Build to understand
Shipping real things is how I learn where the theory breaks. That is true for product, and it is especially true for AI security.
About
I am most useful where product instinct meets systems risk.
jag.sh is a builder's portfolio, but it is also a directional signal. The through-line is simple: build real things, understand where they break, and get better at making that complexity legible.
Translate across layers
I care about the seam between executive intent, product choices, and technical controls. Most risk hides in that translation gap.
Prefer legible systems
The most trustworthy tools are the ones people can reason about quickly, even when the underlying system is complex.
Working Thesis
The most interesting AI security work is not isolated in a red-team deck. It happens earlier, where product, policy, trust, and rollout choices are still negotiable. That is the lane I want to keep moving toward.