Generative AI has truly enabled us to do so many things. I can no longer imagine a future where we live without being involved with AI.
This note is a perfect example. The content I am speaking through voice input right now is being rewoven into proper sentences by an LLM, which then drafts this note for me. AI is integral to both work and hobbies. Even the music production I need for my new DJ hobby can be handled by generative AI like Suno, which whips up several tracks in no time.
I no longer manually click through each Google search link when conducting research. I just ask Perplexity. At work, I use NotebookLM and Gemini Deep Research features to handle everything from analyzing massive amounts of documents to creating reports.
It may be too early to call this infrastructure, but thanks to generative AI, the areas I can cover by myself have expanded dramatically. By paying a small fee to an AI partner, I can now substitute tasks that previously required hiring specialists at a reasonable quality level. We live in such a world.
I have already outsourced the tedious work of writing code from scratch to AI. However, there is something we must not misunderstand here. Being able to use AI and having ability as an engineer are completely different things.
The Transition from Craftsman to AI Commander
Looking back, Japanese engineers used to be a group of craftsmen. During that era, people valued the long years spent honing specialized skills.
And now? By lining up excellent generative AI agents, a single person can give them instructions and produce deliverables spanning multiple specialized fields in an unbelievably short time.
I am constantly amazed by the innovations of our time. At the same time, I feel that what one person should do and can do has exploded.
From Mimeograph to TikTok: A History Walking Alongside Tools
When I was still in elementary and middle school, from the end of the Showa era to the beginning of Heisei, it was primarily an analog era.
We barely had word processors and printers, and I remember struggling to create decent printed materials. Come to think of it, my middle school graduation anthology was mimeographed. Young people today probably do not know what a mimeograph is.
In 1995, when I joined my company, that atmosphere lingered. Presentations meant writing on transparent OHP sheets with markers and projecting them. Right after joining, company training included analog protocols like how to address internal and external mail. Email addresses were something only a select few had.
By the end of the 20th century, we could write papers in Word and create slides in PowerPoint. Entering the 21st century brought further refinement, with demands for eye-catching slides using special animation effects. It also became an era requiring design sense from engineers, and I remember the struggle of needing yet another new skill. I was glad I did design work in middle school.
Then YouTube emerged, making it normal for amateurs to publish videos. Now with just an iPhone, you can livestream on TikTok. Even my daughter does regular streams.
Every time tools evolved, individual possibilities continued to expand. And now, with the emergence of generative AI, that expansion is reaching its limits. That is an exaggeration, Gemini.
Elevated Abilities and the Invisible Current Position
However, with all this convenience comes new worries.
In the narrow village society without the internet, comparison targets were only nearby, so the goal of how far you needed to go to be recognized seemed visible.
But now it is different.
We live in a world where you can see endless geniuses if you look up. Since generative AI can easily elevate your abilities, it becomes even harder to understand your true capabilities and growth. In a sense, it has become a cruel world where it is easy to put on air.
I was fortunate to experience the transition from analog to digital and older technologies firsthand.
Old machine language and assembly were 8-bit worlds. The number of instructions was small, and memorizing them was easy. It was possible to grasp everything.
However, with modern 64-bit instruction sets, developing anything with assembly alone is madness. Things have become far too complex, and the learning method of understanding and building everything from zero is no longer realistic. To produce results in a finite lifetime, we have no choice but to skillfully utilize the black box called AI and hone our skills in compressed time.
What We Veterans Can Leave for the Next Generation
AI will do it, so you do not need to know the internals.
This is correct in a sense, but it can also hinder growth as an engineer. In an era when you can create working things without knowing the internals, what should you do to deliberately acquire skills? This is an extremely difficult challenge to consider.
So what is the role of old-timers like us?
Simply passing down stories of past hardships has no meaning. What is important is conveying the background of how each thing and technology was created and how it works.
The logic behind the code AI spits out, the underlying principles beneath convenient frameworks. Given our intuitive understanding of these concepts, we should be capable of deciphering the hidden meanings and effectively sharing knowledge.
Not just using existing AI, but how do we create new technologies beyond that? As those who know the principles, we have come to need a certain responsibility in educating the next generation and driving innovation.
To Avoid Mass-Producing Working Garbage
Here, the role of people like us who know old technologies changes dramatically. It is not intended to lecture young people about learning assembly. The goal is to use that knowledge to detect code that appears correct but is actually inefficient and dangerous—working garbage produced by AI.
AI lies without hesitation. It proposes implementations full of security holes and presents architectures that wastefully consume resources. At such times, if you have low-level knowledge—the principles of how computers actually work—in your head, you can instantly detect discomfort, like this implementation feels wrong or this coding style will become technical debt.
This aesthetic eye is the only value in this era. It is impossible to beat AI in coding speed. However, in quality assurance and architecture design, humans still have an advantage. But that is a privilege only those who know the internals can have.
Veteran Survival Strategy: From Coder to Auditor
As a reminder to myself, I defined it this way. The job of engineers in the AI era is not creation but auditing.
Knowledge of assembly and memory management does not exist for nostalgia. It is a weapon to break through modern highly abstracted layers and expose the flaws AI tries to hide.
The black-boxing of technology will not stop. Therefore, instead of retelling old war stories, we veterans should delve into the complexity of this technology and analyze its workings.
It is fine to have an AI write code. Let's also maintain a relaxed attitude. However, the moment we can no longer ask why about that output, we stop being engineers and become mere AI operators.
I ask with self-reflection.
That code of yours—do you truly understand its contents before deploying it?
Top comments (0)