diff --git a/assets/sponsors/a/anyshift.png b/assets/sponsors/a/anyshift.png new file mode 100644 index 0000000000..3d1bd03ed2 Binary files /dev/null and b/assets/sponsors/a/anyshift.png differ diff --git a/content/events/2026-los-angeles/program/engin-diri.md b/content/events/2026-los-angeles/program/engin-diri.md new file mode 100644 index 0000000000..7477fc15d6 --- /dev/null +++ b/content/events/2026-los-angeles/program/engin-diri.md @@ -0,0 +1,24 @@ ++++ +Talk_date = "2026-03-06" +Talk_start_time = "13:45" +Talk_end_time = "14:00" +Title = "The Ralph Wiggum Loop: How Autonomous AI Loops Built My Serverless SaaS While I Slept" +Type = "talk" +Speakers = ["engin-diri"] ++++ + +You ask an AI assistant to build something, grab coffee, and return to find it stopped three steps in. "Should I continue?" it asks politely. You're not coding anymore. You're babysitting. Checking every five minutes saves nothing. + +Then I tried something different: the "Ralph Wiggum Loop." Named after the Simpsons character who cheerfully stumbles through chaos, it's a loop that continuously feeds a PROMPT.md file to Claude Code until infrastructure deploys successfully. Simple, maybe reckless, but it works. + +I pointed this loop at a real project: a production-ready URL shortener SaaS on AWS. DynamoDB for storage, Lambda functions for creation and redirection, API Gateway, S3 static hosting, CloudFront with HTTPS, plus a landing page and dashboard. + +Claude forgot IAM permissions on the first try. Put Lambda code in the wrong place. Took several attempts to configure CloudFront correctly. But it fixed each mistake and kept iterating. + +The final output: working infrastructure, passing tests, and a working frontend. + +