Back to homepage

Weeknotes #11

This week mostly about what I learned on machine learning course, design tools, my first Chrome Extension, and what caught my attention on Twitter. With some open-source flavor.

Machine learning for creatives

It was great 4 days full of creative ideas, experiments, and weird randomly generated sounds. To me, I was interested mostly in training larger some data set and generating something from it. So I joined a team with an idea to generate Fictional Constitutions from data of some existing constitutions.

Our project had three parts:

  • Getting data from public API of existing constitutions.
  • Training two data sets for text (for the preamble, and the rest of the constitution).
  • Implementing it together with existing data set for sketching.

And this is the result (published on a pretty cool domain if you ask me) – Not bad for two days of work. :-) You can find it on Github too of course :-)

However – I understand very roughly how the whole machine learning thing works, and I realized some stuff during these four days:

  • It's mostly about data. Also that the first hard part is to get good enough data. Which usually needs to be cleaned and prepared.
  • The open-source community is solid in machine learning – a lot of stuff like data sets, trained models, or even training algorithms are already pre-made and prepared for use. Just to plug it and with ease of ml5.js generate some cool-ish projects.
  • There is this "codepen for machine learning experiments" –
  • Tensorflow can be used with Node.js and written in javascript. Just nice.
  • I think that my place in this "machine learning tech word" is very possibly in the "data part" of the process. I really enjoyed the part of getting, understanding, and cleaning data. I realized that this wasn't for the first time – some of my side projects are heavily relying on collected data, and I always liked that part a lot. I even like it that much, that I wasn't able to push it into production because I was still improving the state of data itself 🤷‍♂️

Different optics on open-source

I mentioned that it's great that there is a lot of open-source stuff to help with training our datasets. So I want this to remind me to check whatever I use for possible vulnerabilities.

Open source & design

Since we started with Orbit and the whole idea of open-sourcing our design system, I've started with checking what is happening in the open-source community around design. Almost immediately, I realized that there is so much space. And this Twitter thread sums it up pretty well.

Design tools

This week was a lot about design tools for me, a lot of new updates and great article I found on Twitter. Check my Twitter thread with the best picks:

+ a few of my opinions on recent design tools updates

  • Figma releasing plugins on 1st August, which means that a lot of old Sketch plugins will be converted to Figma too (e.g. Figma Walker, copy of Sketch Runner, or this Figma version of Select similar layers). Also, because this version of Plugin API will allow you to generate anything into canvas, people are playing with generative design. If any design tool still waits with introducing ways how to extend them, they shouldn't wait too long.
  • Also, Sketch published beta for Sketch Teams. It's just great to see what competition does with the whole field. And this is hard competition, not just fort Figma, but also for tools like Abstract, Zeplin or Avocode.
  • This article on Abstract workflow and "how to explain GIT to designers." It's not that much about Abstract (because I am having doubts about this tool anyway), but ideas for design versioning in general.
  • Redesigned – looks "more 2019" now (with dark mode, of course). However, the best part of this update is publishing their API with raw data. For example, check this analysis of That's a lot of data to play with. 🤓

One-time side project – Chrome Extension for Twitter

Twitter released a new design last week globally. And it was followed by so much hate. Don't believe? Check #newTwitter hashtag 🤦‍♂️

But to be honest – I like this new Twitter. Yes, there are still some issues, but they finally brought two essential features to the desktop:

  • Bookmarks
  • Switch between more Twitter profiles

Another thing I like to see is that someone was able to actually release some big project written in React Native. Hopefully, they will share somewhere their journey.

However, back to my one-time side project. I mentioned that there are still some issues – for me, it was mostly about super-large typography even on the smallest settings. So I fixed it with writing my first Chrome Extension – Polished Twitter.

I published it on Chrome Web Store (because why not) – and after a few days, there is exactly one user who uses it. Me. Not a big deal though, it was created for my needs :-)

All articles