devlog~
AI Engineer in Dresden. I turn language models into reliable production systems — agent architectures, multi-model orchestration, and the tooling that holds it together.
Shipping Vega — an autonomous AI assistant that manages its own context. Running hydra reviews on every PR — up to 10 cross-model agents catching what I miss. Matching freelancers to projects with Project Beat.
Projects
- vegaTypeScript
Autonomous AI assistant that manages its own context, plans multi-step tasks, and executes with minimal supervision.
- hydraTypeScript
Multi-headed code review skill — up to 10 cross-model AI agents (Opus + Codex) with adversarial peer review and chairman synthesis.
- project-beatTypeScript
Freelance project discovery — AI-powered matching of consulting opportunities with relevance scoring.
- schliffPython
Deterministic quality scorer for AI agent instruction files — 8-dimension scoring, multi-format, anti-gaming.
- vault-syncPython
CLI that syncs GitHub repositories into an Obsidian vault as a searchable knowledge graph.
Writing
- The State of AGENTS.md: Scoring the Agent-Makers' Own InstructionsI scored the AGENTS.md files of the 16 biggest AI agent repos with a deterministic engine. ~1.46M combined stars, zero A grades, and the repo that popularized the convention ranks 13th of 16.
- Loop EngineeringWriting an agent loop is one line. Engineering what it carries between wakeups — idempotency, pacing, convergence, a cost ceiling — is the job.
- The Missing Layer: Static Security Analysis for CLAUDE.md and AGENTS.mdOWASP LLM Top 10 risks live in your instruction files. Runtime evals don't lint them. Here's what static security analysis catches — with a concrete finding.