Some parts of the experience I've gained over the years suddenly became "ordinary" with AI tools. While this initially bothered me, over time I realized a completely different truth: AI, instead of making my engineering muscles lazy, allowed me to use them more efficiently. So, have I really become a better engineer, or am I just better at using the tools?
The answer to this question is complex, because the change brought by AI is not just about using a tool. For me, it meant reshaping my engineering approach and problem-solving strategies. I had to add new layers on top of my old habits.
Accelerated Discovery and First Drafts
AI tools act as an incredible accelerator, especially for routine and boilerplate tasks. Writing a systemd unit file from scratch, creating complex regexes for a new fail2ban rule, or preparing an Nginx reverse proxy configuration suitable for a specific scenario, which used to take me hours, now takes much less time. I describe the scenario to the tools, and they instantly provide the first draft.
In a production ERP, generating initial drafts of necessary PostgreSQL queries or FastAPI endpoint skeletons for a new operator screen with AI eliminates the initial "blank page syndrome." This leaves me with more time to focus on more complex workflows or architectural details. Once, when choosing a new Redis eviction policy, getting a quick summary of which policy would work better in different scenarios greatly accelerated my decision-making process.
The Changing Value of the "Why" Question
AI tools answer the "how" question very well. When you ask, "How do I do this?", it provides a roadmap. But the "why" question is still the most critical point. When suggesting a PostgreSQL index strategy, AI might offer options like B-tree, GIN, or BRIN, but understanding in which scenario an index might lead to WAL bloat or how it might affect the connection pool is possible with my 20 years of experience.
When performing network segmentation, I can ask AI for VLAN configuration. However, knowing how to optimize spanning tree settings to prevent switch loops or how DSCP/QoS markings will be handled at the company's egress is still my responsibility. AI only gives you theoretical knowledge; you learn real-world trade-offs and potential side effects through experience. For example, AI can detect an N+1 query problem in PostgreSQL, but distinguishing whether it's an ORM error or a planner's incorrect join decision requires in-depth knowledge.
New Engineering Layers Revealed by AI
With the advent of AI in my life, new engineering domains have also emerged for me. Prompt engineering has become a discipline in itself. I have to learn to "talk" to the model to ask the right question, provide the right context, and get the desired output. This is essentially a form of reverse engineering. Understanding how the model thinks and guiding it... This is also a type of architectural design.
While using different providers like Gemini Flash, Groq, and Cerebras with fallback mechanisms in the backend of my side product, I had to implement RAG (Retrieval-Augmented Generation) patterns. This means not just calling an AI model, but designing the entire information flow interacting with it, data preprocessing steps, and validation of results. This is a new distributed system architecture layer for me.
ℹ️ A New Discipline
AI tools elevate our existing engineering skills to the next level. However, this also entails acquiring new skills and continuous learning. Concepts like prompt engineering and RAG should be on the agenda of every engineer working with AI.
Critical Thinking and the Human Touch
In conclusion, have AI tools made me a better engineer? Yes, but indirectly. They have made me a faster, more efficient engineer capable of focusing on a wider range of problems. Now, instead of tedious, repetitive tasks, I can do more architectural design, more trade-off analysis, and more root cause analysis.
Once, in a customer project, I tried to solve a disk fire problem I experienced in a Docker Compose-based deployment with the help of AI. Its initial suggestions were standard log rotation or volume cleanup. But the problem was caused by OOM-killed processes constantly restarting due to containers exceeding their memory limits and overloading journald. AI alone couldn't find this deep-seated cgroup memory.high soft limit issue. When combined with my knowledge of journald rate limit and cgroups, we arrived at the correct solution.
AI offers you a gem, but the knowledge, experience, and critical thinking to process that gem are still within me. These tools are assistants, accelerators for me. Thanks to them, I have been able to direct my engineering skills to more strategic areas. Have I become a better engineer? Perhaps I have, because the change brought by AI forced me to learn more, question more, and think more deeply.
So, how have AI tools affected your engineering journey? What was your biggest "AI moment"? I'm curious to hear in the comments.
Top comments (0)