
Weekend Showcase Challenge: Daily Muse
A small AI project that started as a weekend build grew into something much more personal. I refined Daily Muse with new ways to generate, save, and revisit inspiration while learning a lot more about building, securing, and shipping on AWS.
A week ago, I built Daily Muse for the “Set Your Creative App Free” challenge, a small, always-on AI agent that generates an original quote and reflection every day based on a randomly chosen theme and publishes it to a little webpage.
When I saw the brief for the Showcase Challenge, I honestly didn't have to think twice about which project to pick.
There were still so many things I wanted to add to Daily Muse the first time around, but simply didn't have enough time for.
So this weekend, I opened it back up and thought:
“Okay... what can I actually make better this time?”
And well... quite a few things happened. 😅
So... what's different now?
The core idea is still the same.
Daily Muse wakes up once a day, creates an original quote and a short reflection, and publishes it.
But this time, I wanted the whole experience to feel less like a demo and more like a little place someone would actually want to visit.
So I gave it a theme:
The Magic Shop. 🌙
I wanted it to feel like a warm, comforting little space where you could come in, slow down, and stay for a while.
There's now a crescent moon, a twinkling star field, floating glass-like orbs, wandering fireflies, and calming ambient music that you can turn on or off.
The atmosphere also changes depending on the time of day. During the day, it's soft and lavender. At night, it turns into a deeper, warmer, starry little world.
I also changed the themes the agent works with.
Alongside resilience, growth, courage, patience, curiosity, gratitude, and change, I added themes that feel a little more personal:
self-love, healing, belonging, coming home to yourself, embracing imperfection, and quiet comfort.
I wanted the agent to sometimes create something that feels like the kind of words you'd want to come across on a difficult day.
Still original quotes.
Always.
And then I added a few things I wanted myself
The first version only had the daily quote.
Now there's a Generate Another button.
If today's muse isn't quite what you're looking for, you can simply ask for another one.
That quote gets added to the shared gallery too, with a label showing whether it was the daily muse or something generated on demand.
But the feature I'm probably most attached to is the Personal Collection.
I have a ridiculous number of quotes saved on my phone, so naturally, I wanted somewhere to keep the ones I actually love.
The collection lets me save favourites privately, and the agent can sometimes pick something from that collection instead of generating something completely new.
When it does, you'll see “From Your Collection” instead of “Today's Muse.”
I also experimented with pairing those saved quotes with AI-generated images.
Since image models and availability have a habit of changing when you least expect it, I made sure the feature can gracefully fall back if an image isn't generated.
And yes, you can actually try the collection feature.
I didn't want people reading this article to just see a screenshot and move on, so I created a separate demo environment.
Use the demo passphrase:
trydailymuse It points to a completely separate sandbox collection, so you can add and view quotes without touching my actual personal collection.
Building the second version
Most of this weekend wasn't really about building everything from scratch again.
It was more like opening up the first version and going:
“Yeah... this needs fixing.” 😭
One of the biggest changes was the gallery.
The first version assumed there would only be one quote per day because the scheduled agent was the only thing creating content.
Now there can be multiple entries on the same day because of the Generate Another feature.
So I had to rethink how everything was identified, stored, and labelled.
Not exactly the most exciting part of the project, but definitely something I needed to get right before adding more features.
Then came the Personal Collection.
I wanted the collection to actually be private while still giving people a way to try the feature.
So I created a separate sandbox file for the demo instead of exposing my real collection.
Sounds simple enough, right?
Well...
I managed to break my own Lambda function while trying to secure it. 😭
I added an S3 policy to prevent public access to my personal collection, which worked perfectly...
Except it also stopped my Lambda function from reading the file.
Turns out an explicit
Deny with Principal: "*" doesn't care that you're your own backend.A deny is a deny.
I initially tried using
NotPrincipal to make an exception for the Lambda role, but things got a little messy with how assumed-role sessions are handled.The cleaner solution was to change the public
Allow statement using NotResource so that the private file was excluded from public access in the first place.Much simpler.
And honestly, that was probably one of my favourite lessons from this weekend.
It's easy to make something work.
Making it work without accidentally exposing the data behind it is a completely different thing.
I also tried going further
I started working on proper user accounts using Amazon Cognito.
The idea is eventually to let Daily Muse remember individual visitors, save their preferences, and make the experience a little more personal.
The architecture I planned uses Cognito for authentication, DynamoDB for storing preferences, and API Gateway with JWT authorization to properly protect requests.
I got quite a bit of the planning done...
And then the weekend happened. 😅
I ran out of time before I was comfortable calling the authentication system finished.
And honestly, I'm okay with that.
I could have rushed together a half-working login system just so I could say I added authentication.
I'd rather leave it for the next version and do it properly, especially when actual user data is involved.
Sometimes not finishing a feature is the better engineering decision.
AWS services behind Daily Muse ☁️
The architecture has grown a little since the first version.
- AWS Identity and Access Management (IAM) - secure account access and scoped permissions
- Amazon Bedrock / Amazon Nova Lite - generates the original quotes and reflections
- AWS Lambda (Python) - handles daily generation, on-demand generation, and collection actions
- Amazon API Gateway - exposes the Lambda actions to the website
- Amazon S3 - stores generated content and collections and hosts the static website
- Amazon EventBridge Scheduler - triggers the daily generation automatically
The main automated flow looks like this:
EventBridge Scheduler → Lambda → Amazon Bedrock → S3
For the interactive features:
Browser → API Gateway → Lambda → Bedrock / S3
Both ultimately feed into the same website and shared gallery.
The next version would extend this architecture with:
Cognito → API Gateway → Lambda → DynamoDB
for authentication, user preferences, and personalised experiences.
What I learned from taking another look at this project
The first version of Daily Muse taught me how to connect AWS services and actually get something working.
This version taught me to think beyond:
“Does it work?”
I had to think about security, data, failure cases, user experience, and what happens when someone other than me starts interacting with the project.
I learned that putting a password on a page doesn't automatically make the data behind it private.
I learned that sometimes the cleanest solution is to rethink the approach instead of adding more and more layers to fix the first one.
I learned that building something for other people to try makes you notice a lot of things you wouldn't notice when you're the only user.
And probably my favourite lesson:
Polishing is also building.
I didn't create an entirely new idea this weekend.
I took something I had already built, looked at everything that felt unfinished, and tried to make it a little more thoughtful, a little more reliable, and a little more me.
And somehow, that was just as satisfying as building it from zero.
A little inspiration
A shoutout to Lewis Sawe , whose The Museum That Grows was one of the projects that stood out to me during the previous challenge.
I loved the idea of an always-on creative agent that keeps adding to its own little world over time, and it made me think differently about what Daily Muse could become beyond generating one quote a day.
It's always fun to see how differently other builders interpret the same challenge. That's probably one of my favourite parts of these community challenges.
What's next?
Proper user accounts are still on the list: Cognito authentication, saved preferences, and eventually a more personalised Daily Muse experience.
I already have a pretty good idea of how I want to build it.
I'm just not going to rush the part that deals with people's actual data.
For now, I'm happy with where this little project ended up.
It started as a weekend challenge.
Then it became something I wanted to keep building.
And now it has its own little world.
Welcome to The Magic Shop.
Come in, stay a while. ☁️🌙
Try it yourself
Demo passphrase:
trydailymuse Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article