VaHive Systems Lab is an independent research operation founded by Titiya Ruangkwam and Calvin Cook, based in Thailand. We have no institutional affiliation, no external funding, and no academic home — just two people who became convinced that the deployment-time alignment problem was both critically important and seriously underserved by existing work.

MAGUS is what three years of first-principles design produced. It is a runtime governance architecture built around a specific insight: that AI systems can be perfectly aligned at training and still drift structurally over time in deployment — through instruction drift, autonomy accumulation, and authority laundering — in ways that no current governance framework is designed to detect or prevent.

The v3.0 specification covers two complete deployment pathways: one for locally-hosted LLM inference environments, one for API-deployed agent systems built on cloud-hosted models. Both pathways are sealed, formally specified, and documented across 14 documents with a combined 46-entry elevation cycle tracker — every entry resolved. The Local LLM pathway is published on Zenodo (DOI: https://doi.org/10.5281/zenodo.19013833). v3.5 is in active development.

We're applying to Manifund because the gap between our specification and a working reference implementation is hardware. That's the whole problem. The architecture is done — we need to build it.