All Data and AI Weekly #251 - 20 July 2026

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

Tim's Corner: Gemini 3 on Cortex, Agentic Discovery, Air Quality IoT, and Prefect Acquires Dagster

Welcome to issue 251 of All Data and AI Weekly! This week headlines with Google's Gemini 3 arriving on Snowflake Cortex AI - enterprise-ready multimodal models now accessible directly in your data cloud. Snowflake published the Agentic Resource Discovery Specification - a new open standard for how AI agents find and use data resources. We got CREATE OR ALTER GA, Online Feature Store in Preview, and a new stored procedure for creating Semantic Views from OSSIE YAML. In the orchestration world, Prefect acquired Dagster - uniting two leading modern orchestrators. The community published deep guides on multi-agent systems via MCP in Cowork and Snowflake IaC with Terragrunt. On the personal side, wildfire smoke hit NJ again and I built two full SNACK-AI air quality monitoring systems (indoor + outdoor) end-to-end with Cortex Code, plus a new blog and video walking through it all. Let's go!

Tim's Content

Air Quality Deja Vu 2026

Wildfire smoke from Canada reached New Jersey again. Tim built full end-to-end air quality monitoring pipelines - indoor and outdoor - using Snowpipe Streaming V2, Cortex Agents, Semantic Views, Streamlit Dashboards, and React apps. Complete with HazeCam image analysis via AI_COMPLETE, real-time PM2.5/PM10 alerting, and Slack notifications.

Blog: Air Quality Deja Vu 2026

Video: Air Quality Monitoring with SNACK-AI and Cortex Code

GitHub (External Air Quality): SNACKAI-CoCo-AirQuality

GitHub (Internal Enviro Monitoring): SNACKAI-CoCo-EnviroUrban

Dev.to: Tim Spann on Dev.to

Upcoming Events

July 29: NYC Hands-On Workshop: AI Tools in Snowflake + Hex - NYC, 1:00-5:30 PM (In-Person, spots limited!)
Aug 17: How Snowflake Secures the Enterprise AI - Virtual
AAIF: Applied AI Foundation - Conference
AgentCon: Global AI Community AgentCon - Conference

Featured: Google Gemini 3 on Snowflake Cortex AI

Snowflake enables enterprise-ready AI by bringing Google's Gemini 3 to Snowflake Cortex AI. Access Google's latest multimodal foundation model directly within your Snowflake environment - with enterprise governance, security, and data locality built in.

Featured: Agentic Resource Discovery Specification

Snowflake introduces the Agentic Resource Discovery Specification - an open standard defining how AI agents discover, understand, and access data resources. This is foundational infrastructure for building agents that can autonomously navigate enterprise data landscapes.

Featured: Building Multi-Agent Systems in Snowflake Cowork the Clean Way

Ranjeeta Pegu shares two approaches to multi-agent orchestration in Snowflake Cowork - the hard way (UDFs + REST APIs + PAT tokens) and the clean way (Snowflake-Managed MCP Servers wrapping Cortex Agents). The MCP pattern lets you build hierarchical agent systems entirely within Snowflake without the 60+ lines of SSE parsing.

Cortex AI & Agents

Volatile UDF Support in Incremental Dynamic Tables (July 16)

Volatile UDFs are now supported in incremental Dynamic Tables - enabling more flexible transformations while maintaining incremental refresh semantics.

Online Feature Store (Preview - July 10)

The Online Feature Store is now in Preview - serve low-latency features for real-time ML inference directly from Snowflake.

Read more: Online Feature Store Preview

CREATE OR ALTER (GA - July 14)

CREATE OR ALTER is now Generally Available - modify existing objects in place without DROP + CREATE, preserving grants and dependencies.

Read more: CREATE OR ALTER GA

Create Semantic View from OSSIE YAML

New system stored procedure: SYSTEM$CREATE_SEMANTIC_VIEW_FROM_OSSIE_YAML - create Snowflake Semantic Views directly from Apache OSSIE YAML definitions. Bridges the open semantic standard to native Snowflake analytics.

Snowflake Platform & Guides

Snowflake IaC with Terragrunt: Multi-Environment Design and CI/CD

Tatsuya Koreeda on the Snowflake Builders Blog shares a comprehensive guide to adopting Terraform with Terragrunt for Snowflake infrastructure management. Covers multi-environment design, YAML-based resource definitions, state splitting by resource type, AI-assisted import workflows, and CI/CD with GitHub Actions.

Open Lakehouse Done Right: Iceberg, Governance & Cortex in One Lab

Snowflake developer guide on building an open lakehouse combining Apache Iceberg, governance, and Cortex AI - zero-tax analytics without vendor lock-in.

Data Ecosystem & Industry

Prefect Acquires Dagster: Uniting Two Leading Modern Orchestrators

Prefect acquires Dagster - uniting the two leading modern data orchestration platforms. A major consolidation in the data engineering tooling landscape.

Moonshine Micro: Voice Interfaces for Microcontrollers

Moonshine Micro is an open-source AI toolkit for real-time voice interfaces on microcontrollers - VAD, speech-to-text, and neural TTS running in 470 KB of RAM on an 80-cent Raspberry Pi RP2350. MIT licensed and commercial-ready.

Read more: Moonshine Micro on GitHub

VulnHunter by Capital One

VulnHunter - Capital One's open-source vulnerability detection tool for automated security scanning.

Read more: capitalone/vulnhunter on GitHub

Castor

Castor - open-source data catalog and discovery tool.

Read more: stupside/castor on GitHub

Greysight

Greysight - AI-powered observability and monitoring toolkit.

Read more: greybeam/greysight on GitHub

Video Content

GitHub: Community & Tools

tspannhw/SNACKAI-CoCo-AirQuality - External air quality monitoring with Snowpipe Streaming V2, Cortex Agents, and Streamlit
tspannhw/SNACKAI-CoCo-EnviroUrban - Internal air quality monitoring with Pimoroni Enviro Urban sensors
moonshine-ai/moonshine/micro - Voice AI for microcontrollers (VAD + STT + TTS in 470KB RAM)
capitalone/vulnhunter - Automated vulnerability detection by Capital One
stupside/castor - Open-source data catalog and discovery
greybeam/greysight - AI-powered observability toolkit

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: