I build harnesses and systems around AI models that shape how they reason, use context, stay within boundaries, evaluate their work, and produce evidence about what they did.
My work spans reasoning and behavior structures, agent evaluation, authorization, developer safeguards, and governed execution.
My public work includes a runnable change-evaluation demo, a local Git safeguard that requires approval for selected file paths, and engineering cases showing how I investigate failures, repair controls, and test the resulting systems.
I'm the founder of Second Mind Systems and open to engineering roles, consulting, implementation work, and technical collaboration.
Contact: [email protected]
Behavioral regression testing for coding agents.
Your agent passed once. What happens when the conditions change?
Blue Jackal tests selective adaptation: whether the agent responds appropriately to what changed without losing the behavior that should still hold.
Unit tests check the code. Blue Jackal checks the agent.
Repository · How it works · v0.1.0
My systems analysis of how purchaser protections can change as Mark Cuban's open-source PBM agreement moves through procurement led to a controlled-adaptation contribution.
Mark Cuban reposted our resulting contribution on LinkedIn.
Inspect the work and evidence →
The engineering portfolio is structured as a navigation system, designed to route you from the responsibility or capability you want to evaluate to the work and evidence that matter.
Choose what you want to evaluate →
I work hands-on across problem framing, architecture, implementation, evaluation design, acceptance criteria, and final technical review.
AI agents assist with scoped implementation and review. I define the work, set the boundaries, check the results against source evidence, and decide what is accepted.
Explore the engineering portfolio · About Second Mind Systems