Five-hundred twenty-five thousand, six-hundred minutes, how do you measure, measure a year?
As someone who’s been working closely with software for the past two years, I’d say in git commits:
For those who aren’t familiar with version-control systems (VCS), a git commit is simply a log or savepoint in any software project. Each log keeps a history of the things you’ve done so far, so that you can track them in the future. If there is a commit history of my Tuesday, it would look like this:
Cook breakfast and lunch Take a bath and dress up Go to work
For software projects, They usually look like this (from my project, PySwarms):
Add logging functionality in Reporter module Update documentation in sphinx Remove deprecated modules
The chart above shows how many logs I have done every single day for the past year. Greener boxes mean more commits, whereas gray boxes are zero commits. It’s not necessarily a measure of productivity, more of an approximation of one’s tech life.
In this post, I’d like to review my year in commits, then discuss in brief some lessons I’ve learned and challenges I’ll set for myself next year. I think, that as I end my Masters degree and transition into a new phase in life, it’s good to look back on the things I’ve done so far.
My year in commits: a review
First, let’s have a brief tour of my life in commits. I’ve included my commit history from January 2017 and added some annotations on significant life events during the past year:
I realized that this commit graph covers a lot of things that happened during grad school. Going through all of them may be too much, so let me give you some highlights:
- The programming bug (July 2017). Perhaps, the “programming bug” bit me when I started PySwarms back in July 2017. Although my blog has been set-up back then, I only started programming seriously with that library. Right now, the project is growing: we have two more maintainers and fourteen contributors all-in-all— not bad for my first soirée.
- Building my research code (August 2017–) Git commits for my thesis were usually done in 2-3 week bursts. PySwarms taught me a lot about software development, and I decided to apply them in my research code.
- Long rests (Parts of July and August, December 2017). I took a good amount of breaks last year, but it seems that I worked non-stop since January. The last few trips I had were in Matsue and Miyazaki. Then, the “final” long break happened during the Christmas holidays, when my mom and girlfriend came. After that, all trips were just business trips: an internship in Kobe and a conference in Tokyo.
- LaTeX manuscript (February–June 2018). I did enjoy preparing my manuscript. I started five months before the deadline, so I think there was no pressure back then. Since we wrote some papers beforehand, I didn’t start from scratch. With LaTeX, I devised a git workflow for my thesis document. I’m really happy and proud of how it turned out!
- Long road home (July–September 2018). The last three months were all dedicated to my other endeavors. I applied for an internship in Preferred Networks (PFN), and did some preparations for my return. PFN had a coding exam, that’s why you’ll see a hefty amount of commits during that period. To be honest, the last few months were the most difficult months for me. I’d probably write more about that in the future.
Overall, I think that I had a good year in programming and software. With PySwarms, it seems that I crossed an imaginary line from programming just to make things work into being aware of design patterns, idiomatic code, and software systems. The latter forced me to become more attuned to a software’s lifecycle and flow rather than just its logic.
I learned a lot, and realized that the tech world is really huge. Before, I thought that knowing Python (or any programming language) is enough. But there’s a whole world of software practices that I’m not aware of: continuous integration, deployment, testing, etc. New and exciting things to discover!
I still consider myself a newbie, and perhaps I’ll always be a newbie for the rest of my career in tech. Perhaps it would be nice to share some nuggets along the way.
Some lessons learned
As I’ve said before, I’m still starting out. I think I just got a lucky break in PySwarms, for it accelerated my learning—mostly by necessity. In this section, I’ll write down some lessons I’d bring as I go forward in this journey in software and tech. I may not be the most qualified person to give nuggets of wisdom, but hey, maybe you’ll find them useful:
- Strive to write readable and idiomatic code. I think I will never consider myself as a 10x programmer, just a 1x programmer with good habits. One good habit is writing readable and well-documented code. They say that research code is frustratingly hard-to-read (just look at OpenAI), but I think it’s an opportunity for improvement. Well-documented code doesn’t just help other developers, it helps you as well.
- Create a large volume of work. As with all creative work, I think one way to acquire “taste” and improve one’s craft is to keep on building things (see Ira Glass on the Creative Process). I realized that every gist, commit, or project never goes to waste. It’s always a chance to build something new, iterate, and improve. I’d love to maintain PySwarms because it’s one of my very first breakthrough projects, but I’d love to see myself building new tools and products in the future.
- Find your heroes. I realized that it’s important to look for models to look up to. Much better if they’re still actively participating in the craft. For me, I look up to Scott Hanselman (Software Eng’g), Saron Yitbarek (Software Eng’g), Reina Reyes (Data Science), and Neil Lawrence (Machine Learning). One common denominator among them is that they go out of their way to make their craft accessible to everyone. Find your heroes and be inspired by them.
- Be nice. Treat everyone with respect, be patient, and be chill. Be it in code reviews, answering issues, or teaching others, a little niceness goes a long way.
Two more things: a certain philosophy that I’m practicing right now is to always learn the stack immediately below and above mine. There’s no need to become full-stack—as long as you can do your own thing with deep expertise, and you’re aware of the context of your field—you’re already fine. It’s difficult to learn everything, especially in a rapidly-growing field. Strive to do one thing extremely well, while understanding the peripheries in an intermediate capacity. With this in mind, you don’t run the risk of spreading yourself too thin nor becoming a one-trick pony.
For example: my current stack is in machine learning, then that means I need to learn how the data is being sourced (stack below me), and how my models can affect an organization’s decision (stack above me). I don’t need to fully-grok all management principles nor ETL techniques, but as long as I’m aware, and I can do them an intermediate person would, then I think I’m good. I can focus my energy in machine learning, but still aware of the larger context of what I’m doing.
Lastly, I learned from a year of seriously doing software is to treat it like a craft: there’s really no level where I can say, “I’ve mastered this.” Truth is, tech is a moving target: the techniques that I know today might be irrelevant tomorrow. Thus, the best way to set goals is to make short-term and lean achievements.
It’s weird to put one’s learning path in public, but maybe this would make me more accountable in the future. Here are some things that I want to learn in the short-to-medium term:
- Scaling and deploying machine learning models. I’ve been very interested in this problem since the ScaledML conference. How to deploy ML models at scale? How to reduce latency? How to update and train models online? Might be nice to learn technologies such as Kubernetes and Ansible for orchestrating reproducible containers.
- Agile and lean management. I’ve been working with collaborators and contributors in PySwarms, and it would be nice to improve our workflow so that we don’t go stale whenever everyone’s busy. I want to learn how to manage IT projects.
- Finally learn Haskell. Yeah, this is something I’ve been putting off for quite some time. I actually started a small project in Haskell, might as well finish it. I also purchased Get Programming with Haskell, hopefully it accelerates my learning.
- Less commits on weekends. For the past few months, it seems that I’ve been writing code every day. Not that I’m complaining, but even if I love programming, doing it too much can lead to burnout. Grayed-out top and bottom rows should be in my commit graph next year.
As a close, I think that doing this yearly review has helped me reflect on my journey in tech and software. Looking back, just one year of programming has provided me with new skills and perspectives. Now, tech has become more than just code—it’s design, community, and craft. As I transition to the next phase of my life, I’m excited to see what the next few years will bring.
- 10-02-2018: Update some grammar mistakes