It Works.
Don’t let your tool make you look like a fool.
“Building” something lends itself to a productive connotation; it feels positive, compounding. And it is.
But the narrative that almost universally gets buried when people say they are building is that they are consistently slamming into small, medium, and large barriers as they create something net new.
This has been my personal experience.
Creating something is really fun and interesting, but the trajectory is meandering.
I’ve been rolling the ball forward on a conversational diagnostic that provides a behaviorally-informed on-ramp to identifying AI use cases.
With colleagues, you have individuals with checks and balances based on their own areas of expertise. This is, relatively speaking, slower.
With Claude as my main only colleague, I can move faster. But the checks and balances are mine to make.
… And if I don’t make them, Claude makes them (up) for me… and doesn’t always (read: usually) tell me.
It’s quite possibly the most applicable “check yourself before you wreck yourself” situation of 2026 to date.
To illustrate, here’s my actual convo avec Claude:
Imagine I go out and create an experiment.
I can choose the fact that I want to study ten people, and we’re going to give five of them Questionnaire A and five of them Questionnaire B.
We’ll get those people from a specific source, maybe the people that live on my street, and we’ll run them through a specific set of questions so that we know their opinions on something.
What I just described is an experimental design, but there are a lot of holes to put in that experimental design, things like:
Why ten people?
Why did we ask those ‘specific questions’? Were they the ‘right’ ones to assess our hypothesis?
What are the implications of asking people that live on my street? Seems biased.
I’ve had a similar experience setting up this product with you. To the extent that we have pushed forward, unless I ask why, we kind of just keep pushing forward.
Sometimes you’ll flag stuff, but in general it’s up to me to push back question decisions leaning on the fact that there might be other ways to do things that are superior that I’m just not aware of.
We spend so much of our AI prose venting about slop, which is whether the output is good (buyer beware).1
But where did the output come from?
This question, I think, gets asked less often.
And when we don’t ask it, we risk thinking the output is good when it is in fact not.
We have to define what “good” means before we can claim our product works.
Then we can talk about whether the output is good.
Tina Huang (aka my pseudo-celebrity AI crush) recently referred to this as taste:
As AI gets better at generating things — writing, code, designs, strategies — the question isn’t really “can AI do this?” anymore. It’s “Can you tell the difference between good output and bad output?”
Can you look at something AI made and know if it’s actually good? Or just looks good on the surface? That’s taste. And I think it’s becoming one of the most valuable things you can have.
But here’s the thing about taste. You can’t shortcut it. The only way to develop taste in something is to actually learn that thing deeply… AI makes the producing part faster. But the judging part? That still comes from you.
Deep knowledge + AI fluency is where the real edge is.
When it comes to AI outputs, beauty is certainly not in the eye of the beholder. It’s in the raw materials from which it was constructed.
Knowledge workers who can generate hypotheses and stress test their own critical thinking are, in my opinion and surely of others (#frontierprofessionals), in the hot seat to adopt an AI sidecar that can help them go fast and far.
Many have already figured this out, particularly if they work in tech or technically adjacent fields.
But for what it’s worth, a lot haven’t.
The narrative that AI is coming for your job is an emotional one that likely distracts. I quoted Tim Pychyl a few days ago, and I’ll do it again here:
“Procrastination is an emotion regulation problem, not a time management problem.”
Stories we tell ourselves impede us from taking action.
But imagine if we (all) did.