Google AI Updates: October 13, 2026

1. TimesFM 3.0 Brings Multivariate Forecasting to BigQuery in Preview

Google. Google released TimesFM 3.0, a 330M-parameter time-series foundation model from Google Research, in preview in BigQuery. Unlike TimesFM 2.0 and 2.5, which were univariate only, version 3.0 forecasts multiple target columns jointly and accepts both past-only and future-known covariates through new TARGET_COLS, PAST_COVARIATE_COLS, and FUTURE_COVARIATE_COLS arguments to AI.FORECAST. Existing AI.FORECAST, AI.DETECT_ANOMALIES, and AI.EVALUATE workflows can switch by setting model => 'TimesFM 3.0', and Google says the model ranks first on the GIFT-Eval, FEV-Bench, and TIME leaderboards. Source

2. Data Agent Kit Connects Coding Agents to Google Cloud Data Services

Google. Google Cloud introduced the Data Agent Kit, which packages MCP tools, Markdown skill files, and IDE tools for working with BigQuery, Cloud Storage, Spanner, Managed Service for Apache Spark, and Managed Service for Apache Airflow from a coding agent. It ships as an extension for VS Code-based IDEs such as Antigravity and Cursor, and as a plugin for agents including Antigravity CLI, Claude Code, and Codex, with the plugin source on GitHub at GoogleCloudPlatform/data-agent-kit-plugin. Google’s walkthrough builds an ingest, dbt, training, and batch scoring pipeline from prompts and converts a YAML pipeline definition into a scheduled Airflow DAG. Source

3. Looker Data Apps Let Agents Generate Governed Web Apps in Preview

Google. Google Cloud launched Looker Data Apps in preview, letting an AI agent generate interactive React applications from a plain-language description and host them natively inside Looker. Apps are grounded in the LookML semantic layer and inherit existing row-level security and access controls, can write back to records and push actions to tools such as Salesforce and Jira, and expose generated SQL for auditing. Developers can also build apps with Claude Code or Antigravity CLI, and Google notes LLM tokens are consumed only at design time, with end-user interactions running on native Looker APIs. Source

4. Google Summarizes Health AI Research Results, Including NHS Mammography Studies

Google. Google published a roundup of its disease research, citing two Nature Cancer studies with Imperial College London and the NHS covering 125,000 women in which its mammography system caught 25% of historically missed interval cancers and could cut specialist reading workload by up to 40%. The post also describes AlphaProtein Novo, a pipeline for designing new-to-nature enzymes detailed in a bioRxiv preprint, and a flu forecasting model that the CDC ranked first among 39 systems for predicting 2025 to 2026 hospital admissions. Google.org has provided over $60 million in 2026 to groups applying advanced models to disease research. Source