AI Strategy, Infrastructure Architecture & Security Research
This glossary introduces terms used in the linked research and project documents. The original publications and project specification remain authoritative.
IICP is an open, Apache-2.0 protocol for intent-based discovery and provider selection. Execution uses a negotiated binding; component maturity is documented separately. The project specification is authoritative. Publication and standards discussions do not establish IETF approval or broad adoption.
SYNAPSE appears alongside IICP in older career and project descriptions. Use the current IICP specification for the protocol name, scope and licensing; this glossary does not establish a separate current protocol or employer product.
A falsifiable systems-level theory explaining how load-induced numerical nondeterminism in LLM inference may propagate autoregressively, and a monitoring index to study stability under real-world deployment conditions. Reframes output drift as an observable stack-level property rather than a model-weight problem.
Read the canonical paper: Inference Stack Degradation Theory v1.5 →
R.E.A.D. × TV is an author-published research proposal for ethical-alignment measures. Its version 1.4 paper describes the proposed method; it does not establish independently validated ethical measurement or regulatory certification.
The linked service describes a writing workflow. Authorship and cross-links do not establish search rankings or independent validation.
See the agentic long-form offer