Newsletter

The Thing AI Can't Do for You

Written by Ash Roy | Apr 3, 2026, 5:00:01 AM

I think something kinda strange is happening right now.

Some of us are using AI to save 10 hours a week while others are using it to make 10 times as much noise aka AI slop.

Two very different outcomes produced by the same technology.

Turns out it's not the technology that makes the difference. It's the clarity we bring to the technology BEFORE we touch it.

AI just amplifies (and perhaps accelerates) our taste and judgement.

I find that if my thinking hasn't been done first I just tend to move faster in the wrong direction.

When I look at AI's output, I can usually feel when something's missing.

It doesn't always feel 'wrong' but it feels kinda incomplete.

I've come to think of that instinct as taste.

And taste isn't something I can prompt my way into.

Taste comes from years of doing the work, making mistakes, and developing a feel for what good actually looks like.

AI can't acquire taste on my behalf. That only comes from doing the work over decades.

That said, if I can bring taste to the table, then AI becomes a genuinely powerful tool.

I can look at what it produces, identify the gap, and direct it back to fill that gap precisely.

That's a very different dynamic than just putting prompts into Ai and accepting what comes back.

The other place AI has genuinely moved the needle for me is as a sparring partner.

Not "write me a plan." More like: "here's my thinking, tell me why I'm wrong."

When I ask AI to push back on my ideas, to show me the opposing view, to stress-test my assumptions, something useful happens. I have to think more carefully.

I have to either defend my position or abandon it. And this sharpens my thinking (rather than blunts it).

I've started calling it my mental gym. A place where my thinking gets stronger through resistance.

That's AI amplifying my thinking. Not replacing it.

Where AI actually moves the needle for me:

  • First drafts I'll reshape, not publish as-is
  • Brainstorming and pushing back on my thinking
  • Research, synthesis, and pattern recognition at scale
  • Systematising decisions I've already made
  • Reducing friction on repeatable, low-stakes work

Where it tends to quietly cost me:

  • Anything that requires taste, and taste only comes from experience
  • Strategy that needs my specific context and judgment
  • Anything that needs my voice to land
  • Work that builds my thinking over time
  • Decisions that depend on relationship or trust

The takeaway

I've come to think of AI as a brilliant assistant but a terrible owner.

I use it to move faster. I use it to think better. I use it as a sparring partner that makes my ideas stronger.

But I don't hand over the parts of the work that require discernment, conviction, and taste.

Because that's still my contribution.

And probably always will be.

Ciao for now,

Ash