GraphNews

Explore ArcadeDB's 2025 year-in-review, highlighting advancements in vector search, Python integration, gRPC, and storage efficiency

“If you skip data modeling before building a knowledge graph, you’re not innovating – you’re gambling.”

Forget context graphs. That's so last month...

How structured knowledge became healthcare’s quiet advantage

Building AI systems that don’t just make decisions, but remember and use the “why”.

...the pieces you need to know to start building an ontology for a knowledge graph

Part 1 — Generative Knowledge Modeling

“Ontology” used to be a cursed word.

Ontologies - why in the world would you ever need one?

Context Graphs: A Tangled Web Of Confusion 🕸️ With Context Graphs, I'm having loads of great discussions (with a lot of awesome critique and feedback - love it, helps me learn) 🙏 But also I see a lot, trying to mis-classify the problem space by overlaying other graph based solutions onto this one. There seems to be confusion about the space we are operating in. So, with that in mind, I thought I'd go back to basics. State Machines - concepts: ⚙️ - States are (where you are and where you can possibly move to) - Transitions (how you move) - Events (what triggers movement) - Guards (conditions that allow or block) The system is always in one state. They have a possibility space i.e. from the current position (state). Various options exist (some known and some unknown - this is key), but the path between them isn't predefined. Using the information available (context) a decision is made, results in a move to a different state. i.e. What happens next determines where you go, (GPS - dynamic re-routing vs routes marked on a Map) The path emerges from execution but isn't known ahead of time. Workflows vs State Machines ⚖️ Workflows (including procedural knowledge approaches): are predefined sequences of steps, the path is predefined and decreed up front e.g. Step 1 → Step 2 → Step 3. All options are not only known but the sequence is typically defined up front (although can evolve- but not typically at run-time) Note: trying to make workflows dynamic is trying to fit a square peg in round hole, probably can engineer, but it's the wrong pattern Both valid but very different behaviours for different problem spaces! 🤖 Agent → Uses State Machines 🔧 Traditional automation → Typically uses workflows How Agents navigate State Machines They observe state (current goal or state in its execution process), evaluate options and choose the next state based on context (all info needed to make the decision), then execute and arrive at new state. Agents also operate in an unknown state space i.e. they can create new options as they go. This fundamentally breaks predefined flow approaches. 💡 This repeats. The path isn't scripted, it's discovered at runtime. Context Graphs: A context graph is the structure an agent navigates: - Current state - Available transitions - Execution history + All the decision and other information used to make the decision to the appropriate transition (judgment, decision criteria etc, the "context" used) It's live at runtime and persisted as decision trace when complete, with the accumulated traces becoming organisational memory. Why The Confusion? 🤔 Context graphs sit at the intersection of data, software, and AI. Each discipline sees it through their own lens. We can def debate if the graph is just a log (the FC Capital POV) or also moves into the actual execution (what we've built) but first we must acknowledge the fundamental problem space we are in. The terminology will sort itself out ✌️ Nebulyx AI | 37 comments on LinkedIn

Starting the year with something we’ve been building quietly 🚀

From instance to structure to code to project ...

The missing step in Decision Intelligence — exposed by AI / Agentic Systems

Wow, my post on ontologies, context graphs, and intent maps got a lot more interactions than I expected.

Large Language Models (LLMs) have gained the ability to assimilate human knowledge and facilitate natural language interactions with both humans and other LLMs. However, despite their impressive...

Today, discussions about ontologies, semantic graphs, and knowledge graphs are often clouded by a semantic blur. Why? Because two fundamental levels are frequently confused: 1️⃣ Human interpretation Symbols (words, concepts, terms) gain meaning through our ability to link concepts to real-world refe

#GraphLandscape2026 We’re starting the year with a fresh view of the graph landscape! After sharing our first version of the Graph Landscape 2026 back in November, we received great feedback from the community. Based on your input, we’ve decided to make additions, refinements, and adjustments — and created a brand new edition in PDF. This report reflects: 👉 emerging trends across the graph ecosystem 👉 evolving technologies and use cases 👉 new and maturing players shaping the market Happy reading! #Technology #GraphAnalytics #KnowledgeGraphs

All the talk around Context Graphs is exciting but misses a few key points:

LLM have fundamentally changed how we interact with data , automate reasoning and build intelligent systems .However , despite their…

There’s a recurring tension in how people talk about the semantic spectrum and especially ontologies. On one side, ontology is treated as the peak.

Foundation Capital's piece on Context Graphs as AI's next trillion-dollar opportunity is making the rounds.

Can you explain what is your function and where it fits in the overall drug development and commercialization process? I am a director of data at BMS and I work with translational data. This includes areas such as translational safety, translational medicine, and reverse translation.

AIM MarketView: Graph Databases 2026 is now live 🔴 Graph databases are rapidly emerging as core platforms for relationship-centric analytics, knowledge graphs, and agentic AI–enabled decision systems. The AIM MarketView: Graph Databases 2026 provides a comprehensive view of the global graph database ecosystem, profiling 30 leading vendors and analyzing how platforms are evolving across scale, intelligence, and enterprise adoption. 🔍 What’s inside the report: - Landscape of 30 graph database vendors spanning property graph, RDF/semantic, multimodel, cloud-native, and open-source architectures - Vendor-specific use cases and commercial models for every profiled graph DB/platform. - Deep dive into Graph + AI convergence, including GraphRAG, vector–graph integration, and graph-based reasoning - Adoption insights across financial services, cybersecurity, life sciences, telecom, retail, manufacturing, and the public sector As enterprises move beyond isolated analytics toward context-aware, connected intelligence, graph databases are becoming foundational to both AI-native architectures and agentic systems. Aerospike | Altair | Amazon | Arango | BangDB | FalkorDB | Fluree | Franz Inc. | Google | Graphwise | HugeGraph | Hypermode | InfiniteGraph | JanusGraph | Microsoft Azure | Memgraph | Neo4j | NebulaGraph, powered by Vesoft | OrientDB | Oracle | Oxford Semantic Technologies (RDFox) | Progress MarkLogic - Progress Data Platform | Rocketgraph | Sparsity Technologies (Sparksee) | Stardog | TerminusDB | TigerGraph | Ultipa #VelocityGraph #GraphDatabases #KnowledgeGraphs #DataEngineering #AgenticAI #GraphRAG #AIMMarketView #AIM

Why Decision-Aware AI Requires Managed Meaning, Not Just Memory

We all feel overwhelmed by information. Data everywhere. Connecting the dots feels impossible. Unstructured, messy data will never power truly smart AI. But there is a solution: structuring knowledge into graphs. 🚀 Introducing 𝘗𝘦𝘳𝘴𝘦𝘶𝘴, a Text-to-Graph SDK by Lettria. Automatically build tailor-made Knowledge Graphs and GraphRAG pipelines in minutes. ➡️ +𝟰𝟱% 𝗶𝗺𝗽𝗿𝗼𝘃𝗲𝗱 𝗽𝗲𝗿𝗳𝗼𝗿𝗺𝗮𝗻𝗰𝗲 on average compared to traditional RAG. Why Perseus? 1️⃣ Super simple to use: Ingest PDF ⇒ convert to .md/.txt ⇒ generate a .ttl knowledge graph ⇒ accurate retrieval. 2️⃣ Edge reliability: Domain-and business specific automatic graph building. No illogical links, no weird nodes. Just clean communities and relationships. 3️⃣ Ontology guided (optional): Bring your own ontology (or use one of Lettria's) to guide the graph building with your custom logic. What you can build: 🔗 GraphRAG pipelines 🧠 AI agent persistent memory 🤖 Multi-agent systems 📊 Analytics & BI Example: python pdf_to_markdown.py mydoc. pdf python index.py mydoc. md python report. py"What are the main topics?" ✅ Coding agents: Claude Code, Cursor, Codex compatible. ✅ Export to Graph DBs: Neo4j native integration. ✅ Compatible with AWS Neptune, Ultipa, TigerGraph, ArangoDB, JanusGraph. 👉 We're opening Premium Early Access for a few weeks. Install the package from GitHub: https://lnkd.in/dKbWV429 🔑 Get your API key from the Perseus web app 👥 Join our Graph Discord community (links in the comments).

🧠 Context graphs are increasingly discussed as a foundational layer for agentic systems—framed by Jaya Gupta and Ashu Garg as a “trillion-dollar opportunity”—but their real value lies in preserving decision context.

Context Graphs and Process Knowledge, It's The Same Thing

Some thoughts on the latest buzzwords

Free, open-source RDF & ontology visualization tools 👇

Building a self-improving knowledge graph that doesn’t just store data, it understands, validates, and evolves it