Weekend Creative Challenge: Word Loom
Word Loom: weave your vocabulary into a story. Write down the English words you're learning. Word Loom turns them into a micro-story or poem, visibly weaving them into the text using Amazon Bedrock (Nova).
Vision & what the app does
Word Loom started from a simple question: what if, instead of memorizing lists of English vocabulary, those words turned into something you actually wanted to read?
The app takes the English words I'm currently learning (many of them while preparing for the AWS Certified Solutions Architect Associate certification) and "weaves" them into a short AI-generated micro-story or poem. Every word the model manages to include gets visually highlighted in the text, like a thread woven into the story, so at a glance I can see how many of my target words made it in and how they're used in natural context.
The idea is intentionally simple: instead of an isolated flashcard ("resilient = resiliente"), the word lives inside a meaningful sentence, with tone and emotional context. That makes it easier to remember, and along the way turns a repetitive study task into a small creative moment.
How to use it
- In the text box, type the English words you want to practice, separated by commas.
- Choose one of the two buttons to generate a Historia (Story) or a Poema (Poem). The option labels are in Spanish, my native language.
- Click "Tejer" (Weave).
- The generated text will appear below, with the words that weave through it highlighted.

How I built it
I started by defining the simplest possible experience: a text box to type comma-separated words, a toggle between "Story" and "Poem," and a generate button with no complex configuration.
The frontend is a single HTML/CSS/JavaScript page with no frameworks, to keep deployment simple. I designed the color palette and typography to make the app feel warm and editorial rather than like a generic technical tool.
For the backend, I wrote an AWS Lambda function in Python that builds a specific prompt for the AI model: it asks it to write a short text (an 80-120 word story or a 6-10 line poem) that includes, unmodified, all the words the user entered. The function uses Amazon Bedrock's Converse API to invoke the model and returns the generated text to the frontend, which handles visually highlighting each woven word.
The biggest technical challenge wasn't the code itself, but the deployment: it was my first time using AWS SAM CLI from start to finish, from installing the tools with Homebrew to configuring credentials, and I ran into a syntax error in the infrastructure template (the Cors property was in the wrong place in template.yaml) that I had to diagnose and fix before the deployment would pass. It was a good reminder that in infrastructure as code, one misplaced property can block the entire deployment.
AWS services used / architecture overview
Services: Amazon Bedrock (Nova Micro model), AWS Lambda, Amazon API Gateway, Amazon S3 (static hosting).
Architecture flow:
Browser (index.html hosted on S3) │ POST /generate { words, mode } ▼Amazon API Gateway (public endpoint, CORS enabled) │ ▼AWS Lambda (Python 3.12) │ builds the prompt and calls bedrock-runtime.converse() ▼Amazon Bedrock — Nova Micro model │ returns the generated text ▼Lambda → API Gateway → Browser (highlights the words used and shows the word count)
The whole project was defined as infrastructure as code with AWS SAM (template.yaml), which means the entire backend can be recreated with two commands: sam build and sam deploy.
What I learned
- How to write more precise prompts for generative models: asking for "all of these words, unmodified" produces very different results than a vaguer prompt.
- The real difference between configuring Cors inside a Lambda function's event versus in the Globals section of a SAM template (a small detail that can bring down an entire deployment).
- How to activate and test an Amazon Bedrock model from the playground before wiring it into code, which saves debugging time.
- That the full flow of a serverless project (Lambda + API Gateway + static hosting on S3) is far more approachable than it looks once you break it into small, verifiable steps.
- And, as a nice side effect, I gained a genuinely useful new tool for my own English study on the way to the AWS SAA certification.
Link to the app
If you are learning English, this app is for you :)
Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article