Chapter 4 · When Everyone Can Make Everything
When everyone can make everything, selection is the skill.
Bad output announces itself. The harder problem is output that looks good enough to publish, good enough to stop revising, and still says nothing.
Production used to be the moat. If you could design, write, code or edit, you had something most people did not. AI is dissolving that moat across many kinds of work at once.
What remains is harder to fake: knowing what is worth keeping.
The almost-good economy
Bad output is no longer the biggest problem. Broken code fails. Wrong grammar is visible. The harder problem is output that is good enough. Good enough to publish. Good enough to convince a manager in a hurry. Good enough to make a creator feel productive.
AI can produce professional surfaces before the underlying judgment is professional. That means the ability to generate is losing value faster than the ability to reject.
The AI Taste Gap
Two people use the same image model. Both generate 500 images. The first writes ever more elaborate prompts and keeps whatever looks impressive. The second rejects 490. The first asks Can the model make this? The second asks Should this exist? One leaves with volume. The other leaves with a point of view.
Ira Glass described the gap beginners feel: their taste is good enough to see that their work is not good yet. AI creates a related gap, the distance between what you can produce and what you can reliably judge. Call it the AI Taste Gap. As production power rises, the gap can widen.
That is why more options from AI can leave you less satisfied. The tool solved generation. It exposed selection.
Build a Taste Library
Pick one area where AI multiplies your output: writing, design, video, sales pages, research memos. Collect twenty examples you consider excellent. Not five hundred screenshots you will never open again. Twenty.
For each one, write down:
- What is it trying to do?
- What specifically makes it work?
- What has been left out?
- What would a weaker version do instead?
- Where is the creator taking a risk?
- What detail would I not have noticed a year ago?
Then add examples you dislike, and explain why. Taste improves when instinct becomes language. Now bring AI in: What standards do you see in my judgments? Where am I inconsistent? Do not accept the answer automatically. The disagreements are where you learn.
Three passes for selection
When AI produces work that matters, use three passes.
- Judge it yourself. Before asking the model to critique its own work, read it unaided. What is generic? What do you not believe? What would you be embarrassed to defend?
- Invite attack. Give AI explicit standards: Evaluate this for clarity, novelty, evidence and unnecessary complexity. Do not rewrite. Name the three biggest weaknesses. Or: Pretend you are a skeptical buyer. Where do you stop believing the page?
- You decide. This is the pass people skip. A model can flatten something distinctive because a conventional version scores better on generic criteria. You own the standard.
One prompt gets you eighty percent. The last twenty percent takes eighty percent of the time, and that is where the work becomes yours.
The anti-generic test
Generative systems are very good at the statistically expected version of a thing. That is why a competent first draft can feel familiar. Ask:
- Which sentence could appear in a thousand other articles?
- Which design choice exists only because it is common?
- Which claim sounds professional but says nothing?
- What did we include because AI suggested it, not because the customer needs it?
Then remove. Originality is not the opposite of pattern. It often comes from knowing the pattern well enough to break it on purpose.
Selection takes courage
Saying yes to everything AI makes is easy. Saying no to 490 images, cutting the clever paragraph, shipping the simpler version: that takes a point of view and the nerve to defend it. AI amplifies who you already are. If you have taste, it gives you more to choose from. If you do not, it gives you more to drown in.
When production becomes abundant, selection becomes scarce.