I write about data platforms, small infrastructure, software boundaries, and using agents in real work. Most entries begin with a system, decision, or failure and then ask what should survive the next change.
Projects show what exists and when I last checked it. Notes record the rules, mistakes, and design choices I expect to need again.
Start here
Four places to begin, across agent work, software assurance, operations, and data systems.
A solo operator's tooling keeps adopting controls that look stolen from an organization. The controls are not for headcount; they are for coordination between actors who share neither memory nor judgment, which describes one operator across time as much as it describes a team.
A reading map for building AI systems that can show what they were asked to do, what they did, why a result should be trusted, and how failures are contained. It starts with established assurance practice and ends with a dated watch list of open work.
Migrating a Talos cluster from a legacy subnet to its own VLAN while it kept serving. The design was mostly an ordering problem, plus one bootstrap loop: the network controller lives inside the network it manages.
A working family of sprint, queue, knowledge, audit, and cockpit tools now composes into one served layer, without giving up explicit state ownership or machine-local execution.
Homelab AnalyticsThe household data and decision platform that owns long-lived semantics, scenarios, policies, and approvals. keeps household reporting, planning, simulation, policy, trust, and agent-facing retrieval in one semantic model. Home Assistant is the device-facing partner, not the system of record for household reasoning.
AppserviceThe private GitOps repository that holds desired state, recovery rules, and operational evidence for the cluster. is the operations repository for a Talos-based Kubernetes cluster. Reconciliation, encrypted secrets, recovery gates, upgrades, and incident evidence live with the desired state instead of in operator memory.
A wrapper-derived effect-intent projection should let an application session correlate instability during the next Talos control-plane upgrade; if it still escalates without correlation, the hypothesis fails.
An EventStormingAlberto Brandolini's workshop method for modeling a domain on colored sticky notes -- commands, domain events, actors, views, and hotspots -- normally run with a room of stakeholders around a paper timeline. pass run solo, over agent session logs instead of a stakeholder workshop, still produced six ranked findings. The notation survived; the workshop did not; the value was in clustering the absences.
A prior "human perpendicular to the loop" claim only describes one interface. The fuller model is two coupled loops at different speeds, with a settlement boundary -- not an empirical/normative split -- deciding when the slow loop must be consulted.
The next-prompt heuristic measures what became visible at the human interface. A merged PR later filed as architecture, and a self-report that turned out to be reconstructable rather than introspective, both passed that test and still needed a different check further down.
RAG, vector search, graph engines, and generic evaluation tooling are moving into managed platforms. The durable work is defining which evidence may count, what failure means, and when an agentic system is acceptable to ship.