Two decades across applications, infrastructure, cloud architecture, and security. Currently researching model harness design and the failure modes that appear when models are given authority over real infrastructure.

I work in the open at alignment.farm and in alignment-farm and stack-research.

• construct-2: how agents accumulate useful experience across sessions, and where it should live: memory, lessons, tools, weights, or runtime rules
• construct: a lab for governed agent memory, judged against the external world rather than retrieval or self-report
• substrate: local-first shared environments where users, agents, and tools share one conversation on append-only Markdown
• materialize: isolated, auditable workspaces built from read manifests, exposing exactly the declared surface to an agent

• The Unaskable Question: mapping the negative space of language models

Independent software engineer and former CTO. Experience in building systems, owning production infrastructure, security, architecture, technical leadership, and self-directed experimental work.

Writing in: Python, TypeScript, Bash, Ruby

Infrastructure: AWS (CDK v2, CloudFormation, Terraform), Docker, Kubernetes

Certified: AWS Generative AI Developer (Professional)