In this blog post, I want to dive into my personal experiences as a 👶 junior engineer 👶 entering the industry during the 🤖 Age of Ultron 🤖. I want to talk about how it has affected me technically, but also how it has affected me emotionally. Despite how many machines engineers work with, around, and now seemingly for, we are humans at the end of the day, and we should talk about that more. For this reason, I want to especially highlight the emotions surrounding these experiences, both figuratively and literally.
Recently, I have been feeling a sense of urgency to “figure out” when I should be writing code by hand ✍️ to deepen my understanding of programming and when I should be allowing LLM’s to generate the code for an architecture that I design. I postulate that I feel this rush to decipher the exact balance to strike because I am a young developer at the early stages of my career. As I write this, I am still in university, and have always had an end-date attached to my various engineering job contracts. I am just a baby, and the world is moving so fast towards something we don’t fully understand. It always is, but it feels faster now for some reason.
Learning the hard-way is paramount in any career and I feel that it directly correlates to the quality of senior engineer that I will become, but I find myself learning the easy way or even not at all in recent history. The industry is moving at the speed that LLM’s enable, so it is often difficult to justify writing code by hand at work, simply because of the speed that development and iteration must move at with timelines adjusting to this newfound “nitro boost button”. I have seen this very direct shift from interning at the same company twice, both in programming related roles.
The first time I joined, LLM’s were not allowed for generating code of any kind, as ChatGPT was the only widely known and available tool, and it was real bad at most technical things in general, and in my experience, especially coding in a close-to-the-metal setting. In this role, I coded all of my embedded software by hand. I learned so, so much. I learned so much about the specifics of C/C++ syntax, best practices in low-level development, and nuances in what creating maintainable and scalable embedded software meant. I remember receiving my first ticket on ClickUp, having absolutely no idea where to start, and subsequently reading a codebase with ten’s of thousands of lines of C++, or what my manager described as C with classes and a few other bells and whistles. I also remember reflecting on that moment 8 months later and feeling an immense amount of pride in the fluency that I gained over the course of my internship. Upon receiving new tickets a few months in, I knew exactly which files, classes, and often lines of code this feature or bug fix was going to touch, and that felt amazing. I felt at one with my “craft”.
The second time I joined (about 8 months later), a huge boom occurred in AI usage in general, but especially with the newfound universal adoption of Claude Code. I was shocked when one of the first things admin granted me access to was a Claude Code license with seemingly infinite tokens. The previous time I was there, which wasn’t very long before, the term LLM was practically a profanity in the context of programming. The attitude towards code generation completely flipped on its head.
I want to be clear though that this is not necessarily a negative thing in of itself, this is simply a reflection of feeling surprised to see the attitude towards some philosophy change so, so fast. That’s part of what feels so intimidating. It’s all changing and moving so, so fast. It makes perfect sense to start using a tool once it becomes promising, and to use it even more when it seemingly proves itself to be useful in so many situations every day, so truly, I get why it has been so heavily adopted.

I feel that an often overlooked side-effect from macrodosing LLM, is the reduction in junior-senior knowledge sharing. Maybe this isn’t something many other juniors are experiencing and is more specific to myself, but I would be surprised if that were the case.
The 🧙magical-wisdom🧙‍♀️ that my more senior teammates bestowed upon me in my earlier internships have and always will be directly represented in the quality of engineer that I am. Their brains store instrumental teachings from years of being a human developing complex systems. Nowadays, when us juniors don’t know something or are stuck between some design decisions, it saves time for both the junior and senior when the junior “just asks Claude”. I put quotations around, “just ask Claude”, because it is a personification. Claude is not a human and you are not asking someone something as you would a human. This is something new, something different, and in my opinion, something scary. Also, to be clear, this “malpractice” of “just ask Claude” is a two-way-street, while I feel that more responsibility lies in the maturity and power of the senior to encourage asking questions, it is also the junior’s responsibility to do the more time-consuming and for some, scarier task of asking your manager something rather than “just ask Claude”.
LLM’s don’t have nearly as valuable knowledge as the seniors on your team do. Claude might be more fluent with a given language’s syntax or the specifics of a well-researched design paradigm, but Claude did not spend 32 hours debugging a segfault back in 2016, your manager and more senior teammates likely did, and they very clearly remember the slurry of subtle human-made decisions that led to the issue and subsequently what led to a better design. When you ask Claude if using the Template Method in Python is valuable for your given situation, it might say something like, but not so far from, “✨Absolutely, you’re so right, so smart, and so good looking✨”. On the contrary, your manager might have a real-lived human experience from 1992 that says otherwise, or even better yet, a mix of experiences that leads to an in-depth conversation about whether or not the Template Method makes sense or if something else does. In this latter and more beneficial case, you are not being fed an answer, whether its your manager or a water-starved, dull, grey, machine that lives in Anthropic’s newest plot of land.

I find myself only speaking about the negative parts from LLM adoption, so I feel a need to express that I do see that there are positives as well, but I just don’t feel like talking about them. Not right now at least. I feel more scared, drained, and frustrated about LLM adoption in the workplace than I do curious, excited, and content. So I feel a need to talk about this side of things first.