Before the model enters
I’m building SealCheck to ask a question that comes before testing an AI model: how secure is the sandbox it will run in?
The tool focuses on assessing sandbox security in preflight mode, before model testing begins. It brings together the two things I keep coming back to: building with AI and understanding the boundaries around it.
Work in progress
SealCheck is in development. This page describes the direction of the project; I’ll share the specific checks, implementation, and findings as the work takes shape.