Why ZFS is perfect for agent workloads
Agents need storage that forks, checkpoints and proves change. Why ZFS, one of the key components under Hive, is the least-wrong substrate for agent workloads — and where S3 fits.
Agents need storage that forks, checkpoints and proves change. Why ZFS, one of the key components under Hive, is the least-wrong substrate for agent workloads — and where S3 fits.
Agents, models, tools and data as nodes you can connect and rearrange. Why i built Portal and what it turned into.
A Verilator packet-router regression as a Kubernetes Indexed Job: deterministic randomized testbenches, per-completion seeds, and a clean pass/fail lifecycle without a shared shell box.
GNU Make as a compact templating engine and orchestration layer for Kubernetes deployments: parameter files, parse-time validation, manifest registries, and rollout targets that render, diff, apply, and capture evidence.
Jail kubelet and containerd with nsjail: restricted rootfs views, projected hostPath allowlists, host-owned networking, and SELinux as the real MAC layer.
Build immutable LinuxKit node images, boot them as Firecracker microVMs, and turn them into a Kubernetes cluster — first with k3s, then a full kubeadm HA control plane behind HAProxy.
An agentic multi-model stress test of a Strait of Hormuz closure: oil and LNG pass-through, country and sector exposure, hidden transmission links, and scenario horizons to twelve months.