WebTerm is a browser-based terminal sandbox for the AI era. As AI coding tools become mainstream, CLI literacy matters for both engineers and non-engineers. But many beginners are afraid to open their local terminal because they might break something. WebTerm provides an ephemeral, no-signup, free, and safe environment to learn Linux commands, Git workflows, and CLI-based AI tools. Currently in beta — feedback is welcome!
Learn the terminal like a game — in a browser sandbox
An interactive platform for learning the terminal, Git, and Vim. Hand-drawn slides teach the concept; a real in-browser sandbox checks every command you type. Progress across a 3D world map with a companion that evolves. 12 courses, 129 lessons, all free.
WebTermSafely run even rm -rf / — all in your browser.
Launched on February 5th, 2026
WebTermA browser terminal sandbox for learning CLI without fear
Launched on January 13th, 2026
Reviews
No reviews yetBe the first to leave a review for WebTerm
The line about refusing to leave the AI's work as a black box is the right one, and it goes further than the terminal. I shipped a render queue an agent wrote, approved the diff twice without reading past the function signature, and lost three days to a stale job id. So the lesson I'd add is reading a diff, not typing commands, because the way people get hurt now is approving code rather than writing the wrong line. Free and no signup is the reason this one actually gets used.
Thanks for sharing this! Completely agree that reviewing agent output deserves its own lesson.
It's something I've been wanting to build, and the fun challenge is planting a bug that feels real instead of a puzzle.
Added to the course backlog!
Report
@daiaoki The bug I'd plant isn't a logic error, it's a fourth copy of a helper that already exists three files over. Nothing breaks, tests pass, and you only feel it six weeks later when you fix one of the four. A retry wrapper around a call that already retries internally is the version that cost me two days.
That’s a much better bug than a simple logic error.
The "fourth copy of a helper" example is exactly the kind of thing that passes every test but quietly increases the cost of maintaining the code. And the retry-on-top-of-retry case is painfully realistic.This makes me think the lesson should be less about "find the bug" and more about "spot what doesn’t belong in this diff."
Definitely adding these to the backlog. Thanks for the examples!
Yes, exactly. I think this is where our "Simulator" format could be a really good fit.
The idea is inspired by a flight simulator for pilots: instead of just learning concepts in a course, you’re put into a realistic situation and have to make decisions and deal with the consequences.
A code review scenario like this — where nothing is obviously broken and you have to go hunting through the code to figure out what doesn’t belong — would be perfect for it.
The line about refusing to leave the AI's work as a black box is the right one, and it goes further than the terminal. I shipped a render queue an agent wrote, approved the diff twice without reading past the function signature, and lost three days to a stale job id. So the lesson I'd add is reading a diff, not typing commands, because the way people get hurt now is approving code rather than writing the wrong line. Free and no signup is the reason this one actually gets used.
WebTerm
@asadmalik901
Thanks for sharing this! Completely agree that reviewing agent output deserves its own lesson.
It's something I've been wanting to build, and the fun challenge is planting a bug that feels real instead of a puzzle.
Added to the course backlog!
@daiaoki The bug I'd plant isn't a logic error, it's a fourth copy of a helper that already exists three files over. Nothing breaks, tests pass, and you only feel it six weeks later when you fix one of the four. A retry wrapper around a call that already retries internally is the version that cost me two days.
WebTerm
@asadmalik901
That’s a much better bug than a simple logic error.
The "fourth copy of a helper" example is exactly the kind of thing that passes every test but quietly increases the cost of maintaining the code. And the retry-on-top-of-retry case is painfully realistic.This makes me think the lesson should be less about "find the bug" and more about "spot what doesn’t belong in this diff."
Definitely adding these to the backlog. Thanks for the examples!
WebTerm
@asadmalik901
Yes, exactly. I think this is where our "Simulator" format could be a really good fit.
The idea is inspired by a flight simulator for pilots: instead of just learning concepts in a course, you’re put into a realistic situation and have to make decisions and deal with the consequences.
A code review scenario like this — where nothing is obviously broken and you have to go hunting through the code to figure out what doesn’t belong — would be perfect for it.
I think we just found a great Simulator idea.