Hack My Agent OR "Alex Under Attack" - Live Meetup
📢 Update 15.02:
Doors open at 17:30 — join us for pizza and mingling!
Talks begin at 18:00 sharp
—-
Want to try and hack Alex yourself? Join here 👇
Hack Alex — How a Team of Strangers Broke an "Unbreakable" AI Bot
Alex Livrent built a self-healing AI twin: Claude 4.5 + Ollama, gated through OpenClaw, layered with SOUL.md, MEMORIES.md, and a heartbeat pulse. Then he fed it to a group chat and dared the world to break it — with his real calendar and contact list on the line.
It auto-hardened with every failed prompt. It summarized attacks. It seemed immune.
But a bunch of strangers, never having met, quietly coordinated in a side group — sharing screenshots, testing rhythms, syncing their nudges. When something leaked, they celebrated like they’d won a cosmic joke.
This meetup is the fallout: • 🧠 Alex on OpenClaw routing, dual-model defenses, and where it blinked. • 🕸️ The strangers on how group timing and indirect choreography cracked what brute force couldn’t.
• Talchaim Damri - how AWS AgentCore could help secured Alexbot
• 🧪 Try your hand in a live prompt injection sandbox.
📌 Agenda still in progress — expect updates and changes.
For anyone building agents that talk too much.