Lazy Where It’s Safe to Be Lazy
Explores which steps in Kubernetes GitOps management can be safely delegated to an AI agent. Distinguishes deterministic gates (cryptography, RBAC, allowlists) from probabilistic weights (prompts and rules).
Strategic insights on Applied AI architecture, high-velocity engineering cultures, and the evolution of technical leadership in the era of generative AI.
Software engineering is shifting from a craft of manual syntax to an architecture of intent and verification. As AI tools accelerate code generation, the developer's role is no longer writing loops, but establishing the constraints—types, schemas, and tests—that prove the result matches the goal.
I advocate for pragmatism over perfection—deploying capable, bounded systems today rather than waiting for hypothetical ultimate solutions. True engineering value comes from building reliable foundations, reducing developer friction, and solving concrete problems at scale.
"We don't need to solve AGI to change the world. We need good enough tools, deployed honestly, at scale."
Treating probabilistic models as standard software components—architecting systems with clear boundaries, safety guardrails, and observable behaviors.
Reducing systemic friction in the daily developer loop. Building environments where testing is fast, deployment is painless, and doing the right thing is the path of least resistance.
Using generative tools to prototype rapidly without accumulating unmanageable code debt. Balancing speed with the engineering rigor needed for production.
Strategic blueprints for the next era of technical management.
Explores which steps in Kubernetes GitOps management can be safely delegated to an AI agent. Distinguishes deterministic gates (cryptography, RBAC, allowlists) from probabilistic weights (prompts and rules).
An architectural case for ephemeral clients driving perpetual agent environments. Examines running headless daemons inside containers with read-only root filesystems and explicit host volume boundaries.
Deconstructs the hype surrounding the 'AI Engineer' title. Introduces the originate vs. relay test to distinguish builders of models from software engineers with AI dependencies.
A weekend spent reverse-engineering a hardware-to-vector tool using Claude Code. Explores the collapsing entry barrier for building complex software, and the threat this poses to businesses whose primary moat was implementation complexity.
An analysis of the viral PocketOS incident where an AI agent deleted a production database. Argues that blaming the AI vendor misses the real root causes: systemic failures in access control, backup validation, and human stewardship.
Staff Software Engineer
A technical leader working at the seam between human judgment and machine-generated code—where the cost of producing software has collapsed, but the cost of understanding it has not. I specialize in scaling distributed systems, reducing developer friction, and designing safe architectures for AI-assisted workflows.
"I believe technology changes the world through practical tools deployed at scale, not theoretical ideals. My goal is to build architectures that reduce friction while keeping teams in command of systems they can still understand."