Research in 2025
At Econic, we’re building the infrastructure for collaborative intelligence—where individuals and organisations can unlock new levels of cooperation and shared growth with their data as we transition to an age of AI and AGI. Our research for 2025 focuses on three interconnected areas that we believe are fundamental to creating collaborative intelligence systems.
Neural Graphs (GNNs).
Transforming how we organise and explore knowledge through adaptive neural architectures that learn through relationships.
Neural graphs offer a fundamentally different approach to capturing knowledge and enhancing AI capabilities—one that promises to be more agile, more powerful, and more cost-effective than current large-model approaches.
By capturing information and developing models as interconnected relationships rather than static embeddings or single large neural networks, we can unlock new AI capabilities that leverage the inherent relationships within information sets and each other. These capabilities include: intelligent search that understands intent beyond keywords, AI personalisation that adapts to individual patterns, advanced reasoning that navigates complex multi-discipline problems, and dynamic information organisation that surfaces hidden insights across fragmented information sets and domains in real-time.
Focus 1: Intelligent Search.
AI models that dynamically capture the unique statistical patterns of data structures, enabling search that understands context beyond what simple keyword and even natural language semantic queries are capable of.
Traditional search systems are limited. They treat information as isolated fragments, force queries into rigid templates, and return static lists. Type a query, get results. Sort by date, filter by category. These approaches worked when information was scarce and well-structured. Today, we’re drowning in data.
The core problem runs deeper than interface design. Current search paradigms treat each query as an isolated event, computing statistical similarities between words without understanding the conceptual relationships that give them meaning. They cannot adapt to evolving contexts or the hidden connections within your data.
Our research explores how neural graphs can redefine search through living models of information. Rather than matching keywords or computing word similarities, we’re developing systems that extrapolate the underlying statistical models revealing how concepts truly relate. Through adaptive self-reflection and continuous data mining, these models refine its comprehension of the information, creating search experiences that are more intuitive and continually evolving compared to traditional methods.
Focus 2: Twin AI Systems.
What one puts down, the other picks up—decoupling ‘real-time inference’ from ‘System 2 thinking structure building’ data flows. This enables a reinforcement learning system where fast transactional queries and deep relationship mining work symbiotically—each agent enriching the other.
Inspired by how human cognition separates immediate responses from deep contemplation, our dual-agent approach aims to balance the requirment for speed with deeper forethought towards greater understanding. The inference agent operates in real time, handling queries with learned efficiency, while the structure agent continuously mines for deeper patterns and relationships, through heirarchical list relationships.
Each agent teaches the other. Every rapid inference provides signals about what matters most, guiding deeper exploration wiht data mining. Every discovered relationship enhances future quick responses. This creates a set of cycles where performance and intelligence amplify each other, achieving what monolithic systems cannot—instant responses that get smarter over time without sacrificing speed for depth or depth for speed.
Focus 3: Dynamic Embeddings For Optimal Graph Traversal.
Embedding atomic operations (sorting, filtering, Boolean logic) and an understanding of Bayesian statistical principles into neural graphs to facilitate advanced query and traversal strategies on large complex knowledge bases.
Traditional embeddings try to capture everything they possibly can through massive, static representations. Our research takes the opposite approach. We’re developing graph neural networks designed for small and dynamically constructed embeddings specifically tuned to the statistical shape of the knowledge base, the underlying data, the users, and the deeper context at hand.
The aim is a lightweight matrix of weights that acts as a customised lens through which to view and interact with information, a key to the information treasure map of sorts—coarser in coverage but far more deeply aligned with both the user’s intent and the data’s inherent structure.
With this approach we can facilitate complex traversal and query strategies that can navigate vast datasets efficiently and accurately, and personalised outcomes at a granularity and speed that massive models cannot achieve.
Collaborative AI and AGI.
Building Infrastructure for Collaborative Intelligence and AGI
The promise of AGI isn’t a single superintelligence—it’s collective intelligence emerging from collaboration.
AI advancement depends on access to diverse data and perspectives, yet the systems we use force organisations into isolation. The result: duplicated efforts, fragmented insights, and development bottlenecked by artificial boundaries.
What if competitive advantage came not from hoarding intelligence but from participating in networks that multiply it? This research is about creating the substrate on which collaborative AGI can emerge, when organisations can contribute to and benefit from collectively-gained intelligence whilst maintaining sovereignty over their data. In this way we can unlock possibilities that existing isolated systems could never achieve.
Focus 1. Massive Graph Database.
A real-time document database designed from the ground up for collaboration and statistical inference. Synchronise massive amounts of changing information (graph structures) across zero-trust boundaries and thousands of diverse actors and roles while maintaining complete data sovereignty and control.
Traditional databases were designed for a simpler time—when data lived in one place and relationships were explicitly mapped. They weren’t built for today’s reality of global collaboration, dynamic businesses, and millisecond updates. Current real-time databases can achieve this partially but assume single ownership, while massive multiplayer online gaming engines have proven it’s possible to synchronise complex state across thousands of participants in real-time, these systems are built for virtual worlds, not for the real-world data collaboration needs of businesses today.
Our research develops a database built from first principles for collaborative intelligence. By implementing a zero-copy delta-propagation architecture we can achieve million-operation-per-second throughput across a global network of thousands of users. Cryptographic proofs replace trust, allowing competitors to share insights without exposure, and by embedding statistical inference capabilities directly into the data layer, we facilitate a distributed computing or intelligence layer between participants.
Focus 2. Collaborative Intelligence Clusters.
The future of intelligence is collaborative. As we develop an intelligence economy, the advantage will belong to those who embrace deeper knowledge sharing and collaboration across traditional competitive boundaries.
Digital collaboration has evolved dramatically over the last 50 years, from the birth of the internet to open source to cloud computing. Yet despite these technological advances, most data remains siloed, trapped in industrial-age thinking.
As intelligence systems become central to each industry, data becomes the critical resource—but only when it flows. For businesses, this means a paradigm shift from “share only if you must” to “share unless there is reason not to.” Those clinging to data hoarding and siloed thinking will be systematically outpaced. Forward-thinking organisations are already reaping the rewards. When supply chains share disruption data publically, all adapt faster. When competitors pool insights, they discover opportunities invisible to any single player. It’s a transformation from zero-sum thinking—where your gain is my loss—to positive-sum outcomes where collective value creation benefits all.
Our research will work with strategic partners for key industries to study how deepening collaboration can create better outcomes. Using AI to facilitate, we track collaboration—measuring how information flows and transforms, observing what knowledge-sharing patterns emerge and how they add value. Through these partnerships, we aim to demonstrate how enhanced collaboration naturally evolves into industry-specific intelligence clusters, creating the foundation for sector-specific AGI systems.
Focus 3. Intelligence markets.
Establishing a foundation for intelligence markets where analysis, insights and optimisation pathways are the primary medium of exchange between collaborating organisations.
Every economic flow, be it capital market, food, materials or information, faces the same fundamental challenge: optimal resource utilisation. Trillions in value evaporate annually through inefficiencies and externalities.
As our intelligence economy takes shape, we predict the emergence of new type of market: an intelligent market. In these markets, intelligence is the commodity that is traded - insights in productivity gains, or waste reduction or more effective resource sharing. The age old “why would I help my competitors” is replaced with a non-zero-sum game.
Our research develops the foundations for an intelligent market. Working with industry partners, we’re creating frameworks for both mining and sharing actionable intelligence. We measure success through time-to-insight improvements and productivity gains. By treating analytical insights as tradeable assets rather than proprietary secrets, we aim to demonstrate how intelligence markets can transform today’s zero-sum optimisation into positive-sum value creation.
Evolving the UX
Bridging Human-Human-AI Interaction
The future of intelligence isn’t just about processing power—it’s about collaboration bandwidth. Today, humans and AI work in isolation, separated by interfaces that often constrain to specific ways of doing things. When a team needs to coordinate insights, they’re forced into separate dashboards, unable to co-create in real-time. When a research team employs hundreds or thousands of specialised AI agents, they juggle dozens of windows, losing insights in the gaps between apps. The bottleneck isn’t computational—it’s our approach to the interface.
The interface must become a collaborative medium, not merely a viewing window. Our vision: interfaces that adapt and morph at the speed of thought, expanding human-AI bandwidth through intuitive exploration and discovery.
Focus 1: Collaborative Intelligence Interfaces
Creating interfaces that can facilitate deeper human and AI interactions.
When experts collaborate on complex problems, isolation limits progress. Current interfaces show each user their own view, blind to the discoveries happening inches away. But deeper understanding comes from seeing patterns through different lenses—the engineer’s structural view, the designer’s flow perspective, the AI’s statistical patterns. We need interfaces that strengthen cognitive diversity as they deepen knowledge-sharing bandwidth.
Our research explores how a collaborative UX and UI might be designed to be both agile and universally applicable to a wide range of knowledge-coordination activities. The interface becomes a shared nervous system where every participant, human or artificial, contributes to the creative collective process in real-time.
Focus 2: Adaptive Spatial Environments
Developing 3D environments where the relationships embedded in information are spacial—creating an actual space for human-AI collaboration at massive scale.
Flat screens force us to compress multidimensional relationships into 2D planes. But collaboration is inherently relational, which means it is spatial—we need to see not just numbers and text, but connections, proximity, density, movement.
Our research develops a spatial computing framework for environments where humans and AI can intuitively navigate massive amouts of information together. The goal: transform overwhelming complex and evolving data (state) on a massive scale into a collaborative 3d space.
Join Our Research
If you are interested in our research, we’d love to hear from you.
Researchers: We’re exploring the intersection of distributed systems, neural architectures, and human-computer interaction.
Developers: Help us build the open-source infrastructure that will power the next generation of collaborative AI.
Investors: Support the development of foundational infrastructure for the intelligence economy.
Contact us at hello@econic.ai or explore our work at econic.ai