I have subscribed to various YouTube channels over the years and originally had a YouTube account to track them all. After I became increasingly aware of privacy and tracking on the internet I wanted to find a way to still be able to consume the same content without needing to log in.

Being a web developer I obviously built my own web app. It was a relatively simple app, especially as YouTube amazingly provides an RSS feed for each channel. I could parse the RSS feeds of all the channels I was interested in, aggregate all the videos together into a single grid, pull in thumbnails. Easy.

The original app was written in 2015(!), and I've used it pretty much every day since. I've added a few features over the years like featured channels, whose thumbnails show up at x2 size, and keeping track of which videos I've watched. There were a couple of annoying things about it:

1. I had to manually run the server
2. Some thumbnails would show the default placeholder
3. The feeds had to be manually refreshed via a terminal

After reviewing the decisions 10-year-ago me had made, I decided to rebuild it from scratch and also because that's a web developers favourite thing to do 😆. I started by putting together a list of what I needed it to do and also what I wanted it to do. Doing this helps me focus on what is important to make a thing actually usable.

Requirements

• Pull RSS feeds for channels
• List videos in a grid - with thumbnail and title
• Order by date
• Have 'featured' channels which show a 2x thumbnail
• Save watched videos to LocalStorage
• Watching a video opens in a dialog which takes up the full viewport
• Video feeds are refreshed in the background - cron?
• Refresh the feeds in parallel
• Cache the feeds where possible
• Lazy load thumbnails - no layout shift
• Simple - no dependencies

Side quests

• Set it up as a docker service which runs on boot
• Auto update the subscription list when new videos are loaded - SSE? polling?
• Only refresh feeds when I'm at the computer - use lock watch?
• Handle different thumbnail formats
• Management of subscriptions - list/add/remove/feature
• Search
• Shared cURL resources - curl_share_init and the new RFC curl_share_init_persistent

Over the course of 3 evenings I (aka me and Claude) wrote the whole thing from the ground up. I also managed to achieve some of the side quests or realise that I didn't need them.

Refreshing Feeds

The core of the app is the feed refresh, see this post for details, but the short version is that it is really fast, 600 feeds per second fast.

I realised, while thinking about how best to process updates in the background, that the requirement to update in the background had gone, it's so fast that it can happen every page load. Part of me wanted to be able to keep it working offline, but it's a video player, if you're offline the whole thing is useless.

Thumbnails

YouTube thumbnails are annoying... I don't mean clickbait, that is annoying too, I mean that to get the right thumbnail you have to blindly request the different versions and then determine if a placeholder image was returned based on the filesize. What would've been so hard about returning a 404 status code?

For the thumbnails in the app I created an endpoint which abstracts this nonsense and allows them to be cached easily.

OSS

Of course the whole thing is open source if you want to run your own version: