You don't have to be ignorant to work here, but it helps

Apr 20, 2026

Recently[1] in a freeform discussion in the Redwoods Design System community, the topic had once again turned to the impact of AI on the work we do.

As usual, I was doing my best impression of an old man yelling at a cloud, when the following popped out:

I trust humans to be ignorant better than I trust AI to be ignorant.

This extremely engaging conversation had started out with another classic Design System conundrum - just how much should a designer know about coding, and vice versa.

I have the fortune of working on a phenomenal team at HP with colleagues all across the spectrum of design and development, but I maintain that having people with gaps in their expertise of Design Systems is a Good Thing™.

Two of the twenty theses of The Zen of Python are as follows:

If the implementation is hard to explain, it's a bad idea.

If the implementation is easy to explain, it may be a good idea.

In HP we've been developing and exploring some fairly radical concepts in the Design System space, and much of the work in this has been working out how to explain it.

This isn't just to torture my colleagues[2], but to help evaluate whether or not it's actually a well-designed architecture. If I can't explain it to the writing team, how are we going to document it for our users? Is it even a good idea?

As ever, AI reared its ugly head in the conversation. What happens when you try to explain something to AI, or when you ask to document a feature?

Well, one of the things AI is quite good at is comprehending an extremely large context, so it will probably, in its usual sycophantic way, produce comprehensive and possibly even accurate documentation of your feature, irrespective of whether it's good or not.

Removing the human technical writer from that loop removes not only the skill and expertise that that human will have, but the equally valuable ignorance[3].

AI can make for a passable simulacrum of skill, but ignorance is much harder to replicate[4], and even harder to verify.

Replacing your colleagues with AI doesn't just devalue the work they do, you're missing out on that most uniquely human quality of ignorance.

I trust the ignorance of humans.

  1. Actually, not that recently really, as I'm really bad[5] at procrastinating ↩︎

  2. Though they may disagree on that point ↩︎

  3. You know who you are, and you know that I love you ↩︎

  4. I recall an anecdote from the development of Steve Davis World Snooker, the first snooker game to feature AI. After creating a fiendishly good computer opponent, it turned out to be much harder to make it bad at playing snooker. ↩︎

  5. Well, good, technically ↩︎

Hyrum's Law in Design Systems