Price the abstraction

Lately I’ve been thinking: modern software isn’t “too complex” by accident. It’s complex because we rewarded shipping speed over understanding.

Abstraction over abstraction has made it possible to build apps faster than ever… but also easier than ever to:

• waste memory,
• hide performance problems,
• and ship systems nobody can truly reason about.

And now AI-generated code is here—trained on that same ecosystem. If we don’t raise our standards, we’ll mass-produce the same mistakes at scale. The irony: GC languages and “easy stacks” (JS/Java, MERN, etc.) are still the most in-demand. So the move isn’t to boycott them. The move is to stop treating them like magic.

So my new goal is not to “reject” abstractions. It’s to price them.

• Use abstractions to move fast
• But understand the cost model: memory, latency, failure modes
• Measure instead of assuming
• Learn how to pop the hood when things break

I think the real differentiator going forward won’t be “who can write code faster” (AI will win that).

It will be: who can build systems that stay understandable, efficient, and reliable at scale.