The AI Creativity Paradox: Why Smarter Isn't Always More Creative

Building Maya

My wife went to India for a month to train as a yoga teacher. When she came back, she wanted a good app for building and planning yoga lessons. The existing tools were clunky.

So I built her one.

I created Maya, an AI yoga instructor that could generate sessions based on what users asked for. Tell her you’re stressed and she’d build a calming sequence. Say you want energy and she’d design a morning flow.

Maya in action

Click “Play Demo” to see what a conversation looks like:

Maya

AI yoga companion (Demo)

Play Demo

Want to try Maya yourself? Check out the live project at joshdesk.live

It worked. Then my wife tried it.

The Reality Check

“That’s not how yoga works,” she said, with the patience of someone who’s watched their partner overthink a simple problem.

“You’re treating yoga like it’s just a collection of poses to arrange. But yoga sequences tell a story - they have purpose, progression, meaning. Sun Salutations aren’t random poses; they’re a complete practice refined over centuries.”

She was right. I’d built a pose-ordering system, but yoga sequences aren’t Lego blocks. They’re more like music: notes form phrases, phrases form compositions.

Back to the drawing board.

The Rebuild: Understanding Structure

Two weeks later, I had completely rewritten the data model.

Maya got smarter too. Instead of just creating sessions from scratch, the agent could now search through proper sequences, understand traditional flows, and suggest existing practices.

When success creates new problems

As Maya got better, I started planning the next step: creating Classes, combinations of multiple sequences that form a complete practice.

If Maya can create sequences, should she also search for and reuse existing ones? Of course. Why recreate what already exists?

But watch what happens when she gets good at reusing:

Maya

AI yoga companion (Demo)

Play Demo

In the second demo, Maya keeps finding existing sequences but never creates anything new.

The efficiency trap

AI systems tend toward reuse. Given the choice between creating something new and recycling something that works, they’ll pick reuse. It’s faster, safer, and the immediate results are often better.

But creativity requires trying things that might be worse. As Maya gets better at finding existing sequences, she stops making new ones:

• Efficiency: “I found a perfect hip-opening sequence for evening practice. I’ll use that.”
• Creativity: “This user mentioned they’re feeling anxious about work. Let me create a custom sequence that addresses that specific emotional state.”
• Personalization: “They said they have 22 minutes. Let me adjust this existing sequence to fit exactly.”

All three are valuable, but they pull in different directions.

The lazy AI problem

This isn’t specific to yoga. Give any AI system the ability to reuse content and it drifts toward conservatism, picking the safe option over the risky one.

I’ve seen it in other domains:

• Code generation: AI that always suggests the same boilerplate instead of exploring new patterns
• Content creation: Systems that remix existing content rather than generating truly novel ideas
• Design tools: AI that converges on safe, familiar patterns (you’ve seen the “AI look”)

Possible Solutions (None Perfect)

I’ve considered several approaches to this problem:

Preventing AI laziness

The human version of this

Artists and designers face the same tension. A veteran illustrator knows what works, so they keep doing it. A junior tries weird things because they don’t know better, and sometimes the weird thing is better. Humans push through that because they get bored, or they want to grow, or they care about making something new. AI has no such drive. It optimizes for whatever metric you gave it, and reuse scores well on every metric except novelty.

Open questions

Maya is in limbo while I figure this out. She works well, but I don’t want to make her more capable until I have an answer for the creativity problem. A few questions I’m sitting with:

• Should AI systems have a built-in drive toward novelty?
• How do you measure creativity alongside quality?
• When is reuse the right call and when is it laziness?
• How do you build systems that evolve instead of just optimizing?

I don’t have answers yet. Maybe hybrid systems, where one agent optimizes for reuse and another is forced to create. Maybe better metrics that reward novelty. Maybe you accept that some inefficiency is the cost of making something new. Or maybe the right answer is simpler than all of that: build the creative path and make it the default, so the system has to opt out of novelty instead of opting in.

I’ll update this when I figure it out.

---

Technical Details: Maya’s stack:

• Runtime: Bun
• Backend: Elysia.js with Swagger for self-documenting APIs
• AI: Google’s Generative AI SDK (@google/genai) with function calling
• Frontend: React Router v7 (framework mode) with full SSR
• Database: Turso (libSQL) with Drizzle ORM
• Type Safety: EDEN for end-to-end type safety without codegen
• Forms: TanStack Form with Zod validation
• UI: Radix UI components with Tailwind CSS v4
• Auth: Clerk for user management
• Hosting: Self-hosted on a Digital Ocean Linux droplet

The creativity vs. efficiency dilemma emerged during the transition from generating individual sessions to creating multi-sequence classes.

Live Project: You can view the live version of the project that inspired this post at joshdesk.live.