By Rustan Ace Corpuz —

Relearning to Be a Software Engineer in the LLM Era

I've been using Docker since 2016. Nine years of docker-compose up and docker build, shipping containers to production, debugging the occasional "works on my machine" issue. I was dangerous enough to get things working, but if you asked me about the nuances of layer caching or multi-stage builds? I'd change the subject.

This wasn't unique to Docker. Looking at my skillset honestly, I realized I'd been operating at surface level across many technologies. I knew enough to ship features, but not enough to make informed architectural decisions.

With AI making code generation trivial, I've been thinking hard about how to stay relevant. My conclusion: it's time to move beyond just writing code and focus on product-level thinking and system design. The interesting part? AI is actually helping me get there.

The Docker Deep Dive That Changed My Perspective

Last month, I decided to finally understand Docker properly. Not just the commands, but the why behind everything. Previous attempts at learning from documentation had failed—either too basic or too dense, and I'd forget everything within days.

This time, I opened Claude's Research Mode and just started asking questions as my curiosity led me:

  • "Why do Docker layers exist?"
  • "What actually happens during a build?"
  • "Explain multi-stage builds with real examples"
  • "How does the build cache work internally?"

The conversational format clicked for me. I could ask follow-ups, request clarifications, and explore tangents without losing the thread. No searching through scattered blog posts or parsing dense documentation.

Within a few hours, concepts that had been fuzzy for nine years suddenly made sense.

Immediate Results

The best part? I applied what I learned that same day. Our Docker images in Google Artifact Registry were bloated—some over 2GB. Using multi-stage builds and understanding layer optimization, I reduced them by 30% to 80% depending on the project.

Before:

FROM node:18
WORKDIR /app
COPY . .
RUN npm install
RUN npm run build
CMD ["npm", "start"]

After understanding the concepts, I could intelligently prompt the LLM for optimization strategies and correct it when needed. The result was properly structured multi-stage builds that only included production dependencies.

Reclaiming the Full Engineering Role

This experience crystallized something I'd been feeling: AI isn't taking our jobs; it's giving us permission to reclaim the parts we gave away to specialists.

For years, the industry pushed us toward specialization. Frontend developers, backend developers, DevOps engineers, security specialists. We optimized for depth in narrow domains. But AI changes this equation.

When code generation becomes commoditized, the value shifts to:

  • Understanding the full problem space
  • Making informed architectural decisions
  • Thinking about business impact
  • Designing systems that evolve gracefully

These are exactly the areas I'd been neglecting while focusing on cranking out code.

My New Learning Pattern

Now I regularly use AI as a learning companion for topics I've been skating by on:

  • System design patterns I implement but don't deeply understand
  • Performance optimization beyond "add an index"
  • Security concepts beyond "use HTTPS"
  • Operational excellence beyond "add more logs"

The key is using Research Mode for actual learning, not just quick answers. I explore topics conversationally, building mental models rather than memorizing syntax.

Looking Forward

I'm genuinely excited about where this is heading. As AI handles more of the implementation details, we get to focus on the interesting problems: How should systems be designed? What are the right trade-offs? How do we build software that truly serves user needs?

The engineers who thrive won't be those who can write code fastest—AI wins that race. They'll be those who understand problems deeply, design thoughtful solutions, and know how to leverage AI as a force multiplier.

We're not becoming obsolete. We're becoming more complete engineers than we've been in years. And ironically, AI is helping us get there.