An engineering lab that publishes its decisions.

Background

Years in systems and platform engineering teach a specific instinct: the interesting problems live at the seams. Credentials, routing, evaluation, workflows — the parts nobody demos and everybody depends on. Applied AI is full of those seams right now, which is why the lab's work concentrates there: skill management for agents (Beskar), git-native ticketing for agent workflows (Rohrpost), and AI platform architecture in client engagements.

Philosophy

• Problems before techniques. A tool starts with a constraint that hurts in real work, not with a model looking for a use case.
• Boring on purpose. Plain files, git, single binaries, offline-first. Cleverness is a cost, not a feature.
• Decisions are content. Rejecting an alternative with reasons is more useful than shipping without one.
• Open where possible. Everything that can be open-source, is; everything covered by confidentiality is labeled as such.

Editorial split

The project sites — beskar.ventrion.dev and rohrpost.ventrion.dev — document what the tools do and how to use them. This site carries the other half: why they were built, which alternatives lost, what failed, and what it cost. Cross-linked, never duplicated.

Colophon

This site is static HTML with hand-written CSS, built with a zero-dependency Node script, and served from Cloudflare Pages. No frameworks, no trackers, no cookies. The design shares its language with the project sites — dark, compact, one restrained accent — with the parent site in steel rather than amber.