snowflake

Introduces Spider 2.0-AIFunc, a benchmark of 465 instances across 125 real-world databases for evaluating AI-native SQL queries that use cloud platform AI functions. Evaluates ten state-of-the-art models, finding proprietary models reach 67-70% accuracy while open-source models lag behind.

Discusses the evolution from text-to-SQL to autonomous data agents, comparing custom-built agents using LangGraph with managed platforms like Snowflake Cortex Analyst, Databricks Genie, and PowerBI Copilot.

Snowflake AI Research releases Arctic RL, an open-source unified RL backend that integrates with VeRL and SkyRL, enabling up to 6x actor-update acceleration and 3.5x end-to-end training speedup. It includes recipes for text-to-SQL and multi-hop QA, achieving competitive accuracy on enterprise benchmarks.

A Twitter thread highlights key takeaways from a Latent.Space podcast episode with Databricks co-founders, covering why Databricks beat Snowflake, the rise of metaharners, Neon's success, HTAP via LTAP, MosaicML's fate, and maintaining startup culture in a large company.

This paper proposes MERIT, a dynamic multi-horizon memory retrieval framework for interactive text-to-SQL agents that uses episode-level and turn-level memory with learned retrieval policies optimized via reinforcement learning and a process reward model for dense rewards. Experiments on BIRD-Interact and Spider2-Snow show that MERIT outperforms static and single-horizon dynamic baselines in success rate while requiring fewer interaction turns.

Snowflake now supports job-based batch inference powered by Ray, enabling distributed GPU execution for scaling model inference over millions of unstructured datapoints with a single API call.

Beau Rothrock used Devin, an AI coding agent, to migrate 14,000 dashboards from Redshift to Snowflake in five weeks, overcoming a delayed project.

An analysis of three new Postgres-compatible cloud databases—Snowflake Postgres, Databricks Lakebase, and Azure HorizonDB—highlighting their distinct architectures and the vendor lock-in implications for enterprise data platforms.

Snowflake and OpenAI announce a $200 million partnership to integrate OpenAI's frontier models directly into Snowflake's data platform, enabling enterprises to build AI agents and generate insights from their data using natural language without coding.