The Event Loop at Doorloop

​18:00 - Mingling 🎉, SWAG 🎁, pizza 🍕, and beers 🍻

​18:30 - lectures

​

​AI Agentic Development Feedback (Ralph) Loop

​by Neriya (Neri) Rosner, Senior Full-Stack Engineer, DoorLoop

​What if your AI didn’t just write code, but reviewed, tested, deployed, and fixed it on its own?

​Most AI workflows stop at generation. A developer writes a prompt, reviews the output, finds bugs or architectural issues, rewrites the prompt, and repeats, generating wasted iterations and unused code along the way.

​We built a different loop: AI that checks itself end-to-end.

​In this talk, we’ll show how our agent starts with fast static feedback using ESLint and TypeScript, validating its code instantly. It runs unit, integration, API, and third-party contract tests, executes real user flows via Playwright MCP, inspects logs during development with Sentry SpotlightJS, commits and pushes to GitHub, and deploys through Vercel, and solves CI issues.

​Finally, we’ll show how, using Ralph Loop — built on a Research → Plan → Execute → Review framework — the agent can iterate overnight and continuously develop.

​You’ll leave with a practical blueprint for building fully agentic, self-correcting Node.js workflows.

​

​Exactly once processing is a lie : Building Bulletproof Idempotency in Node.js by Achiya Haviv | Full Stack Developer at Bilton

​"Exactly once" delivery is a great marketing promise, but in real distributed systems, it's an engineering myth. From dropped networks to a personal double-charged bank account, message duplicates are inevitable. In this session, we’ll explore the underlying failure modes of queues like SQS and Kafka, and walk through practical Node.js patterns to build bulletproof, idempotent consumers. Learn how to stop trusting your infrastructure blindly and handle duplicates by design.

​

​Going ACID-less: Surviving Distributed Data Chaos

​by Harel Dagan | Software Engineer

​Moving to microservices and the Saga pattern means losing a critical luxury: Isolation. Without global transactions, your database is exposed to asynchronous network chaos, leading to dirty reads, race conditions, and dangling resources.

​In this session, we’ll break down three common data failures in distributed systems: Dirty Reads, Lost Updates, and Out-of-Order Compensations. We will evaluate the engineering trade-offs of solutions like Optimistic Concurrency Control and Idempotency Tombstones.

​You'll walk away understanding why our job isn't to build a perfectly sterile system, but to design one that recovers safely when things get dirty.

​

​Kindly be informed that attendance at the meetup is on a first-come, first-served basis until capacity is reached. Although a place has been reserved for you (after click on attend button), please note that it does not guarantee your spot.