https://luminx.one/ Research, notes, and software from LuMiNx — a multidisciplinary research institute studying AI and cybersecurity. en Tue, 05 May 2026 14:01:00 GMT https://luminx.one/notes/on-reading-the-model/ https://luminx.one/notes/on-reading-the-model/ Wed, 22 Apr 2026 00:00:00 GMT notes Notes on what it means to take a language model's output as evidence, and what a research institute owes the people on the other end of that output. https://luminx.one/research/orbitid-threat-model/ https://luminx.one/research/orbitid-threat-model/ Sun, 12 Apr 2026 00:00:00 GMT research We present the threat model behind OrbitID, an SSO provider that eliminates shared-secret authentication entirely. We enumerate the attacks that legacy providers leave open by retaining password fallback and email-link recovery, and show how a possession-and-presence model closes them while remaining compatible with OIDC relying parties. https://luminx.one/research/edge-native-homeservers/ https://luminx.one/research/edge-native-homeservers/ Mon, 02 Mar 2026 00:00:00 GMT research We describe MatrixFlare, a Matrix-spec homeserver implemented entirely on globally distributed edge runtimes. Room state lives in durable per-room actors; media in object storage; federation is asynchronous by default. We discuss the latency and cost characteristics of this architecture compared to a region-bound monolithic homeserver, and the protocol-level adjustments required to make federation tolerable under arbitrary geographic dispersion. https://luminx.one/notes/why-astro/ https://luminx.one/notes/why-astro/ Sun, 15 Feb 2026 00:00:00 GMT notes A short note on the engineering posture behind this site, and why a zero-runtime static generator carries the institute's voice better than a SaaS landing page would.