Getting Back to the Blind75

None of my content is AI-generated. Learn more about my use of AI.

---

I initially started doing the Blind75 problems last year as interview prep. The problems come up often in interviews, and I was right at the beginning of my transition to full-time software development.

It thankfully didn’t take me long to find a great position, and as I was starting a new role with a then-unfamiliar tech stack, I didn’t really have time to keep going with the Blind75 problems.

Now, a full year later, I’m coming back to it. I have three main motivations for doing so:

1. to plug any gaps in my knowledge
2. to stay sharp as the industry embraces changing workflows

Covering my bases

I’m self-taught, and I want to make sure I don’t have gaps. In a lot of computer science programs, students will cover DSA (data structures and algorithms) — a fundamental part of software development — in their curriculum. I have a wide breadth of experience with differnet tech — from Javascript frontend frameworks to different backend languages, DevOps, Linux, Docker, etc. But I want to make sure I also have the depth I need in some nitty-gritty technical areas.

Staying sharp (and having fun)

I have seriously mixed feelings about it (I’ll probably write about it in the near future), but the tech industry is moving more and more towards AI and “agentic” workflows¹. As a result, developers are doing less and less actual coding, offloading that particular part of our job to AI models like Claude and Codex. We are already seeing a cognitive decline in software developers.

I actually find programming a lot of fun. It gives a similar satisfaction that many find from crosswords and sudoku. It’s also a craft, and I feel like we are losing that craft as an industry, in favor of accelerated development and faster time to market. (More on that in a future post.)

Returning to the problems

When I came back to the problems yesterday, I didn’t realize how many I’d already done — about 25 of the 75. I started in Python, which is a great choice for interviews, as it’s very concise and doesn’t have a lot of ceremoy².

I’ve decided to begin again, but this time with a second language: Java. Why Java? I have always had a fascination with it — from a distance. I didn’t do a computer science degree, and, while it is extremely popular in the industry, I don’t get to use it often. It is widely regarded as the ultimate learning language, and so I decided to give it a go for the second language of my Blind75 solutions. I might do a third pass with a more low-level language like C++ or Rust… we’ll see.

Footnotes

1. Agentic workflows are where the AI model acts on its own (in response to a prompt). In comparison to something like ChatGPT, where it is just responding to you with text or images, agentic workflows allow AI models to plan and then execute coding tasks. ↩

2. By this I mean it doesn’t require a lot of code to get up and running. Almost all of the code you write is for solving the actual problem at hand, rather than setup and boilerplate. ↩