AgentGuard
Agents on a leash — policy, audit trails, and least-privilege tooling so the autonomous parts stay accountable.
Open dossier →A working notebook for verifiable systems, honest agents, and the quiet infrastructure beneath both. This is the public surface — production lives elsewhere.
Bias toward verifiable builds. Reproducible infra, observable systems, explicit trust boundaries — every layer should answer for itself.
Move fast, break things. Move fast with stable infra.
Optimize the output. Reason from first principles.
Automate the human away. Keep the human in the loop.
Trust the model. Trust, but verify the trace.
Hide the seams. Show your work.
Each tile is a working surface. Some are shipping; some are still in the sketchbook. Click through to read what's there.
Agents on a leash — policy, audit trails, and least-privilege tooling so the autonomous parts stay accountable.
Open dossier →Opinionated workflows around CRDs — schema, rollout, drift.
Open dossier →An audio experimentation stack — pipelines and taste in equal measure.
Open dossier →Vigilant signals for anomalies and misuse. Humane automation, with a low false-alarm rate.
Open dossier →Spatial and semantic maps drawn from quiet signals.
Open dossier →A workbench for marginalia — notes that travel with the code.
Open dossier →What I reach for, and where each thing earns its keep. No exhaustive list — just the tools I trust to behave.
More links and case studies are coming. This surface stays public-only by design — the rest sits behind your controls.