Weekend Creative Agent Challenge: Word Loom
Word Loom: weave your vocabulary into a story, now with an agent that creates a new one every day, without being asked!!
Vision & what it does
Last weekend I built Word Loom for the "Build a Creative App" challenge: an app that took English vocabulary words and wove them into an AI-generated micro-story or poem. When I saw this second challenge's call (turning that creative app into an always-on agent) the decision was easy: give it continuity, not just to complete the challenge, but because the real goal behind Word Loom from the start has been to reinforce my own English learning while I prepare for the AWS Certified Solutions Architect – Associate certification.
The previous version depended on me typing the words in and clicking "Weave." It was useful, but it was still a tool that I had to activate and feed. This second version flips that role: now an autonomous agent runs once a day, without anyone asking it to, picks a new set of words from a vocabulary bank and a short theme on its own, generates a story or poem with Amazon Bedrock, and leaves the result ready and waiting for me when I open the page.
That makes the app genuinely more complete as a study tool: instead of always reviewing the same five words I chose, every day I run into new words I didn't pick myself, used in a different context and a different tone (the agent also alternates between story and poem and changes the theme rain, a midnight train, a letter found years later, depending on the day). It's the difference between studying with my own usual flashcards and having someone else surprise me with new vocabulary every morning.
How I built it
I reused the whole foundation of the first version (the frontend, the S3 bucket, the original on-demand Lambda) and added a second Lambda function dedicated exclusively to the agent. The most important design decision was where the autonomy actually lives: I considered simply generating new content every time someone reloads the page, but that would still be reactive at its core, the agent wouldn't be doing anything while no one was watching, it would just move the trigger from a click to a reload. For it to be a real agent, in the sense the challenge asks for ("ready when you return"), the generation had to happen on its own, on a schedule, without depending on a human opening anything.
That's why I used Amazon EventBridge Scheduler to trigger the agent function once a day. The function picks 5 random words from a pool of 30, alternates between story and poem depending on the day of the month, selects a short theme from a rotating list, builds the prompt for Bedrock, and saves the result as latest.json in the same S3 bucket where the site lives, plus an archived copy by date in a history/ folder, to keep a record of what the agent has created over time. The frontend simply reads that file on page load; it doesn't trigger any generation itself.
AWS services used & architecture overview
Services: Amazon EventBridge Scheduler, AWS Lambda (two functions: on-demand generation and the autonomous agent), Amazon Bedrock (Nova Micro model), Amazon API Gateway, Amazon S3 (static hosting and storage for the generated content).
Agent flow:

Everything was defined as infrastructure as code in the same AWS SAM template.yaml, now with two Lambda functions sharing the same stack.
What I learned
- How to use Amazon EventBridge Scheduler to trigger automatic Lambda executions, something I hadn't done before.
- A more conceptual than technical lesson: the difference between "looking automatic" and "being autonomous", consciously deciding that generation shouldn't depend on a page reload, but on a clock, so the agent actually matched the challenge's real intent.
- To think about more granular IAM permissions: the agent function needed S3 write access (s3.PutObject) in addition to the Bedrock permission the original function already had.
- And, once again, a direct benefit for my own studying: I now have a daily source of new vocabulary in context, without having to remember to generate it myself.
Link to the app
Are you lerning English? this app is for you!!! Enjoy it!
Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article