Treating evaluation as a first-class cached graph artifact.
- Graph-native systems design
- Declarative infrastructure (Nix, Kubernetes)
- Evaluation, caching, and reproducibility
- Composable architectures beyond file-based repos
- FleetNix + CAPI integration
- Cluster bootstrapping + secret flows (SOPS, Clan)
- Exploring graph-based program representation
- Evaluation as a persistent artifact (not ephemeral)
- Language-agnostic program graphs
- Function-level composability + caching
- Replacing file-based repos with graph materializations
- Nix / NixOS / Home Manager
- Kubernetes (k3s, GitOps direction)
- Terraform / OpenTofu
- Linux + cloud infra
I care about:
- correctness
- reproducibility
- composability
- leverage