March 08, 2019
Here are some things I did this past week…
I interviewed Thomas Blower last Saturday. We did the interview at Nourish at 10 AM. It was fun having it at a different place for once. Tom was a great interview again. I love it when people are open about their lives and honest about who they are at that moment. Tom is 6 years younger than me but I’ve always felt like he’s one of my teachers as well. It’s humbling to say the least.
This past week I’ve spent a fair amount of time dedicated to Sizable and getting tracks ready but I’ve also spent a fair amount of time with my roommate Caleb Salyer making some music and radio shows / podcasts. More than likely, the things we’re making won’t see the light of day. We’re just experimenting with format. It’s freeing to make things that have no obligation to please anyone else but ourselves.
As I mentioned last week, I’ve decided to start exploring AI in music a bit more.
This past week I basically got my computer set up to work with Magenta and I learned and relearned some aspects of Tensorflow and Python. I see that my first goal is going to be producing datasets that are in the same format that Onsets and Frames requires. I’ve been investigating how that data is organized into TFRecords. It doesn’t look too hard but it is different than anything I’ve done. With a little luck, I’ll be able to generate my own TFRecords and get good results. That will be a proof of concept as to whether my infinite dataset idea is worth exploring further or not.
At Carfax, I’ve been working solo on refactoring my favorite app: a Graphql API that provides shop related data to some of our team’s products. And I might spend some personal time this weekend refactoring an older product into a more modern style.
I love making big refactors and rewrites because I love the feeling of going in there and making things right. Too often in the coding business is it the case when a project keeps getting older and instead of getting the love and attention it requires to be pleasant to work on, it gets neglected. Months and years pass where people keep adding in code thoughtlessly, copying and pasting to just get the feature done as fast as possible. Then design principals fall out of favor and instead of reworking the app, there just becomes a divide with more than 2 styles of code in there. Then a new framework version comes out that makes things easier but the app is too far gone to easily update it and the technical debt just keeps accruing.
But the real incentive for refactoring this older app is: I can essentially experiment with using gRPC as a means to communicate server to server instead of the overused REST and demo a proof of concept that might affect which future tech choices get made at Carfax.
Written by Joseph Weidinger.