DEV Community

Cover image for All Data and AI Weekly #234 - 23 March 2026
Timothy Spann
Timothy Spann

Posted on

All Data and AI Weekly #234 - 23 March 2026

All Data and AI Weekly #234 - 23 March 2026

( AI, Data, Agentic AI, Cortex Code, NiFi, Iceberg, Polaris, Streamlit, Python, Java, SQL, MCP, LLM, RAG, Cortex AI, AISQL, Search, Unstructured Data )

The Agentic Enterprise Takes Shape

Welcome to issue 234! This is a landmark week for the Snowflake ecosystem. Snowflake officially unveiled Cortex Code to the world with a major press release, the Agentic Enterprise Control Plane vision dropped, Cortex Agent Evaluations are now documented, and DCM Projects bring infrastructure-as-code to Snowflake. On the Iceberg front, bidirectional interoperability with Horizon Catalog is here, plus Azure Data Lake Storage support for external volumes. I've been shipping new SNACK-AI projects all week - including a Situation Desk, OpenFlow Python Processors, and a DJ app built entirely in Cortex Code. Tomorrow I'll be at the NYC Snowflake Users Group for a hands-on workshop. Let's dive in!

alldataaiweekly23march2026


New SNACK-AI Repositories

Repository Description
SNACKAI-CoCo-situation-desk A real-time situation awareness desk built with SNACK-AI and Cortex Code, providing unified monitoring dashboards for streaming data events, alerts, and operational intelligence
SNACKAI-OpenFlow-Python-Processors Custom Python processors for Snowflake OpenFlow, enabling data transformation, enrichment, and routing logic within OpenFlow pipelines using Python
SNACKAI-CoCo-Quick Quick-start templates and utilities for bootstrapping SNACK-AI projects with Cortex Code, get up and running fast
Semantic Layer Quality Scoring Framework A comprehensive framework for scoring and evaluating the quality of your Snowflake semantic layer - covering completeness, consistency, and accuracy metrics
SNACKAI-CoCo-DJ-Snowflake A DJ and music application built entirely with Cortex Code on Snowflake, showcasing creative AI-powered app development
SNACKAI-CoCo-MarkdownBrowser A markdown-based browser built with Cortex Code, enabling rich document rendering and navigation within the SNACK-AI ecosystem

Video


Upcoming Events: Catch Tim Live!

All Snowflake Events: https://www.snowflake.com/about/events/

Snowflake Virtual Hands-On Labs

Get hands-on with Snowflake in these upcoming free workshops:

Date Lab Link
Mar 18 Agent to OpenFlow: End-to-End Unstructured Data Insights Register
Mar 18 Zero to Snowflake in 90 Minutes Register
Mar 24 From Prompt to Production: Real-Time Fraud Detection Register
Mar 25 Build Faster with Cortex Code AI Register
Mar 25 Getting Started with Snowpark on Iceberg Register
Apr 1 Cortex Code: Your AI Coding Assistant in Snowsight Register

Browse all labs: https://www.snowflake.com/en/webinars/virtual-hands-on-lab/?tags=region%2Famericas&page=1&pageSize=6&offset=6


Major Announcement: Snowflake Unveils Cortex Code

Snowflake officially announced Cortex Code - an AI coding agent that drastically increases developer productivity by understanding your enterprise data context. This is a major milestone: an AI assistant that doesn't just write code, but understands your Snowflake environment, your data, your semantic layers, and your governance policies.

Key Highlights:

  • AI coding agent native to Snowflake
  • Deep understanding of enterprise data context
  • Integrated with Snowsight and CLI
  • Custom skills and extensibility support
  • Governance-aware code generation

Press Release: https://www.snowflake.com/en/news/press-releases/snowflake-unveils-cortex-code-an-ai-coding-agent-that-drastically-increases-productivity-by-understanding-your-enterprise-data-context/


Featured: The Agentic Enterprise Control Plane

Snowflake published a vision piece on the Agentic Enterprise Control Plane - a unified layer for orchestrating, governing, and monitoring AI agents across the enterprise. This is where enterprise AI is heading: agents that operate within defined guardrails, coordinated by a central control plane built on trusted data.

Key Concepts:

  • Centralized agent orchestration
  • Policy-driven governance for AI agents
  • Observability and audit trails
  • Integration with existing data infrastructure

Read more: https://www.snowflake.com/en/blog/agentic-enterprise-control-plane/


Cortex AI & Agents

Cortex Agent Evaluations

The Snowflake engineering team published a comprehensive guide on evaluating Cortex Agents - covering metrics, benchmarks, and systematic approaches to measuring agent reliability and accuracy. Critical reading for anyone deploying agents in production.

Resources:

Cortex Code Governance Skills

New engineering blog on building governance-aware skills for Cortex Code. These skills understand your data policies, masking rules, and access controls, ensuring AI-generated code respects your security posture.

Read more: https://www.snowflake.com/en/engineering-blog/cortex-code-governance-skills/

Cortex Code in Snowsight (Preview - Mar 12)

Cortex Code AI Suggestions are now available in Preview directly within Snowsight, bringing intelligent code completion and AI assistance to the web IDE.

Documentation:

Cortex Agents Documentation

Updated comprehensive documentation for building, deploying, and managing Cortex Agents.

Documentation: https://docs.snowflake.com/en/user-guide/snowflake-cortex/cortex-agents

Snowflake Intelligence Artifacts (Mar 19)

New Artifacts feature for Snowflake Intelligence allows agents to produce, store, and reference structured outputs like reports, charts, and data summaries.

Documentation:

Classification ML Function - Predict Method

Updated documentation for the Classification predict method, with new examples for applying trained classification models to new data.

Documentation:

Cortex Code Best Practices

The definitive guide to getting the most out of Cortex Code CLI - covering data exploration, code generation, skill development, and workflow optimization.

Read more: https://www.snowflake.com/en/developers/guides/best-practices-cortex-code-cli/


Apache Iceberg Updates

Bidirectional Interoperability with Horizon Catalog

Major engineering blog on bidirectional interoperability between Apache Iceberg and the Snowflake Horizon Catalog. Read and write Iceberg tables from Snowflake while external engines (Spark, Trino, etc.) simultaneously access the same data.

Read more: https://www.snowflake.com/en/engineering-blog/bidirectional-interoperability-iceberg-snowflake-horizon-catalog/

External Query Engine with Horizon Writes (Mar 16)

New feature allowing external query engines to write to Iceberg tables managed by Snowflake Horizon.

Read more: https://docs.snowflake.com/en/release-notes/2026/other/2026-03-16-tables-iceberg-query-using-external-query-engine-snowflake-horizon-writes-feature

Azure Data Lake Storage External Volumes (Mar 20)

Support for Azure Data Lake Storage as external volumes for Iceberg tables, expanding multi-cloud options.

Documentation:


DCM Projects: Infrastructure-as-Code for Snowflake

Database Change Management (DCM) Projects bring declarative, version-controlled infrastructure management to Snowflake. Define your databases, schemas, tables, and roles in code and deploy them reproducibly.

Key Features:

  • Declarative resource definitions
  • Version-controlled database changes
  • Reproducible deployments
  • CI/CD integration

Documentation:


Snowflake Platform Updates

Release 10.9 - TSS History View (GA)

The TSS_HISTORY account usage view is now Generally Available, providing historical data on Tri-Secret Secure configurations.

Read more: https://docs.snowflake.com/en/release-notes/2026/10_9#tss-history-account-usage-view-general-availability

OpenFlow BigQuery Connector (Preview - Mar 17)

The OpenFlow connector for BigQuery enters Public Preview, enabling CDC from Google BigQuery into Snowflake.

Read more: https://docs.snowflake.com/en/release-notes/2026/other/2026-03-17-openflow-bigquery-pupr

User-Defined Data Types

New support for user-defined data types in Snowflake SQL, enabling custom type definitions for stronger data modeling.

Documentation: https://docs.snowflake.com/en/sql-reference/data-types-user-defined

Listing Access History

New listing access history view for tracking who accessed shared data listings and when.

Documentation: https://docs.snowflake.com/en/sql-reference/data-sharing-usage/listing-access-history

Listings BCDR

Business Continuity and Disaster Recovery documentation for Snowflake Marketplace listings.

Documentation: https://docs.snowflake.com/en/collaboration/listings-bcdr

Alerts UI (Limited Access)

New Alerts UI for managing and monitoring Snowflake alerts visually.

Documentation: https://docs.snowflake.com/en/LIMITEDACCESS/alerts-ui

Snowflake Notebooks

Updated product page for Snowflake Notebooks - collaborative, interactive notebooks natively in Snowflake.

Read more: https://www.snowflake.com/en/product/features/notebooks/

Snowflake Postgres High Availability

Documentation on high availability for Snowflake Postgres deployments.

Documentation: https://docs.snowflake.com/en/user-guide/snowflake-postgres/high-availability


Project Snowwork: AI for Business Users

Project Snowwork aims to bring AI capabilities directly to business users - no coding required. Build intelligent workflows, automate routine tasks, and get AI-powered insights from natural language.

Read more: https://www.snowflake.com/en/blog/project-snowwork-business-users/


Snowflake Open Source

Arctic Training

Open-source training framework for Arctic language models from Snowflake.

GitHub: https://github.com/snowflakedb/ArcticTraining

Arctic Inference

Optimized inference engine for Arctic models with high-throughput serving.

GitHub: https://github.com/snowflakedb/ArcticInference

Snowflake Telemetry Python

Python SDK for Snowflake telemetry - instrument your applications with traces and metrics.

GitHub: https://github.com/snowflakedb/snowflake-telemetry-python

SSv2 AI Webinar

End-to-end demo for Snowpipe Streaming V2 with AI, including live dashboards and Cortex Agent integration.

GitHub: https://github.com/sfc-gh-chathomas/SSv2-AI-Webinar

CoCo Skills Library

Community-contributed Cortex Code skills library with reusable skills for common workflows.

GitHub: https://github.com/vinodhini-sd/coco-skills-library


From the Community: Medium & Blogs

Deep Technical Looks at Cortex Code

Article Author Description
A Deep Technical Look at Cortex Code from a Snowflake Architect's Perspective Cadarsh Hegde Comprehensive architectural analysis of Cortex Code internals, extensibility model, and integration patterns from an architect's viewpoint
Meet Your New Teammate: A Deep Dive into Snowflake Cortex Code Abhishek Mittal Getting started guide exploring Cortex Code as an AI pair programmer for Snowflake development
Cortex Code Decoded: Teach Your Agent New Tricks with Custom Skills Maseed Ilyas Step-by-step tutorial on building custom skills to extend Cortex Code capabilities

Architecture & Strategy

Article Author Description
From Self-Service BI to Autonomous Intelligence Snowflake Architectural blueprint for evolving from traditional BI to autonomous, AI-driven intelligence platforms
Architecting Multi-Agent Workflows with Snowflake Cortex Snowflake Design patterns for orchestrating multiple AI agents in Cortex, covering coordination, state management, and error handling
From Legacy BI to an Intelligence Layer Snowflake Lessons learned from Snowflake's own BI migration journey to an AI-powered intelligence layer

External Blogs

Article Description
NanoCortex Blueprint: Snowflake Cortex Agents Minimal, clean blueprint for building Cortex Agents with a focus on simplicity and reproducibility
Snowflake Cortex Proxy for AI Coding Agents How to set up a Cortex inference proxy so external AI coding agents can use Snowflake's LLM endpoints
Markdown as Agentic UI Using markdown as a universal interface for building agentic user experiences
Agent-Native Design Guide to designing applications that are built from the ground up for AI agent interaction
How Booking.com Uses AI and Data How Booking.com creates connected customer trip experiences using AI and data at scale

Semantic Views: Why We Need Them

The Snowflake engineering blog published a foundational article on why semantic views matter - bridging the gap between raw data and business meaning, enabling consistent metrics across tools, and powering natural language queries.

Read more: https://www.snowflake.com/en/engineering-blog/why-we-need-semantic-views/


dbt Labs Updates

ADE-Bench: Data Engineering Benchmarking

The Analytics Data Engineering Benchmark (ADE-Bench) from dbt Labs provides standardized benchmarks for evaluating AI-assisted data engineering. Measures how well AI agents can write correct, efficient dbt models.

Resources:


Videos

Title Description Link
SNACK-AI Demo Walkthrough Tim walks through the latest SNACK-AI architecture Watch
Cortex Agent Evaluations Deep dive into evaluating Cortex Agents Watch
Snowflake AI & Data Engineering Snowflake AI workflows and data pipelines Watch
Building with Cortex Code Hands-on Cortex Code development session Watch
Agentic AI Patterns Patterns for building AI agent systems Watch
Data Pipeline Architecture Modern data pipeline design Watch
Snowflake Intelligence Deep Dive End-to-end Snowflake Intelligence walkthrough Watch
SNACK-AI Architecture SNACK-AI architectural patterns explained Watch
Cortex Code Skills Building custom skills for Cortex Code Watch
Real-Time Data with OpenFlow OpenFlow for real-time ingestion Watch
Iceberg on Snowflake Apache Iceberg integration patterns Watch

MCP & Agent Tooling

Tool Description
Snowflake Cortex Inference Proxy Proxy server that lets external AI coding agents use Snowflake Cortex LLM endpoints with policy controls and prompt management
MCP-MAT MCP management and testing toolkit for developing and debugging MCP integrations
Meshtastic MCP MCP server for Meshtastic mesh networking devices, enabling AI agents to communicate over LoRa
Coalesce MCP MCP integration for Coalesce data transformation platform
iMCP iOS/macOS MCP client for running Model Context Protocol servers on Apple devices
Ask Search MCP-compatible search tool for AI agent web search

AI Research & Models

LumberChunker: Document Segmentation

Carnegie Mellon published research on LumberChunker - an intelligent approach to segmenting long-form narrative documents for RAG pipelines. Beats naive chunking strategies significantly.

Read more: https://blog.ml.cmu.edu/2026/03/17/lumberchunker-long-form-narrative-document-segmentation/

Mistral Small 4

Mistral Small 4 released - the latest compact model from Mistral AI with improved reasoning and instruction following.

Read more: https://mistral.ai/news/mistral-small-4

MiniMax M2.7

MiniMax M2.7 is now available on Ollama - a compact but capable model for local inference.

Run it: https://ollama.com/library/minimax-m2.7

Open SWE by LangChain

Open SWE from LangChain - an open-source software engineering agent framework for autonomous code editing and debugging.

GitHub: https://github.com/langchain-ai/open-swe

AutoResearchClaw

AI-powered automated research assistant that can search, read, and synthesize academic papers.

GitHub: https://github.com/aiming-lab/AutoResearchClaw

Agent Safehouse

Secure sandbox environment for testing and developing AI agents without risk to production systems.

Website: https://agent-safehouse.dev/

DiMOS

Dimensional OS - an operating system abstraction for building multi-modal AI agent systems.

GitHub: https://github.com/dimensionalOS/dimos


Developer Tools

Browsers, Crawlers & Headless

Tool Description
Lightpanda Browser Lightweight, fast headless browser purpose-built for AI agents and web automation
HeadlessX Advanced headless browser automation framework for scraping and testing
SiteFetch Turn any website into LLM-ready text with a single command

Terminal & Shell

Tool Description
Ghostty Fast, feature-rich terminal emulator built for modern workflows
NVIDIA OpenShell NVIDIA's open-source shell environment for GPU-accelerated workflows
Open Terminal Terminal emulator from the Open WebUI team, designed for AI-powered interactions
Tmux Guide Comprehensive visual guide to tmux terminal multiplexer
MPP.dev Modern process and project management from the terminal

Code, Search & Productivity

Tool Description
TinyCode Minimal, fast code editor for quick edits
QMD Quick markdown processing and rendering utilities
Kanban Todo Simple kanban board in the terminal for task management
OpenCode Open-source AI coding assistant
Goal-MD Goal-oriented markdown for structured project planning

Audio & Voice

Tool Description
NeuTTS Neural text-to-speech engine with natural voice synthesis
Audiblez Convert text documents into audiobooks with AI voices
VoiceTag Voice-based tagging and annotation system
KittenTTS Lightweight text-to-speech library

PDF & Documents

Tool Description
TinyPDF Lightweight, fast PDF processing library

Networking & Infrastructure

Tool Description
Rustunnel Fast, secure tunneling written in Rust
Project Nomad Self-hosted infrastructure and networking toolkit
Cockpit Web-based server management interface
Mailrise SMTP gateway for push notifications
Apprise Universal push notification library supporting 100+ services

Security & OSINT

Tool Description
Sn1per Automated penetration testing reconnaissance scanner
Awesome OSINT for Everything Curated collection of OSINT tools and resources
Shadow Fleet Tracker Open-source maritime vessel tracking and analysis
RedGridLink Tactical communication and coordination tool

Data & Search

Tool Description
Sonar Semantic search engine for local documents
Hacker News Dataset Complete Hacker News dataset on HuggingFace for analysis and training
Brave Search API Privacy-respecting search API for building AI applications
PURL.dev Package URL specification for universal software package identification

Design

Tool Description
Stitch Design MD Google's markdown-based design system documentation framework

ROI of Generative & Agentic AI

Snowflake published extensive research demonstrating the radical ROI organizations are seeing from generative and agentic AI implementations.

Resources:


Data Sharing & Open Table Formats

Snowflake's vision for data sharing with open table formats enables seamless data exchange while maintaining interoperability across engines and platforms.

Read more: https://www.snowflake.com/en/blog/data-sharing-open-table-formats/


SNACK-AI: The 2026 Pattern

As we look forward in 2026, SNACK-AI is the architectural pattern for powering AI applications at any scale - from mobile phones and robots to massive enterprise clusters.

SNACK-AI integrates:

  • Snowflake Platform
  • NiFi (Apache)
  • Apache Iceberg
  • Cortex AI (Agents/Search/MCP)
  • Kafka (Apache Kafka)

Deep Dive into SNACK-AI:


Thanks

https://sessionize.com/tspann

https://github.com/timothyspann

© 2020-2026 Tim Spann https://www.youtube.com/@FLaNK-Stack

Generated with Cortex Code for All Data and AI Weekly - March 23, 2026

Top comments (0)