To detect if a link is an image or a video, we used to load it on your browser and then load it again to show it to you. That was wasteful, and we now detect if a link is an image or a video with a microservice, for better reliability.

That worked! Our microservice was able to detect that the link was a video (by downloading its first bytes) and the editor indeed tried to load it.
Loading media happens client-side and in your browser, basically just <video src="https://video.twimg..." />. Probably Twitter/X reads the origin of the request (stacker.news) and rejects it with 403 (Forbidden).

I'll check this out, maybe there's more to it, and come back with results.