engblogs

Technical guide to using ChatGPT within Excel to enable new financial data integrations and streamline spreadsheet workflows.

Databricks for Good unifies GO Project's data into a scalable lakehouse with Unity Catalog governance, delivering near real-time KPI dashboards and AI-powered personalized newsletters to strengthen stakeholder engagement.

A concise exploration of GPT-5.4's Thinking System Card, detailing its architecture, capabilities, and potential impact on AI reasoning.

A technical overview of the GPT-5.4 release, highlighting new features, performance improvements, and deployment considerations.

A concise technical guide to deploying AI in educational settings that maximizes opportunity through scalable implementation and strong governance.

Explores why reasoning models struggle to exert full control over their chain-of-thought and why this limitation can be a net positive for safety and reliability.

Salesforce's Data 360 Compute Fabric team redesigned Kubernetes scheduling with a proactive, high-density bin-packing strategy powered by a MostAllocated placement to co-locate Spark executors, achieving 13% compute-cost savings at scale.

Open-source, fully reproducible large-scale pretraining of Olmo Hybrid on Lambda infrastructure demonstrates reliable, scalable training across 64 NVIDIA HGX B200 GPUs with a hybrid DeltaNet-transformer architecture and an open stack.

A technical synthesis of GTIG's 2025 zero-day review, highlighting the shift to enterprise and edge-device exploitation, the rise of Commercial Surveillance Vendors, sustained PRC-nexus espionage activity, and the growing role of AI in accelerating attack and defense.

Make security simpler: Introducing the Google Cloud recommended security checklist distilled from MVSP principles into a scalable, automatable, AI-ready program featuring 60 controls across six domains to harden cloud environments and accelerate risk management.

Prompts the Nano Banana image-generation suite (Gemini 3) to achieve precise, high-quality visuals, covering model capabilities, tech specs, best practices, prompting frameworks, and workflows with Veo and Lyria.

GKE gains native support for custom metrics in the Horizontal Pod Autoscaler, enabling agentless autoscaling and removing adapters and IAM complexity, while delivering lower latency and improved reliability.