Timothy Spann

Posted on Mar 29

All Data and AI Weekly #235-30March2026

#mcp #llm #agenticai #snowflake

All Data and AI Weekly #235-30March2026

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

Tim's Corner: Cortex Code Everywhere

Welcome to issue 235! This week Snowflake made Cortex Code in Snowsight generally available and shipped native Windows CLI support, Agent Teams for multi-agent orchestration, and a wave of new agent skills. Batch Cortex Search entered public preview for high-throughput offline workloads, Agentic ML landed with Cortex Code for automating the full ML lifecycle, and the engineering team published a deep dive on solving the identifier mismatch problem in multi-engine lakehouses. On the community side, new guides cover dbt pipelines with Cortex Code, Snowflake Postgres migration via logical replication, and quantitative research with AI functions. I've been shipping new SNACK-AI projects all week - including an OSINT toolkit, a Cortex Code cheatsheet, and SnowGhostBreakers Advanced. Let's dive in!

Tim's Content

Recent Presentations

Beyond the Prompt: Future of AI Data Cloud - March 2026 - Slide deck
24 March 2026 Demo Notes - NYC Snowflake Users Group hands-on demo

New SNACK-AI Repositories

Repository	Description
SNACKAI-CoCo-UnstructuredDemos	End-to-end demos for processing unstructured data with Cortex Code and SNACK-AI, including a dev ports example
SNACKAI-CoCo-BeyondthePrompt	Companion repository for the Beyond the Prompt presentation on the future of AI in the Data Cloud
SNACKAI-CoCo-Snowghost2	Ghost-busting application v2 built with Cortex Code on Snowflake
SNACKAI-CoCo-SnowGhostBreakers-Advanced	Advanced ghost-busting analytics with enhanced detection and investigation workflows
SNACKAI-Cortex-Code-Cheatsheet	Quick-reference cheatsheet for Cortex Code commands, skills, and workflows
SNACKAI-CoCo-OSINT	Open-source intelligence toolkit built with Cortex Code and SNACK-AI

Upcoming Events: Catch Tim Live!

April 14: D4BNA - Virtual US
April 15: Nodes AI Conference - Virtual
April 13-17: D4BV - Virtual APAC
May 20: Budapest Data and AI - Budapest, Hungary - Tim Virtual speaking

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

Featured: Cortex Code - Now More Widely Available and Ready for Bigger Tasks

Snowflake published a major update on Cortex Code this week with four key announcements: Cortex Code in Snowsight is now generally available, the CLI now supports native Windows environments, Agent Teams enable coordinated multi-agent orchestration for large tasks, and new agent skills standardize expert workflows for cost intelligence, ML, Streamlit, OpenFlow and more.

Key Highlights:

Persistent AI coding agent in Snowflake Workspaces and Notebooks
Agentic coding, discovery, administration and cost management in Snowsight
Agent Teams spin up parallel subagents with shared task lists and dependency management
New skills: cortex-code-guide, developing-with-streamlit, openflow, cost-intelligence, machine-learning, cortex-ai-functions

Cortex AI & Agents

The Agent Context Layer for Trustworthy Data Agents

Snowflake published a foundational blog on the Agent Context Layer - explaining why "talk to your data" is table stakes and why trustworthy agents need context spanning semantics, identity, constraints, policy and provenance. Internal experiments showed that augmenting agents with a plain-text data ontology improved accuracy by +20% and reduced tool calls by ~39%.

Agentic ML in Snowflake

Agentic ML through Cortex Code automates the path from data to predictive insights. Data science teams can build production-ready ML solutions from natural language prompts, with skills covering training, deployment, distributed training, hyperparameter tuning, and monitoring. First National Bank of Omaha reports 10x productivity gains.

Best Practices Guide: https://www.snowflake.com/en/developers/guides/agentic-machine-learning-best-practices-cortex-code/

Batch Cortex Search (Preview)

The CORTEX_SEARCH_BATCH table function is now in public preview. Submit batch queries to a Cortex Search service for offline use cases with high throughput requirements - entity resolution, deduplication, or clustering tasks. Batch search can also query services that are currently suspended.

Resources:

Cortex Code AI Function Studio Skill

New open-source skill for creating, evaluating, and optimizing custom AI functions using Snowflake Cortex AI Complete.

GitHub: https://github.com/Snowflake-Labs/cortex-ai-function-studio-skill

Cortex Code Skills Library

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

GitHub: https://github.com/kellykohlleffel/cortex-code-skills

Apache Iceberg Updates

Solving Identifier Mismatch in Multi-Engine Lakehouses

A deep engineering dive into the identifier case sensitivity problem in multi-engine lakehouses. When Spark, Trino, Flink, and Snowflake share the same Iceberg catalog, identifier normalization conventions clash. Snowflake's solution: database-level CATALOG_CASE_SENSITIVITY configuration for catalog-linked databases, supporting both Hadoop-style (case-insensitive) and ANSI SQL-style (case-sensitive) conventions.

Apache Iceberg from Zero

New comprehensive guide from Snowflake Labs for getting started with Apache Iceberg from scratch.

GitHub: https://github.com/Snowflake-Labs/apache-iceberg-from-zero

Snowflake Platform Updates

Release 10.10 (Mar 22-25)

Key updates in the latest Snowflake release:

Interval Data Types (Preview): New INTERVAL YEAR TO MONTH and INTERVAL DAY TO SECOND data types for duration calculations
Batch Cortex Search (Preview): High-throughput batch search via CORTEX_SEARCH_BATCH table function
Cortex Updates: See the full list of Cortex updates in the release notes

Full Release Notes: https://docs.snowflake.com/en/release-notes/2026/10_10

Dynamic Table SCHEDULER Attribute (GA - Mar 26)

New SCHEDULER attribute for dynamic tables lets you control whether automatic refresh is enabled or disabled. When SCHEDULER = DISABLE, the table can only be refreshed manually with ALTER DYNAMIC TABLE ... REFRESH - critical for external orchestrators like dbt that need to manage individual refreshes without triggering the entire pipeline.

Block Public Stage Access with IP Exceptions (GA - Mar 20)

Block public access to Azure internal stages while maintaining an allowlist of permitted IP addresses or CIDR blocks via SYSTEM$BLOCK_INTERNAL_STAGES_PUBLIC_ACCESS_WITH_EXCEPTION.

DML Error Logging for Tables

Configure DML error logging for data load operations, capturing failed rows for investigation instead of failing the entire load.

Documentation: https://docs.snowflake.com/en/user-guide/data-load-overview#configure-dml-error-logging-for-a-table

ANOMALY_INSIGHTS: Get Hourly Consumption by Service Type

New method for the ANOMALY_INSIGHTS class providing hourly consumption metrics broken down by service type.

Documentation: https://docs.snowflake.com/en/sql-reference/classes/anomaly-insights/methods/get_hourly_consumption_by_service_type

Snowflake Notebooks in Workspaces

Updated documentation for Notebooks in Workspaces - collaborative, interactive notebooks with container runtime for AI/ML production workflows.

Documentation: https://docs.snowflake.com/en/user-guide/ui-snowsight/notebooks-in-workspaces/notebooks-in-workspaces-overview

Notebook Project Objects (NPO): Run Python Scripts in Snowflake

A new schema-level object type that lets you deploy and run Python scripts non-interactively on a warehouse or compute pool. Write and test locally, deploy as a Notebook Project Object, then run on-demand or schedule with a Task. Supports CI/CD via Snowflake CLI.

Resources:

OpenFlow Updates

OpenFlow Connector for SQL Server

The OpenFlow Connector for SQL Server replicates data from SQL Server databases to Snowflake in near real-time or on schedule using SQL Server Change Tracking. Supports SQL Server 2016-2022, Azure SQL Database, Azure SQL Managed Instance, AWS RDS, and Google Cloud SQL.

Documentation: https://docs.snowflake.com/en/user-guide/data-integration/openflow/connectors/sql-server/about

From the Community: Medium & Blogs

Article	Description
Building dbt Pipelines with Snowflake Cortex Code	Hands-on guide to building dbt pipelines using Cortex Code as your AI pair programmer
How to Build a Governed Metric Registry on Snowflake	Building trusted KPIs for humans and AI with a governed metric registry
Open Trust Center Severities in Your Organization	Snowflake SQL guide for querying Trust Center severity findings across your organization
What is Cortex Code and Why Does It Matter?	phData's overview of Cortex Code's capabilities and impact on data engineering workflows
The Data Engineer's Cortex Code Cheat Sheet	Quick-reference cheat sheet for data engineers getting started with Cortex Code
Building an MCP Ecosystem at Pinterest	How Pinterest built an enterprise MCP ecosystem for AI agent integrations
Spartera + Snowflake Partnership	New technology partnership announcement between Spartera and Snowflake

Quickstarts & Guides

Guide	Description
Quantitative Research with AI Functions and Cortex Code	Build quantitative research workflows using Cortex AI functions and Cortex Code
Build Real-Time Fraud Detection Model with Natural Language	Build a fraud detection model using natural language prompts in Snowflake ML
Supply Chain Assistant with Snowflake Intelligence	Build a supply chain assistant powered by Snowflake Intelligence
Intro to Feature Store	Getting started with Snowflake's Feature Store for ML workflows
Migrate to Snowflake Postgres Using Logical Replication	Migration guide for moving PostgreSQL workloads to Snowflake Postgres
Powering AI-Ready Data Products with HL7 Data	Unpack complex HL7 healthcare data into AI-ready data products
Neo4j Fraud Detection on Snowflake	Graph-powered fraud detection with Neo4j and Snowflake
Neo4j Manufacturing on Snowflake	Graph analytics for manufacturing use cases with Neo4j and Snowflake
Neo4j Subways on Snowflake	Subway route analysis with Neo4j graph analytics on Snowflake
Route Optimisation & Vehicle Route Plan Simulator	Build a route optimization and vehicle route planning simulator
Cortex Code Cookbook: SiS Dashboard Deploy	Cookbook recipe for deploying Streamlit in Snowflake dashboards with Cortex Code
dbt + CoCo Demo	Demo project showing dbt development with Cortex Code

GitHub: Snowflake Labs

Repository	Description
apache-iceberg-from-zero	Comprehensive guide to getting started with Apache Iceberg on Snowflake
snowflake_dcm_projects	Database Change Management project examples and templates
cortex-ai-function-studio-skill	Cortex Code skill for building and evaluating custom AI functions
blueprint-manager	Blueprint management tool for Snowflake project templates
sfguide-powering-AI-Ready-Data-Products-by-unpacking-complex-HL7-data	Companion code for the HL7 healthcare data guide
postgres-full-day-training	Full-day training materials for Snowflake Postgres

Snowflake Postgres

Full-Day Training

Comprehensive training materials for getting started and going deep with Snowflake Postgres.

GitHub: https://github.com/Snowflake-Labs/postgres-full-day-training

BlueBox

Open-source toolkit for Snowflake Postgres workflows and utilities.

GitHub: https://github.com/ryanbooz/bluebox

Data Trends & Resources

Data Trends 2026: Financial Services

Snowflake's report on key data and AI trends shaping the financial services industry in 2026.

AI Futures Model

Interactive resource exploring trajectories and scenarios for AI development.

Read more: https://www.aifuturesmodel.com/

How to Enable Tracing and Logging in Power BI Desktop

Community guide for configuring tracing and logging in Microsoft Power BI Desktop with Snowflake connections.

Security

How to Secure Agentic AI: Agent Identity Protocol (AIP)

All Things Open article on securing agentic AI with the Agent Identity Protocol - a framework for establishing trust and identity for AI agents.

Go Implementation: https://github.com/openagentidentityprotocol/aip-go

TeamPCP Supply Chain Attack on LiteLLM

Wiz research on the TeamPCP campaign trojanizing LiteLLM - critical reading for anyone running open-source AI inference proxies.

MCP & Agent Tooling

Tool	Description
Agent Browser (Vercel Labs)	Browser automation framework designed for AI agents from Vercel Labs
Agent Skill Creator	Framework for creating reusable skills for AI agents
x1xhlol/system-prompts-and-models-of-ai-tools	Collection of system prompts and model configurations from various AI tools

AI Research & Models

Cohere Transcribe

Cohere Transcribe 03-2026 - new speech-to-text model from Cohere Labs.

HuggingFace: https://huggingface.co/CohereLabs/cohere-transcribe-03-2026

MolmoWeb by Allen AI

MolmoWeb - multimodal language model interface from the Allen Institute for AI.

Read more: https://allenai.org/blog/molmoweb

Deer Flow by ByteDance

Deer Flow - agent workflow framework from ByteDance for building complex multi-step AI applications.

GitHub: https://github.com/bytedance/deer-flow

Meetily

Open-source AI meeting assistant for transcription, summarization, and action items.

GitHub: https://github.com/Zackriya-Solutions/meetily

Developer Tools

Terminal & Shell

Tool	Description
lnav	Advanced log file navigator for the terminal
cliamp	CLI audio/media player

Data Visualization & Infrastructure

Tool	Description
Terravision	Automated infrastructure visualization from Terraform files
MiniStack	Minimal cloud infrastructure stack for development

Documents & Screenshots

Tool	Description
ProofShot	Screenshot verification and proof-of-work tool
RevPDF	PDF reverse engineering and extraction tool

Frameworks & Libraries

Tool	Description
Yggdrasil	Event sourcing and CQRS framework
Keploy	API testing platform that auto-generates test cases from API calls
VoidLLM	Lightweight LLM inference engine

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.

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: