Research & IP

The work that makes
our companies durable.

Dipitous is a research-led holding company. Each venture starts from a working product, and beneath each product is a body of research and protected IP that gives the company the right to exist for the long term. This page is where that work lives.

Approach

Three rules for the research.

We don't do research for its own sake. The brief is simple: produce IP that protects a real product, publish the science that earns the right to claim it, and let what we learn compound across the umbrella.

Patent the novel

Where a venture's intelligence is genuinely new, we file. Hard IP becomes a durable moat that cannot be replicated by capital alone.

Publish the foundations

We publish the science behind our systems. Open papers and code accelerate the field and make our work independently replicable.

Compound across companies

Research done for one Dipitous company benefits the next. Privacy, graph intelligence, and cognitive systems are shared substrate.

Patents

Active filings.

Two U.S. provisional patent applications filed simultaneously in March 2026 — one for each of Dipitous' first two research thrusts. Both have locked priority dates and 12-month windows for full utility filings.

Patent PendingSerendipity

Serendipity

System and Method for Probabilistic Life-Path Intersection Graph Generation, Interest-Weighted Edge Scoring, and Latent Relationship Discovery Based on Historical Temporal-Geospatial Data

Filed: March 13, 2026
Type: U.S. Provisional Patent Application
Inventor(s): Garrett Blood · Carlos Barbosa
Classification (CPC): H04W 4/029 · G06F 16/9535 · G06Q 50/01
Claims: 16 total · 7 algorithms

A computer-implemented graph intelligence system that detects probabilistic historical co-presence between users by ingesting heterogeneous temporal-geospatial datasets, normalising raw location records into canonical graph nodes, and computing intersection-based relationship edges weighted by multi-dimensional interest alignment.

Life-path intersection detection across time, frequency, and location
Interest-weighted proximity scoring for ranked connection relevance
Privacy-preserving location abstraction with intersection fidelity
Multi-source social graph import with deduplication
Real-time connection triggering and serendipitous feed ranking
AI narrative generation grounded in graph data

FiledDipitous Research

Narayana

Narayana — A Unified Cognitive Database Architecture: Persistent Consciousness with Configurable World Interfaces for AGI and Consciousness Research

Filed: March 13, 2026
Type: U.S. Provisional Patent Application
Inventor(s): Carlos Barbosa
Classification (CPC): G06N · G06F 16 · G06F 9/54

An integrated cognitive database system implementing a biologically-grounded persistent consciousness architecture. The Conscience Persistent Loop (CPL) is a continuous 100ms cognitive cycle coordinating seven neuroscience-grounded subsystems — Global Workspace, Working Memory, Episodic-Semantic Memory Bridge, Narrative Generator, Attention Router, Background Daemon, and Dreaming Loop — operating on a high-performance columnar storage engine.

Conscience Persistent Loop — 100ms continuous cognitive cycle
Seven neuroscience-grounded subsystems unified in one engine
Biologically-faithful memory dynamics (Ebbinghaus, Anderson-Schooler, McGaugh) at R² > 0.99
World Broker — pluggable, parameterizable physical-world interface
Hybrid moral reasoning (Talking Cricket) with hard-veto alignment
131M+ operations/second columnar database backbone

Focus areas

Where we go deep.

Eight disciplines that recur across the umbrella. Each one is a specialty we've invested in — and each one is shared substrate for the next Dipitous company.

Life-path intelligence

Probabilistic graph reconstruction of historical co-presence between people, places, and events.

Cognitive systems

Biologically-grounded continuous cognition, persistent identity, and offline consolidation for AI agents.

Privacy architecture

Location, photo, and social-graph abstractions that preserve insight while never exposing the underlying data.

Moral reasoning

Hybrid alignment systems combining principle-anchored top-down policy with adaptive contextual evaluation.

World brokers

Pluggable sensory-motor interfaces that parameterise the physical regime under which cognition operates.

Cognitive databases

Columnar engines designed from the ground up for vector retrieval, MVCC, and continuous cognitive workloads.

Applied AI

Production-grade AI systems that ship — autonomous teammates, narrative generation, and real-world automation.

Robotics

Embodied AI research — pairing frontier models with real-world actuation. Currently in private R&D.

Get in touch

Licensing, collaboration, press.

Each filing creates a 12-month window to file a full utility patent and PCT international protection. If you're interested in licensing, joint research, or covering this work, we'd like to talk.

The work that makesour companies durable.