CtxEngineer @ctxengineering

The Azure Functions MCP extension has had a breakout year 🚀 From "tools only" → a full MCP platform: ✅ tools, resources, prompts ✅ MCP Apps (interactive UI) ✅ built-in auth + rich types and more! Check out: aka.ms/functions-mcp-… #MSBuild

Go support! yihaaaa

Azure Functions @AzureFunctions

3 days ago

Azure Functions at Build 2026: serverless agents in a .agent.md file, 1,400+ M365 and SaaS connectors as triggers, Go as a first-class language, full MCP support, new CLI, On-demand Sandboxes for DTS and much more! techcommunity.microsoft.com/blog/appsonazu…

0 12 30 6K 15

Introducing OmniVec: An Open-Source Embedding Platform for AI Apps on Azure - Azure Cosmos DB Blog devblogs.microsoft.com/cosmosdb/intro…

@Michael_Kissig Egal ob pro oder contra bitcoin. Ein Anlageberater, der mit „innerer Wert“ argumentiert, wird mich nie beraten. Es gibt keinen intrinsic value zum 100. Mal. Konsens über Wertevorstellung ist subjektiv, sonst nichts.

idk who created this website, but it's awesome and i love it. sophon.at/papers

!

The Year of the Graph @TheYotg

a week ago

Graph Engines: Introducing Neo4j Virtual Graph - Graph reasoning on the data you already have Unlocking the power of graph intelligence on all your enterprise data - no ETL needed. Most graph conversations start with "move your data into a graph database." But what if you

1 3 12 531 7

Excited to launch @neo4j Virtual Graph 🚀 Graph reasoning directly on enterprise data. no ETL, no copies, Cypher pushed into Any Lakehouse. Helping make enterprise data accessible to AI agents. lnkd.in/eeSTkaCP #GraphAI #EnterpriseAI #AIAgents #Snowflake #Databricks

Today we’re launching Command Line, a new technical blog about how and why we build. Our first stories go behind the architecture, design decisions, and research shaping the next wave of developer tools. Stay tuned for more next week at #MicrosoftBuild msft.it/6013vkN8f

@Proton_Pass @ProtonPrivacy Penis?

HNSW is fast. But it is also memory-hungry. DiskBBQ is the alternative. Here's how it works: 1. Hierarchical K-means partitions all vectors into clusters, each with a representative centroid 2. The centroids are arranged into a tree and loaded into memory. The vectors themselves stay on disk 3. At query time, the centroid tree is searched first to find the most promising clusters. Only those get fetched and fully explored. 4. BBQ (Better Binary Quantization) compresses each vector to 1 bit per dimension, so bulk cluster scoring stays fast with minimal memory and disk overhead 5. SOAR (Spilling with Orthogonality-Amplified Residuals) assigns border vectors to multiple clusters, which reduces the recall penalty of hard cluster boundaries Use HNSW when you need maximum recall and have plenty of memory. Use DiskBBQ when you're cost-sensitive and can live with slightly lower recall.

Understanding Microsoft Fabric Ontology jamesserra.com/archive/2026/0… #DataScience #MachineLearning

🎉 Generative UI, AG-UI and A2UI at Google I/O Google 'gets' this space, and we love collaborating with them on it. quick brain dump on where Generative Interfaces stand today: 1. All UI will be AI. This is CopilotKit's mantra. Every interaction between humans and technology will be mediated by agentic systems — from agentic SaaS, to Slack agents, to an agentic fridge. Agentic intelligence will be a part of every loop, and many loops will contain an interaction step with a user. 2. GenUI sits on a spectrum. On one end the developer controls the UI at build-time on the other the agent controls the UI at runtime. From tightly controlled, to declarative, to fully open-ended. CopilotKit is the only toolkit with primitives across that whole spectrum. 3. Agents break request-response. The contract that has defined the frontend-backend boundary for 30 years doesn't fit agents. They're long-running, they share control with users, they stream output. This was the original reason we built AG-UI — an event-based transport layer made for the agentic age. 4. Builders feel this pain constantly. Fragile glue code, unclear handoffs, terrible UX wiring agents into frontends. We hear it all the time, and folks are always relieved to learn someone has tackled it. we build this stuff because Generative Interfaces are one of the most meaningful shifts happening in software right now. if we can make starting on it less scary, that's a win github.com/CopilotKit/Cop…

Context graph apps shouldn’t take a week to scaffold. Create Context Graph is an interactive CLI scaffolding tool for building domain-specific context graph applications with Neo4j. It helps you get from idea to working app faster by asking for your domain, agent framework, and Neo4j setup, then generating the backend, frontend, schema, agent tools, and sample data around that choice. Key features: • Interactive CLI wizard – pick your domain, framework, and Neo4j connection instead of wiring the stack by hand • Generated full-stack app – FastAPI backend plus a Next.js/Chakra UI frontend with streaming chat and graph visualization • Domain-specific graph setup – creates Neo4j constraints, indexes, GDS projections, ontology-driven models, and tailored Cypher tools • Demo or real data path – use synthetic documents and decision traces, or import from tools like GitHub, Slack, Gmail, Jira, Notion, Calendar, and Salesforce • Multiple agent frameworks – supports options including PydanticAI, Claude Agent SDK, OpenAI Agents SDK, LangGraph, CrewAI, Strands, Google ADK, and Anthropic Tools It’s open-source (Apache-2.0 license). Link in the reply 👇

Exact kNN on 1M vectors: ~10 QPS HNSW on the same dataset: ~1000 QPS This is because HNSW builds a layered graph: shortcuts at the top, all vectors at the bottom. A query enters at the top layer, finds the closest node, uses it as the entry point for the layer below, and repeats until it reaches a tight local search at layer 0. It also let you tune your indexing speed vs query speed vs recall, so you can get the index that suits your needs.

INTRODUCING HELIX-DB v2!!!💥🚀 HUGE changes coming to Helix-DB: - New query methods 🔎 - Object-storage backed 🪣 - Query insights 📝 - Vectors on nodes AND edges 🌐 More info below 👇🏻

Microsoft’s Intelligence Stack Explained: Work IQ, Fabric IQ, Foundry IQ & Project Opal medium.com/@umeshp2188/mi… #MachineLearning #ChatGPT

Spending half your budget on memory just to keep your vector index running? There's a better way. HNSW is the gold standard for vector search, but it needs everything in memory. As datasets grow, that gets expensive. HFresh flips the model by storing vectors on disk while keeping only a compact centroid index in memory. 𝗛𝗼𝘄 𝗶𝘁 𝘄𝗼𝗿𝗸𝘀 • Divides vectors into small regions called postings • Uses a compact in-memory HNSW index over centroids to identify relevant regions • Fetches only the relevant postings from disk for search • Applies Rotational Quantization at two levels for compression The result is significantly lower memory usage with predictable latency, even at billion-vector scale 𝗪𝗵𝗲𝗻 𝘁𝗼 𝗰𝗼𝗻𝘀𝗶𝗱𝗲𝗿 𝗛𝗙𝗿𝗲𝘀𝗵 𝗟𝗮𝗿𝗴𝗲 𝗱𝗮𝘁𝗮𝘀𝗲𝘁𝘀 𝘄𝗶𝘁𝗵 𝗵𝗶𝗴𝗵-𝗱𝗶𝗺𝗲𝗻𝘀𝗶𝗼𝗻𝗮𝗹 𝗲𝗺𝗯𝗲𝗱𝗱𝗶𝗻𝗴𝘀 — Memory savings become substantial compared to HNSW 𝗖𝗼𝘀𝘁-𝘀𝗲𝗻𝘀𝗶𝘁𝗶𝘃𝗲 𝗱𝗲𝗽𝗹𝗼𝘆𝗺𝗲𝗻𝘁𝘀 — Run the same workload on smaller infrastructure 𝗪𝗿𝗶𝘁𝗲-𝗵𝗲𝗮𝘃𝘆 𝘄𝗼𝗿𝗸𝗹𝗼𝗮𝗱𝘀 — The cluster-based design avoids the write amplification that HNSW can experience during large imports HFresh is currently a technical preview, so we recommend testing in non-production environments first. Read our blog for more details: weaviate.io/blog/weaviate-… Try HFresh now in Weaviate Cloud console.weaviate.cloud/signin?utm_sou…

@CandideSurLaMer @ProtonDrive @ProtonPrivacy Agree, it is sn oligopoly, but on the other side innovation and disruption have never needed market influence. If you you look for the same market segment, it is difficult.

Peekaboo 3.0 is live. Biggest release since 2.0. ⚡ Action-first macOS computer use 👁️ Unified screenshot + UI detection 🧩 Cleaner JSON across CLI + MCP 🛠️ Better snapshots I started this last year, but the models just weren’t good enough. Now they are. peekaboo.sh

LLM Wikis + HTML Artifacts are insanely powerful. You should seriously consider this in your workflows. LLM Wikis captures all the important information that lets you and your agents do meaningful work. HTML artifacts present that information in interesting ways that allow you to take important actions along with your agents. My HTML artifacts sit on top of my LLM wikis. They are dynamic and are easily extended as needs arise. I have hooked my Artifacts to talk to my agents, and similarly, the agents can talk to artifacts. This has allowed me to build powerful artifacts that reduce my inbox to zero, keep me updated on any topic of interest, fast prototyping, do deep research, design/trigger new experiments, generate figures to improve understanding, schedule research, search relevant information, discover topics, and so much more. What you see in the clip is not a website. It's a simple interactive HTML artifact. HTML artifacts are useful for designers, engineers, researchers, students, and anyone working with agents. Lastly, HTML doesn't replace Markdown. They are a much better combination working together.