About Self Research IP & Patent Contact

Get in touch

hello@thegeometrical.com

Research & Publications

Research.

Papers, protocols, and technical contributions from Geometrical. Open science where possible. Protected where necessary.

2

Papers published

P202631047

Patent filed · OEPM

1

Open-source protocol

2026
Jul 2026

Paper · Zenodo

Personal Episodic Identity Substrate (PEIS): A 9-Dimensional Architecture for Human Memory in the Agent Economy

Introduces the PEIS framework — a 9-dimensional model for representing human episodic memory as machine-native substrate. Covers architecture, role inference, and the Minimum Viable Context principle.

zenodo.org/records/21923601
Jul 2026

Paper · Zenodo

The AIX Protocol: A Machine-Native Format for Contextual Identity in Multi-Agent Systems

Specifies the .aix format — a portable, spatiotemporally traceable identity envelope for agent-to-agent context sharing. Covers Merkle-SHA256 integrity, federation, and the Nexus permission layer.

zenodo.org/records/21923809
Jul 19, 2026

Patent · OEPM · Filed

P202631047 — Privacy-Native Memory Encoding for LLM Inference Pipelines

Two original inventions: Session Privacy Protocol (intra-session coherence with inter-session opacity) and PII Tokenization Without Fine-Tuning. Filed by Alan Daniel Lapid. PCT/EPO deadline July 2027.

View IP page
Jul 2026

Open Source · GitHub · Apache 2.0

AIX Protocol v0.2.0 — Reference Implementation

Open-source reference implementation of the AIX Protocol. Includes .aix format specification, Nexus v0.1 federation layer, 23 tests green on Python 3.10/3.11/3.12.

github.com/alanlapid/aix-protocol
Forthcoming
2027

Paper · cs.CR + cs.AI · Forthcoming

IP Technical Detail Paper — Privacy Architecture of P202631047

Full technical disclosure of IP1 and IP2. Publication deferred until PCT international filing is established.

After Jul 2027
2026–27

Paper · cs.AI · Forthcoming

Self Empirical Results — Multi-User PoC Analysis

Recall@5, role inference accuracy, and causal graph density across a multi-user pilot cohort. Extends single-user PoC results (38% direct, 12% indirect recall@5).

In preparation

Collaboration, press inquiries, and research partnerships.

hello@thegeometrical.com