Members-Only
Recent Talks & Demos are for members only
You must be an AI Tinkerers active member to view these talks and demos.
January 28, 2026
·
Waterloo
Blueberry Muffin: Semantic Injection
This talk demonstrates the "Blueberry Muffin Exploit," showing how injecting text into agent memory poisons reality, weaponizes trust, and triggers logic denial of service attacks.
Overview
𝗧𝗵𝗲 𝗕𝗹𝘂𝗲𝗯𝗲𝗿𝗿𝘆 𝗠𝘂𝗳𝗳𝗶𝗻 𝗘𝘅𝗽𝗹𝗼𝗶𝘁 is a piece of security research proving that the Agentic AI systems we are trusting to write our code are fundamentally insecure.
Tech stack