AWS Builder Center
Weekend Showcase Challenge: The Idea Forge — Daily Autonomous Productivity Agent

Weekend Showcase Challenge: The Idea Forge — Daily Autonomous Productivity Agent

An autonomous AI agent built on AWS that generates unique productivity and life ideas every day. Powered by Lambda, DynamoDB, Bedrock (Claude), EventBridge, and API Gateway — fully serverless, fully autonomous, running on AWS Free Tier.

Weekend Showcase Challenge: The Idea Forge — Daily Autonomous Productivity Agent

Tags: #application



Vision & What It Does

Most productivity advice is generic, repetitive, and easy to ignore. What if you had a personal muse — an AI that wakes up every morning, reviews everything it has already suggested, and crafts one fresh, actionable idea just for you? That is The Idea Forge.
The Idea Forge is a fully autonomous agent that generates a unique productivity or life idea every single day. At 9 AM UTC, a Lambda function fires, reads every past idea stored in DynamoDB, and calls Amazon Bedrock (Claude 3 Haiku) with a carefully constructed prompt that ensures novelty. The new idea is stored with metadata — category, timestamp, and a unique ID — and added to a growing public gallery served through API Gateway.
The result is a living collection of ideas that gets richer every day, covering ten rotating categories: morning routines, productivity hacks, wellness practices, learning tips, habit building, focus techniques, creative exercises, relationship strengthening, financial micro-habits, and gratitude practices. No two ideas are ever the same.
The problem it solves: People want fresh, specific, actionable inspiration — not the same "wake up at 5 AM" advice recycled endlessly. The Idea Forge delivers exactly that, autonomously, every day, without any human intervention.



How I Built It

I started by scaffolding the project with AWS SAM CLI ( sam init ), choosing Python 3.11 as the runtime. The architecture is intentionally simple — one Lambda function handles both idea generation and the gallery API, while DynamoDB provides persistent memory.

The Core Loop

  1. Read — The Lambda function scans the DynamoDB table to retrieve all past ideas
  1. Reason — It checks which categories were used recently and picks an underrepresented one
  1. Generate — It calls Amazon Bedrock with Claude 3 Haiku, passing the last 5 ideas as context to prevent repetition
  1. Store — The new idea is written to DynamoDB with a UUID, timestamp, and category tag
  1. Serve — A second handler renders a responsive HTML gallery or returns JSON via API Gateway

Key Decisions

  • Claude 3 Haiku over larger models: Haiku is fast, cheap, and creative enough for 1-2 sentence ideas. It keeps costs near zero on Free Tier.
  • Category rotation: Instead of random selection, I track recently used categories and pick from the least-used ones. This ensures variety across the ten domains.
  • Single Lambda, dual handler: Rather than two separate functions, I used one Lambda with two handlers —  lambda_handler  for generation and  gallery_handler  for the API. This simplifies deployment and reduces cold starts.

Challenges

  • Novelty over time: As the collection grows, ensuring freshness becomes harder. I solved this by passing recent ideas as negative examples in the prompt — "don't repeat these."
  • SAM CLI + Python version mismatch: My local machine had Python 3.14, but SAM CLI only supports up to 3.12. I worked around this by deploying Lambda functions directly via the AWS Console and setting up API Gateway manually.



AWS Services & Architecture

Services Used

Service
Role
AWS Lambda
Serverless compute — runs the idea generator and gallery API
Amazon DynamoDB
NoSQL database — stores all ideas with metadata
Amazon Bedrock
AI inference — Claude 3 Haiku generates unique ideas
Amazon EventBridge
Scheduler — triggers Lambda daily at 9 AM UTC
Amazon API Gateway
REST API — serves the gallery HTML and JSON endpoint
AWS SAM CLI
Infrastructure-as-code — project scaffolding and deployment

Architecture Overview

EventBridge Schedule (Daily 9 AM UTC)
│
▼
Lambda Function (idea-forge-generator)
│
├── Read DynamoDB (past ideas)
├── Call Bedrock (Claude 3 Haiku)
├── Generate unique idea
└── Store to DynamoDB
API Gateway (REST API)
│
├── GET / → Lambda → HTML Gallery
└── GET /ideas → Lambda → JSON API
All services run within AWS Free Tier limits. Lambda processes approximately 30 invocations per month, DynamoDB stores less than 1 MB, and Bedrock uses roughly 3,000 tokens per month — all well within free allowances.



What I Learned Across the Summer

This summer's build series pushed me to think beyond one-off projects and into autonomous systems — agents that run without human intervention and produce value over time.
Key takeaways:
  • SAM CLI makes serverless deployment frictionless — one  sam deploy --guided  command creates an entire stack
  • EventBridge cron expressions are simple yet powerful for scheduling autonomous tasks
  • DynamoDB's on-demand pricing is perfect for unpredictable, low-volume workloads
  • Bedrock's Haiku model strikes the ideal balance between quality and cost for high-frequency generation tasks
  • Building autonomous systems forces you to think about state management, novelty, and graceful failure — skills that transfer to any production system
  • The AWS Free Tier is remarkably generous — this entire project runs at zero cost
The biggest lesson: an autonomous agent is only as good as its memory. Without reading past outputs, The Idea Forge would repeat itself within days. Adding memory (DynamoDB scan + prompt engineering) transformed it from a random generator into a curated, ever-growing collection.



Links




Builder Shoutout

Huge thanks to Lewis Sawe (@LewisAwsome) — his  Museum That Grows  project inspired me to think about what autonomous systems could create over time. The idea of an agent that reads its own history and builds something new each day changed how I approach AI projects.


Built for the AWS Summer Builds Showcase 2026 | Weekend Challenge: August 28–31
Any opinions in this article are those of the individual author and may not reflect the opinions of AWS.
Enjoyed reading this content? Let the author know!

Your likes, comments, shares, and saves help creators reach more builders.

Loading recommendations

Loading article