Why can't I sleep

Why can't I sleep architecture: buyer loop and human-approved Strands campaign agent

Why can't I sleep is an Everyday Agent for people who wake tired, wake at 3am, or see low deep-sleep on their watch — and get nowhere with generic tips.

Two loops: 1) Buyer loop (live product) at https://why.longevitygreenlight.com/ — pay $0.88 (R16), play a 58-second browser game (12 green/red bedtime scenes), leave with one named habit stealing tonight. No app, no subscription, not medical advice. Payfast + Lambda ITN credits the find; SES emails access after payment. 2) Campaign agent loop (Strands Agents SDK) on Amazon Bedrock (Nova Lite), optionally Amazon Bedrock AgentCore: probe sleep-symptom questions → score habit clarity → draft a habit-fix page → await human approval → re-check. Never auto-publishes.

Judges can run the agent offline with fixtures (./demo.sh) or against live Bedrock / AgentCore (see AGENTCORE.md in the repo).

Who it's for: people searching for sleep answers at night, and the LongevityGreenLight builder who needs campaign drafts without shipping unapproved copy.

Why it matters: chatbots lecture and trackers score — neither names tonight's thief. This agent does the repetitive probe/score/draft work in the background and only surfaces when a human must approve.

How it works: Strands Agents tools (probe, score, draft, approve, recheck), Amazon Bedrock amazon.nova-lite-v1:0, optional AgentCore runtime_app.py, Lambda Function URLs, S3+CloudFront static site, DynamoDB find credits, SES access email.