Skin deep. You will clone the interface and then meet the wall. Realistically a long weekend.

What stops a copy of Loom working?

The link has to work before the upload finishes. That means chunked uploading during recording and a playable manifest for a video that is still being written, which is the actual product insight and the actual engineering. Transcoding at scale. Every viewer device wants a different rendition, and a video pipeline is a cost centre with opinions. Recovering a recording when the browser tab crashes at minute nineteen, which is the moment that decides whether anyone trusts you again. Bandwidth and storage economics that get worse exactly as you succeed.

Can AI copy

Loom

Record your screen and your face, get a link before you have finished thinking about it.

Skin deep

Verdict: Skin deep
Time to copy: a long weekend
Category: ◉ Media & capture
Source: Proprietary

The prompt

Written to be pasted straight into an agent with no editing. It describes the whole job, including the parts that will go wrong.

Build an async screen recorder. Capture via getDisplayMedia and getUserMedia, composite the camera as a draggable circular overlay on a canvas, and encode with MediaRecorder. Upload in chunks while recording is still in progress so the share link is live within seconds of stopping. Server transcodes to adaptive renditions and serves HLS. Viewer page has a transcript synced to playback, timestamped comments and emoji reactions. Persist chunks to IndexedDB so a crashed tab can resume rather than lose the take.

Open inClaude ↗Terminal agents take it from the clipboard.

What copies cleanly

Capture. The browser gives you screen, camera and microphone streams for free, and compositing the webcam bubble over the screen is a canvas call.
The share page, the viewer, comments and reactions.
Transcription, which is now an API call.

Where it stops

The link has to work before the upload finishes. That means chunked uploading during recording and a playable manifest for a video that is still being written, which is the actual product insight and the actual engineering.
Transcoding at scale. Every viewer device wants a different rendition, and a video pipeline is a cost centre with opinions.
Recovering a recording when the browser tab crashes at minute nineteen, which is the moment that decides whether anyone trusts you again.
Bandwidth and storage economics that get worse exactly as you succeed.

What actually protects them

Nothing structural, which is why the category has many credible competitors.
Being the default verb inside companies that adopted it early.

Editor's note

A good example of a product where the one clever decision, publish before the upload finishes, is worth more than the ninety obvious ones.

Somebody already did it

Capcap.soOpen source and actively chasing exactly this.

Go and look at the original

We would rather you formed your own opinion than took ours. They built the thing; we only measured it.

loom.com ↗