Why Clyep exists

Clyep (pronounced “clip”) is a gallery of short video explainers on AI and machine learning. Here is why we built it.

An explanation works better watched than read

Nothing teaches like doing the work yourself: writing the code, running it, watching it break. But before that, and alongside it, you learn from explanations, and most explanations of AI and machine learning are pages of text. A page can hold every fact about attention or a retrieval pipeline. What it cannot do is move. It leaves the mechanism for you to animate in your head, and in this field the mechanism is most of what matters: what flows where, in what order, and what changes when a setting changes.

An animated explanation does that work for you. A diagram builds up one piece at a time, code runs as it is explained, and a voice paces the idea so you can follow it. For learning from a source, that is the most effective form we know of, and it is why Clyep is video first.

The reader is not left out. On every video page the transcript and the chapters sit beside the player, every line clickable, so you can skim, search, and jump to the part you need.

So why doesn't every topic have a good video? Two things get in the way.

Making a video the usual way is slow

A technical explainer is usually made by hand: screen captures, a recorded narration, hours of editing, and a re-recording when something changes. A ten-minute video takes days, sometimes weeks. So most technical topics never get a video at all, and the ones that exist go stale the moment the library they describe changes.

Clyep's videos are made differently. Each one is a written specification of its scenes: diagrams, statements, lists and code, in plain text files. A compiler turns the specification into animated scenes, and a synthetic voice reads the script. Because the whole video is text, AI agents can draft it and revise it. A person decides what to teach, directs the agents and checks every scene. The agents do the production work. Our videos are human-directed and AI-delivered. A topic gets a video in hours, and a correction is a change to a file.

Elsewhere, a published video cannot be corrected

Speed solves half the problem. The other half is what happens after publishing. On a video platform, any change to a published video, however small, means uploading a new one. That leaves two choices. Keep the old video up, and it stays wrong while the link, the views and the comments keep pointing at it. Delete it, and its history goes with it. Either way, nobody can see what changed, or why.

On Clyep a video keeps one address for life. When it is improved, a new version is published and every earlier version stays watchable. The change log says what changed, when, and who made the change. Views and likes stay with the video across versions, so a re-cut never starts from zero.

That is the idea: video first, made quickly, and kept right over time. There is more on versions, credits and views in how it works. Is there a topic you would like to see explained and the library does not have it yet? Ask for a video.