The Creative Process With AI: It's Not All Prompts and Speed

The Part Nobody Talks About

The creative process with AI isn't what the tutorials show you. It's not a clean demo where someone types a prompt and gets a perfect website in 30 seconds. The real workflow — the one that produces genuinely unique work — has phases that look nothing like prompting. Some of them don't even involve a computer.

I've been building this website with AI as my creative partner. Not a template generator. Not a code autocompleter. A collaborator I direct, challenge, and occasionally argue with. The experience taught me something nobody in the "AI will replace developers" discourse mentions:

The best work happens in the spaces between the prompts.

The Four Phases of AI-Powered Creative Work

Most creative AI workflows follow four distinct phases. Each one requires a different mindset, and skipping any of them produces mediocre work:

Phase	What It Looks Like	What It Produces
Sprint	Feature after feature, high velocity	Rapid execution of clear ideas
Stare	Scrolling your own page without thinking	Taste-driven decisions and gut feelings
Download	Walking away from the computer entirely	Breakthrough ideas from your subconscious
Expand	Coming back with vision and conviction	Directed, ambitious creative sessions

The sprint is the only phase that involves AI. The other three are entirely human — and they're what make the sprint productive.

The Sprint Phase: Feature After Feature

When you have crystal clarity on what you want, AI collaboration becomes pure velocity. You describe a feature. You see it. You refine it. You move on. Feature after feature, component after component.

On this site, some sprint sessions shipped five interactive features in an afternoon:

• The fleeing blog cards that dodge your cursor
• The AI companion that comments as you read
• The hidden puzzle system woven across the site

But here's the thing: you only get sprint sessions because of the other phases. The sprint is the payoff, not the process. Sprint without slow thinking, and you build fast and mediocre.

The Stare Phase: Looking Without Thinking

This is the phase nobody talks about because it looks like doing nothing.

You're scrolling your own page. Not reading. Not analyzing. Just feeling it. Is something off? Does the flow feel right? Your attention drifts without purpose, and occasionally — if you're lucky — it snags on something.

This is where taste lives. AI can't replicate it because it requires:

• Absorbed experience — thousands of websites you've visited, unconsciously catalogued
• Emotional pattern recognition — the instinct that says "the spacing is fine but the rhythm is wrong"
• Non-verbal judgment — knowing something is almost right before you can articulate what's missing

The stare phase isn't laziness. It's the most human part of the process. No prompt, no model, no framework can replicate the gut feeling of a creator looking at their own work and sensing what needs to change.

The Instant Feedback Loop: Why Speed Changes What You Try

Before AI tools, the creative feedback loop in web development was punishingly slow. With AI, it compresses dramatically:

Traditional workflow:

1. Have an idea
2. Spend days to weeks building it
3. Look at it
4. Realize it's not what you pictured
5. Decide: start over or ship something mediocre

AI-assisted workflow:

1. Have an idea
2. Describe it
3. See a working version in 60 seconds to one hour
4. "No, make it feel more like a terminal"
5. See the new version
6. "Actually, make the whole form a conversation"
7. See that version too

Why This Changes Creativity

This compression doesn't just save time — it changes what you're willing to try.

When building takes weeks, you hedge your bets. You pick the safe idea. You don't experiment because failure is expensive. When you can see a concept in under an hour? You try the weird idea. The risky one.

"What if the contact form is a terminal that talks to you?" — that's not a thought you pursue if it costs a week to prototype. But when testing costs an hour, your creative ceiling goes way up.

You stop asking "will this work?" and start asking "what if?"

The Cost Is Real

It's not free, either. Every prompt, every iteration, every "actually make it more like this" — that's API costs ticking with every keystroke.

Cost Model	What Changes
Traditional dev	"I can't afford to try this" — failure costs weeks
AI-assisted dev	"I can afford to try this, but I'm paying per attempt" — failure costs dollars

That distinction matters. It keeps you intentional. You still think before each swing because each swing has a cost. AI didn't make creativity free — it made it affordable enough to be brave.

The Download Phase: Walking Away

You close the laptop. You go for a walk. You cook dinner. And somewhere between the pasta sauce and the shower, it clicks:

"What if the whole form was a terminal?"

This is your subconscious processing. It needs two inputs:

1. Raw material — from the staring phase
2. Idle time — from physically walking away

Your brain keeps working on the problem in the background, connecting dots you didn't know you had. AI can't do this. It doesn't have a subconscious. When you close the chat, the AI stops thinking. When you close your laptop, your brain is just getting started.

The most productive thing you can do for a creative project is sometimes nothing at all. Step away for a day or two. The ideas that survive the walk are always better than the ones from hour six of a marathon session.

The Expansion Phase: Planning With Vision

Now you're back — and you're dangerous. You've stared, you've walked, you've fermented. Now you know exactly what you want and why.

This is where you direct the AI with conviction:

• ❌ "Build me a contact form" — generic, gets generic results
• ✅ "I want the contact form to feel like a conversation. The AI asks questions one at a time. The user types answers. There are witty reactions." — specific, gets specific results

In vision engineering, this is the moment when vague ideas crystallize into clear creative direction. The AI becomes a power tool — but you're the one pointing it at the right wall. No model improvement changes that.

You Get Paid for the Training, Not the Race

A marathon runner doesn't get paid for the 20 minutes at the finish line. The career is built on 10,000 hours of training nobody saw.

Web development with AI works the same way:

	What It Is	Who Does It
The race	Deploying the feature	AI executes
The training	Staring, walking away, three rejected ideas	You
The coach	Analyzing form, suggesting improvements	AI assists

There's a measurable part — pure hours, output, billable. Then there's the creative part that's harder to measure: the taste, the vision, the decade of experience that lets you look at a design and know something's wrong before you can say what.

Both parts are essential. Only one shows up on a timesheet.

What AI Actually Can and Can't Do

If AI could completely replace human creativity, it would have by now. The models are powerful. But every creative AI project that works has a human making the decisions that matter.

AI Excels At	AI Cannot Do
Execution speed — ships features fast	Taste — pick the option that feels right
Breadth — knows more patterns than you	Conviction — tell you your idea is bad
Consistency — follows patterns precisely	The blank page — sitting with uncertainty
Iteration — produces variants quickly	Subconscious processing — walking away
Code quality — structured, typed, tested	Cultural context — what resonates with humans

The Rhythm Is the Skill

The creative process with AI isn't a single mode. It's a rhythm: sprint, stare, walk away, come back with vision, sprint again — harder, weirder, braver.

The people who get the most out of AI aren't the fastest prompters. They're the ones who know when to prompt and when to close the laptop. They've developed the taste to evaluate AI output, the vision to direct it, and the courage to try ideas that sound ridiculous until they work.

AI didn't make creativity free. It made it affordable enough to be brave. And bravery — that's still a human trait.

---

But there's a deeper question behind all of this — what exactly makes the human contribution irreplaceable? We explored that in You Blow the Soul Into It — about what happens when you lend a piece of yourself to the machine.

This is part of our AI Productivity series — real workflow patterns, not prompt tutorials. See also: Stop Prompting, Start Planning and Clearer Vision, Not Better Prompts.