Thread by @pbshhh

Twitter's bookmark feature is so inadequate. About to start building something to fix it. Planning to build it in public this time. Will update this thread as it progresses. Let's see how it goes 💫

Finalised the stack 🎀

🦚 Frontend - Next.js on @vercel
🦅 Database - Firestore at @Firebase
🦊 Servers - Node.js on @Vultr
🐰 Analytics - @usefathom

Implemented authentication 🔒 using Next-auth. Lightweight and customizable. Tried Firebase and Auth0, but both feel bloated and intrusive. Next-auth is 5x faster ⚡. Also, not a fan for having 3rd party cookies 🍪 on this, so custom implementation.

Built an activity listener with @TwitterAPI 's autohook running on @Vultr and connected it to @Firebase Firestore via @vercel serverless functions. Getting 502 at times, need to debug it.

I'm planning to take a small detour and explore @fauna for a while. I'm really excited by the GraphQL implementation and this seems like a nice project to try it out. After all, the whole project is to learn a few things I've been meaning to xD

Was able to move everything to @fauna as the serverless database. No cold start time, in contrast with Firestore. Using FQL for now, hiping to migrate to GraphQL over time.

Running a custom server on Vultr didn't really work out well. Need to set up https for the webhooks, need to ssh through terminal everytime to deploy changes. Monitoring needs to be set up. So considering all this - I'm moving to @heroku. When bootstrapping, less overhead is ❤️❤️

All three subsystems - Webapp, admin panel and server are deployed. Just the bare skeleton, but it works smoothly. Need to take a break for a few days to take care of academics. Hoping to get back later this week, starting with design. ✨

What should I name it? This is always the toughest part of any project :(

featherd.io it is ✨

When embedding 10s of tweets, it takes a long time to load as each embed is basically an iframe. So implemented custom rendering of tweets via data from Twitter api. Now the loading time is practically zero. And is possible to retain deleted tweets and can do custom styling.

Now that most of the core modules are done, moving on to design. Next week or so would be spent on designing the first version of Featherd.

Took a long long break from this to get some academic work done. Starting with featherd again. Aiming for a early 2021 launch.

@threadreaderapp compile