DEV Community

Cover image for We Let Users “Create Moments” With AI — Here’s What We Learned
Cloyou
Cloyou

Posted on

We Let Users “Create Moments” With AI — Here’s What We Learned

The idea sounded simple… until we tested it

In the previous post, I talked about moving AI from just “responding” to actually “participating.” That idea became Aaradhya on CloYou.

But the interesting part wasn’t the idea. It was what happened when people actually started using it.

Because once you move beyond answers and let users create moments, the system behaves very differently.


People don’t use it like a tool

One thing became clear quickly: users don’t treat this like a normal AI tool.

They don’t come in with:

  • structured prompts
  • specific tasks
  • “optimize this” mindset

Instead, they do things like:

  • “let’s create something together”
  • “imagine this moment”
  • “what if we try this scene”

It’s less like using software, more like exploring something.


The role of image upload changed everything

We initially thought image upload would be a small feature.

It wasn’t.

Once users could upload their own image:

  • they became part of the generated scene
  • identity started to matter
  • outputs felt less random

This shifted the system from:

generic generation → personalized experience

And that’s a big difference.


Consistency is not a feature — it’s the system

Most generative systems fail at one thing: consistency.

You can generate something impressive once, but across multiple interactions:

  • faces drift
  • styles change
  • nothing connects

We realized quickly that without consistency, the entire idea breaks.

So we focused on:

  • keeping the AI character stable
  • aligning outputs with the user’s identity
  • making each generated moment feel related to the last

Without this, you don’t have an experience. You just have outputs.


Memory had to be intentional

Another thing we tested was automatic memory.

At first, it sounds like a good idea: just save everything.

In practice, it becomes noise.

So we switched to a simple model:

  • user creates a moment
  • system generates it
  • user decides if it should be kept

This keeps memory:

  • clean
  • relevant
  • user-controlled

And it changes how people value what they create.


Recognition made the system feel aware

One unexpected layer came from recognition.

When users uploaded images where the AI character was already present, the system could identify that context.

This added something subtle but important:

  • awareness of the scene
  • continuity across interactions
  • stronger connection between input and response

It didn’t make the system “intelligent” in a new way, but it made it feel more consistent.


The interaction model is different now

If you look at the full loop, it’s no longer:

input → output → done

It becomes:

  • conversation
  • imagination
  • generation
  • optional memory
  • continuity

That loop keeps going.

And that’s what makes it feel different.


This is where Aaradhya fits in

Aaradhya isn’t just a chatbot layer on top of a model.

It’s a combination of:

  • conversational interface
  • identity system
  • visual generation pipeline
  • user-driven memory

All working together.

You don’t just get answers. You build something across interactions.


What this means going forward

We’re starting to see a shift in how AI systems are used.

Not just for:

  • solving tasks
  • generating outputs

But for:

  • creating experiences
  • maintaining continuity
  • building interaction over time

This is still early, but it points toward a different direction.


Where we’re building this

This is part of what we’re exploring with CloYou.

Not replacing traditional AI systems, but extending them into something more interaction-driven.

Aaradhya is one implementation of that idea.


Final thought

AI is already good at answering.

The next step might be making interactions feel like they actually go somewhere.


🚀 If you want to try it

You can explore it here: https://cloyou.com

Try a normal conversation, but instead of asking something useful, try creating a moment.

That’s where the difference shows up.

Top comments (0)