In essence I treated it like a junior dev who already knew the tech stack. So tell it what I wanted to achieve then pair program to get it working. Then iterate to add functionality. Once a feature is working, I could then ask it to help me understand how the tech stack was being used. Lots of fun, and very productive.
Webp encoded tiles extracted from the CoGs available on AWS then embedded within a pmtiles file.
I was seeing almost an order of magnitude reduction in file size compared to the CoGs with no visible image degradation.
They’re still not at feature parity with 2x the team, 2x the time and 3x the lines of code.
They kept changing stuff, breaking my work flow, with every change it felt like there was more friction to listen to the music and podcasts I wanted, and less friction to listen to algorithmically selected slop I didn't want to listen to. Eventually I just said fuck it. I don't need this source of stress and frustration in my life. I've never looked back.
I'm on youtube music now. In many ways it's a worse product, but it at least stays the same and doesn't keep trying to make me change my listening behavior.
I suspect Spotify's problem is that they have (or had) too many developers, so you get this pressure to look busy "improving" the product, with endless lateral change as a result instead.
shadow-cljs[2] makes using npm libraries easy.
i’ve settled on go backends and reagent frontends as my default setup[3].
1. https://reagent-project.github.io/
I hate to be that guy, but there is 1 silver bullet. Erlang. Whether you are writing code for a cluster of machines or a single node, the developer experience is the same.
I thought that Elixir could do better and created a replacement Elixir monolith. The replacement happily handles the workload while running on my MacBook Pro.