
Weekend Creative Agent Challenge: Dear Tomorrow
What if you could send a message to the person you are going to become? That was the idea behind Dear Tomorrow, my project for the Weekend Creative Agent Challenge.
Weekend Creative Agent Challenge: Dear Tomorrow
What if you could send a message to the person you are going to become?
That was the idea behind Dear Tomorrow, my project for the Weekend Creative Agent Challenge. I wanted to create something simple, but also personal: a time capsule that lets people leave an encouraging message for their future selves and receive it at exactly the moment they choose.
Sometimes we go through difficult periods and wish that our future selves could hear a few words of encouragement. Sometimes we want to celebrate a future milestone, such as graduation, a new job, or the completion of an important goal. Other times, we simply want to imagine what our future might look like and leave a message for that version of ourselves.
Dear Tomorrow was created around that idea.
🌅 Vision and What It Does
Dear Tomorrow is an AI-powered time capsule service. A user writes a message, chooses a future delivery date and time, and provides the recipient's email address. The message is stored securely in the cloud and delivered by email when the scheduled time arrives.
The basic experience is intentionally simple:
Write a message ↓Choose a future date ↓AI transforms the message ↓Store the capsule ↓Wait ↓Receive the message in the future
The service can be used for many different situations.
You can write a message to yourself when you are having a difficult day and ask your future self to remember that you made it through. You can write a graduation congratulations message and schedule it for the day you expect to graduate. You can write down your current goals and ask your future self to reflect on whether you achieved them. You can also send a time capsule to another person as a meaningful future message.
The goal is not simply to store a note.
The goal is to create a moment in the future when someone opens their inbox and receives something from the person they used to be.
🤖 Adding AI to the Experience
I wanted the project to do more than simply save and resend text.
Dear Tomorrow uses Amazon Bedrock to transform the user's original message into a warmer message written from a third-person perspective. The idea is that the final message should feel as if someone who has been watching the user's journey is speaking directly to their future self.
For example, a user might write:
"Today was really difficult, but I kept going. I hope I am happier when I read this."
The AI can transform it into a more personal message while preserving the original meaning.
One of the most important decisions was to store both the original message and the AI-generated message. I did not want the AI to overwrite what the user actually wrote. The original message remains available in DynamoDB while the transformed version is used as the final message.
I also spent time improving the prompt. The first outputs were sometimes too creative and invented details that were not present in the original message. For example, the model could assume that the user was sick or had experienced a specific event just because the user said they had a difficult day.
I changed the prompt to explicitly prohibit invented events, people, memories, and circumstances. I also instructed the model to preserve the user's emotional intent, avoid repetitive encouragement, and produce a complete message rather than an overly dramatic AI-generated speech.
That experience reinforced an important lesson for me: good AI behavior depends not only on the model, but also on clear constraints and careful prompt design.
🏗️ How I Built It
I chose a serverless architecture because this project did not need a continuously running backend server.
The initial architecture was:
Frontend ↓CloudFront + S3 ↓API Gateway ↓Lambda ↓DynamoDB
After the basic application worked, I added Amazon Bedrock to the capsule creation Lambda.
The capsule creation process now works approximately like this:
- The user submits the recipient email, subject, delivery date, and message.
- API Gateway sends the request to Lambda.
- Lambda validates the required fields and verifies that the delivery date is in the future.
- Lambda sends the original message to Amazon Bedrock.
- Bedrock returns the transformed message.
- Lambda stores both the original and AI-generated messages in DynamoDB.
- A separate Lambda checks for capsules whose scheduled delivery time has arrived.
- The ready capsule is sent through Amazon SES.
- DynamoDB is updated from
pendingtosent.
For the AI component, I used Amazon Nova Micro through Amazon Bedrock's Converse API. This provided a simple way to integrate model inference directly into the serverless backend.
☁️ AWS Architecture
The main AWS services used in the project are:
- Amazon S3 — static frontend hosting
- Amazon CloudFront — HTTPS and content delivery
- Amazon API Gateway — HTTP API endpoint
- AWS Lambda — application logic and scheduled email processing
- Amazon DynamoDB — capsule storage and delivery state
- Amazon Bedrock — AI-powered message transformation
- Amazon SES — transactional email delivery
- AWS Certificate Manager (ACM) — HTTPS certificate
- Route 53 — explored during DNS configuration (Domain/Korean DNS provider (내도메인.한국) — domain registration and DNS record management)
The architecture can be summarized as:

For HTTPS, the application uses an ACM certificate associated with CloudFront. The Lambda, Bedrock, and SES components are deployed in the Asia Pacific (Sydney)
ap-southeast-2 region.🔧 Challenges
One of the biggest challenges was not actually writing the application code. It was connecting all the cloud services correctly.
Custom domain configuration required understanding the relationship between domain registration, DNS records, ACM certificate validation, and CloudFront alternate domain names.
Email delivery introduced another set of challenges. Amazon SES requires identity verification and proper email authentication. I configured a domain identity for
dear-tomorrow.kro.kr , enabled Easy DKIM, and added DNS records for authentication. I also requested SES Production Access because the service is intended to send transactional messages to user-selected recipients rather than only verified test addresses.Another challenge was learning that AWS regions matter. ACM certificates used by CloudFront must be created in
us-east-1 , while the application runtime and SES configuration for this project are in Sydney. Understanding these regional differences was an important part of getting the architecture working.📚 What I Learned
This project gave me practical experience with several AWS services, but more importantly, it taught me how those services work together as one system.
I learned that a simple product idea can become a distributed cloud application very quickly. Even a small time capsule application involves frontend delivery, API management, serverless compute, database storage, AI inference, scheduling, email delivery, authentication, HTTPS, and DNS.
I also learned that AI integration is not just about calling a model. The quality of the result depends heavily on the prompt, the constraints, and the way the model output is incorporated into the rest of the application.
Finally, I learned the importance of preserving user-generated data before applying AI transformations. By keeping both the original and AI-generated messages, the system can improve the writing experience without losing the user's actual words.
🚀 What's Next
There are several features I would like to add next.
The first is an AI preview step. Instead of automatically saving the AI-generated message, users could first see the transformed version and choose whether to use it, edit it, or keep the original.
I would also like to add different AI styles such as:
- Warm and supportive
- Friendly
- Calm and reflective
- Inspirational
- Graduation congratulations
- Future-self letter
Other future improvements include secure attachment uploads, user accounts, capsule history, stronger scheduling with event-driven services, and better monitoring for email delivery failures.
🌐 Try the Project
Final Thoughts
Dear Tomorrow started as a small idea about sending a message to the future, but building it became an opportunity to explore how AI and AWS can be combined to create a more personal experience.
The most meaningful part of the project for me is that the technology is designed around a very human idea: sometimes the person who needs encouragement the most is the person we are going to become.
At the same time, this challenge was especially valuable because it gave me an opportunity to move beyond simply developing an application and actually deploy a working service within the AWS ecosystem. Through the project, I was able to gain hands-on experience connecting services such as Lambda, DynamoDB, S3, CloudFront, API Gateway, Amazon Bedrock, SES, and ACM into a functioning application. It was getting to experience how an idea can move from code on a local machine to an actual cloud-based service using AWS.
#agents #AWS #AmazonBedrock #Serverless #Lambda #DynamoDB #AmazonSES #CloudFront
Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article