AI Is a Great Developer, Not an Architect

I genuinely think of AI today as an amazing mid-level developer. It’s fast, tireless, and surprisingly good at turning clear requirements into working code.

Ask it for:

• A CRUD API
• A React component
• A unit test suite
• A refactor of a messy function

and it will usually produce something solid enough to run and iterate on. But “this works” is not the same as “this is the right way to build it.” Architecture lives at a different layer.

What Fundamentals Actually Mean to Me

When I say fundamentals, I’m not just talking about data structures and algorithms. I mean the habits and mental models that shape how I build software, with or without AI.

For me, fundamentals include:

• Understanding how the web actually works - HTTP, state, caching, security basics
• Knowing how to design boundaries - modules, services, layering, separation of concerns
• Being able to reason about complexity - what breaks when this grows 10x
• Reading and understanding code you didn’t write - including AI-generated code

These are the things that decide whether a project is maintainable, safe, and scalable. No AI can do that thinking for me. It can only assist with implementation after I’ve made those calls.

Where AI Shines (And Where It Doesn’t)

AI shines when:

• The problem is well scoped
• Patterns are well known - CRUD, common UI patterns, typical integrations
• The main work is turning a description into code

In those cases, AI feels almost unfairly strong.

But it struggles when:

• Requirements are fuzzy, political, or half-baked
• You need to prioritize tradeoffs - performance vs simplicity, speed vs robustness
• You are designing a system meant to live for years with many teams touching it

That is the realm of architecture - and that’s exactly where fundamentals matter most.

Why I Still Study Fundamentals Even as AI Improves

Caring about fundamentals is not nostalgia. It’s risk management.

Without strong fundamentals:

• I can’t tell when a “working” solution is actually fragile
• I can’t see over-engineering or under-engineering
• I can’t debug subtle issues outside common patterns

With fundamentals:

• I can use AI as a power tool, not a crutch
• I can quickly reject bad suggestions and refine good ones
• I can design systems where AI-generated code has a clear, safe place to live
  The better I am, the more value I can squeeze out of AI.

The Architect’s Job Isn’t Going Away

Architecture is about:

• Understanding the domain deeply
• Choosing the right boundaries, protocols, and data flows
• Setting constraints a team can live with for years
• Anticipating failure modes and growth paths

AI can help explore options, draft diagrams, and prototype designs. But deciding which design to bet on - and why is still a human responsibility. That’s why fundamentals still matter in the AI era.

AI is an amazing developer. But it’s not a good enough architect.

And until it is, the ceiling of what we can build will still depend on how well we think - not just how fast we generate code.