Everyone keeps saying writing code is dead. They're not wrong — but not quite right either.

What's dead is the value of typing code. AI writes functions faster than I can describe what I want. It has more patterns in memory than any engineer could accumulate in a career.

What's alive — more than ever — is knowing what to build and why.

There's a new archetype emerging. I've been calling it the meta engineer.

Not the person who writes the best code. The person with taste.

AI generates clean, functional code at absurd speed. But ask it to decide which modules to build, how they talk to each other, what to leave out — it stumbles. It optimizes locally. It doesn't see the whole board.

The meta engineer looks at what AI produced and says "no, not like that." They know that just because you can build something doesn't mean you should.

Taste is making good decisions with incomplete information. Pattern recognition plus judgment plus restraint.

Honest question I keep asking: will models develop taste too?

Right now they write perfect functions but can't answer "should this function exist at all?" They'll build you an elegant solution to a problem you shouldn't be solving.

The gap might close. But taste is downstream of values, context, and lived experience. It's shaped by watching systems fail in production. By shipping too early and feeling the pain. By choosing simplicity when every instinct says add more. That's harder to replicate than pattern matching.

And even if models get there — someone still decides what "good" means for this product, this user, this moment. That's not a technical problem.

Most engineers don't have strong taste. They build what they're told. They'll implement the spec even when the spec is wrong.

Most people with taste don't have agency. They see the problems but can't act — wrong role, wrong seniority, wrong culture.

The combination is rare. And becoming more valuable, not less.

I saw this with a 19-year-old intern at a portfolio company. He doesn't write code the traditional way. He spends hours exploring tradeoffs with AI as his sparring partner — generates five approaches, evaluates each one, picks the best fit or rejects them all and reframes the problem.

He's not grinding LeetCode. He's developing intuition. His architectural instincts are sharper than some senior engineers I've worked with. He can't always explain why something feels right, but he can feel it. That's taste in formation.

This is where it gets personal.

How do you develop taste if you don't write code yourself? Chefs taste thousands of dishes. Writers read thousands of books. You need deep exposure to the medium — but do you need to produce in it?

Can a so-called non-technical person develop technical architecture taste?

Let's find out. Because that's me.

I don't have a CS degree. But I'm actively building — making architecture decisions, evaluating tradeoffs, shipping real code with AI as my co-pilot. Not as a tourist, but as someone who lives with the consequences.

What I've learned: taste isn't about syntax. It's about structure, simplicity, and knowing what to cut. It's the instinct to ask "do we actually need this?" before building it.

But there's a floor. You need enough technical literacy to evaluate what AI produces. You can't edit a novel if you've never read one.

Can a non-technical person develop real technical taste? The answer isn't yes or no yet. It's: keep building and find out.

Writing code is dead the way typing manuscripts is dead. The physical act got automated. The thinking behind it matters more than ever.

The meta engineer isn't a worse version of an engineer. They're a different kind — and an increasingly essential one.

The question isn't whether AI will replace engineers. It's whether you'll develop the taste to direct it.

Everyone keeps saying writing code is dead. They're not wrong — but not quite right either.

What's dead is the value of typing code. AI writes functions faster than I can describe what I want. It has more patterns in memory than any engineer could accumulate in a career.

What's alive — more than ever — is knowing what to build and why.

There's a new archetype emerging. I've been calling it the meta engineer.

Not the person who writes the best code. The person with taste.

AI generates clean, functional code at absurd speed. But ask it to decide which modules to build, how they talk to each other, what to leave out — it stumbles. It optimizes locally. It doesn't see the whole board.

The meta engineer looks at what AI produced and says "no, not like that." They know that just because you can build something doesn't mean you should.

Taste is making good decisions with incomplete information. Pattern recognition plus judgment plus restraint.

Honest question I keep asking: will models develop taste too?

Right now they write perfect functions but can't answer "should this function exist at all?" They'll build you an elegant solution to a problem you shouldn't be solving.

The gap might close. But taste is downstream of values, context, and lived experience. It's shaped by watching systems fail in production. By shipping too early and feeling the pain. By choosing simplicity when every instinct says add more. That's harder to replicate than pattern matching.

And even if models get there — someone still decides what "good" means for this product, this user, this moment. That's not a technical problem.

Most engineers don't have strong taste. They build what they're told. They'll implement the spec even when the spec is wrong.

Most people with taste don't have agency. They see the problems but can't act — wrong role, wrong seniority, wrong culture.

The combination is rare. And becoming more valuable, not less.

I saw this with a 19-year-old intern at a portfolio company. He doesn't write code the traditional way. He spends hours exploring tradeoffs with AI as his sparring partner — generates five approaches, evaluates each one, picks the best fit or rejects them all and reframes the problem.

He's not grinding LeetCode. He's developing intuition. His architectural instincts are sharper than some senior engineers I've worked with. He can't always explain why something feels right, but he can feel it. That's taste in formation.

This is where it gets personal.

How do you develop taste if you don't write code yourself? Chefs taste thousands of dishes. Writers read thousands of books. You need deep exposure to the medium — but do you need to produce in it?

Can a so-called non-technical person develop technical architecture taste?

Let's find out. Because that's me.

I don't have a CS degree. But I'm actively building — making architecture decisions, evaluating tradeoffs, shipping real code with AI as my co-pilot. Not as a tourist, but as someone who lives with the consequences.

What I've learned: taste isn't about syntax. It's about structure, simplicity, and knowing what to cut. It's the instinct to ask "do we actually need this?" before building it.

But there's a floor. You need enough technical literacy to evaluate what AI produces. You can't edit a novel if you've never read one.

Can a non-technical person develop real technical taste? The answer isn't yes or no yet. It's: keep building and find out.

Writing code is dead the way typing manuscripts is dead. The physical act got automated. The thinking behind it matters more than ever.

The meta engineer isn't a worse version of an engineer. They're a different kind — and an increasingly essential one.

The question isn't whether AI will replace engineers. It's whether you'll develop the taste to direct it.