminal can really "do"—which is exactly why it has to be designed to be controllable.
Which of these is NOT one of the three core design principles the lesson covers?
答案：Answers must rhyme to be memorable
The three principles are visible process, stoppable anytime, and actions need authorization—nothing to do with rhyming.
Fill the chant: see it, stop it, ____ (confirm by default before a dangerous action).
答案：ask first
"Ask first" maps to "actions need authorization": ask for your consent by default before a dangerous action.
An AI terminal pops up "Confirm delete?" before deleting a file—which design principle does this mainly show?
答案：Actions need authorization
Deleting is a dangerous action that affects the real environment; asking your consent first is exactly "actions need authorization."
Judge: showing in real time everything the AI terminal is about to do, so you can "see and control" it, is good design.
答案：Correct
Showing actions in real time is "visible process," which serves the goal of letting people see and control it.
Judge: because the AI terminal touches your real files and repo, once it misunderstands, the cost is real.
答案：Correct
Its actions land on the real environment, so the consequences of errors are real too—which is exactly why it must be designed to be controllable.