I’ve been spending a lot of time with AI image models lately, trying to get good ad creatives out of them.

One thing keeps coming up:

Writing prompts feels weirdly similar to practicing Daoism.

Most people (including past me) assume: if you want a “perfect” image, you must control everything:

lens, lighting, angle, composition

every style tag under the sun

10 adjectives for mood + 5 for color + 3 for vibe

Basically: design the whole image in your head, then force the model to follow.

What usually happens?

The prompt is heavy, the intention is blurry, and the image looks stiff and over-produced. Exactly the opposite of what we wanted.

This is where Daoism quietly laughs at us.

1. Over-controlling the model = going against the “Way”

Daoism has this idea:

“The Dao does nothing, yet nothing is left undone.”

It’s not about being lazy. It’s about not fighting the natural flow of things just because we’re anxious.

Translate that into prompting:

We try to control every tiny variable

The model receives a wall of instructions, many of them conflicting

It squeezes out something that technically “checks all the boxes” but feels… dead

Too many rules = no real direction. You’re basically telling the model: “Do everything, in every style, all at once.”

2. Good prompts are more “go with the flow” than “command the universe”

Now when I look at a prompt, I ask myself:

Is this a clear intention, or is this a control freak moment?

The “control freak” prompts usually:

are very long

try to lock down every parameter “just in case”

come from fear: “if I don’t specify this, the model will ruin it”

The “go with the flow” prompts are different:

One clear idea: what is this image really about?

A few key constraints: scene type, mood, rough composition

Then space for the model to do its job

And the funny thing is:

Short, sharp prompts often give me images that feel more natural, human, and alive.

3. Daoism for prompt engineers: very practical version

If I compress all this into something usable, it’s probably:

① One center

Pick one core priority:

Is this about the product look?

The real-life scene?

The emotion on the person’s face?

Choose one. A single image can’t carry five main messages without breaking.

② Know when to stop

Every time you add a line to the prompt, ask:

If I remove this, will the model still understand what I want?

If the answer is “yes”, you probably don’t need it.

Most good prompts can be explained in 5–7 key pieces of information, not 25.

③ Leave room

Your job is to set direction, not micro-manage:

“Morning light, casual kitchen scene, young person using X for their daily ritual” is usually enough.

You don’t have to decide every highlight and shadow. Some of the “happy accidents” from the model are exactly what make the image feel alive.

4. Follow the “nature” of the model instead of fighting it

The model has its own “Dao”:

It’s very good at expanding from a clear, slightly open-ended idea

It’s very bad at satisfying 10 pages of strict, sometimes conflicting rules

When we prompt from a place of panic (“I must specify everything or it will mess up”) we are basically fighting the system.

When we prompt with more Daoist energy:

clear

restrained

directional

with space

we usually get results that are more:

natural

believable

and actually closer to how humans see the world

Closing

Prompting looks like a purely technical skill from the outside. But the more I do it, the more it feels like a mirror:

How much do you need to control?

How much can you let go, once the direction is right?

Personally, I’m trying to move from “control-everything prompts” → “Daoist prompts”.

Curious where you are on that spectrum. If you’ve noticed similar things in your own prompting, I’d love to hear it.