TLDR IT 2026-09-24

Google's Antigravity SDK now supports local models through LiteRT and compatible inference servers, enabling agents to run offline while keeping code on-device. Hybrid workflows can combine cloud-based planning with local execution to reduce API usage and limit sensitive data transfers.

Meta unveiled VR glasses weighing about 100 grams, with computing and battery hardware housed in a tethered puck, launching in Spring 2027 for $1,299.99. The device offers a private multi-screen workspace, virtual keyboard and touchpad, and photorealistic avatar calling through apps including Zoom and WhatsApp.

Meta has announced plans to bring its Muse personal AI agent to its glasses, alongside new Ray-Ban Meta Audio frames starting at $349 and shipping October 13. The expanded lineup includes Ray-Ban Meta Gen 3 with improved battery life and controls, extending Meta's push toward hands-free AI assistance.

Undocumented configurations and manual workarounds become operational risks when AI agents start managing infrastructure. Reliable automation requires infrastructure and policies defined in code, built-in cost and access controls, and clear boundaries for actions that need human approval.

Fragmented security tools force analysts and agents to rebuild context and coordinate actions across separate systems, slowing incident response. Microsoft's proposed answer is an integrated security operations center, now in Defender preview, that combines SIEM, threat protection, and response controls around shared context.

WhiteFiber Continuum runs DriveNets AI Fabric and WEKA NeuralMesh across 83km of dark fiber, hitting 136 Tbps of bandwidth and 0.9ms round-trip latency, so two data centers behave like one GPU cluster with both sites carrying live traffic at once, no failover delay, no idle capacity.

Read the architecture breakdown

Barracuda launched AI Data Security to stop sensitive company data from reaching ChatGPT, Copilot, Gemini, and 1,300+ other AI tools. It adds prompt-injection screening, policy controls, and audit logs for SMBs and MSPs, available in October with SecureEdge Premium Access. A Barracuda survey found 49% of CIOs at 500 to 1,000 employee firms lack AI governance skills.

Databricks made its Genie One MCP server generally available, giving any AI agent governed access to structured and unstructured data plus business context through Genie Ontology. The MCP runs as a managed service in Unity Gateway, adding centralized governance, fine-grained policies, and audit logging while keeping answers consistent across tools like Claude, ChatGPT, and Cursor. It aims to reduce agent sprawl by letting agents query data, retrieve results, and explore visualizations and citations without custom connectors.

Microsoft plans to invest more than $10 billion through 2030 across Kuwait, Qatar, Saudi Arabia, and the UAE to expand cloud and AI infrastructure. Its commitments also include more than $400 million for network connectivity, alongside sovereign cloud capabilities and business continuity programs.

Dropbox says scaling AI requires redesigning end-to-end workflows, aligning incentives, and measuring outcomes like revenue, quality, and retention. It reports ~70% AI-generated code, top-5% PR throughput versus similar peers, and quality in line with industry, while using internal coding agents like Nova to connect AI spend to engineering value. Leaders argue that as AI lowers engineering costs, human problem-solving, judgment, and leadership matter more, and trusted organizational context will separate effective adopters.

NVIDIA Topograph is an open source toolkit that discovers cluster network topology from cloud or on-premises providers and normalizes it into a common model. It publishes that topology as Kubernetes node labels, Slurm configurations, Slinky ConfigMaps, or NFD resources so schedulers like KAI Scheduler, Kueue, and native Kubernetes can place tightly coupled GPU workloads near each other. By keeping topology current as clusters change, Topograph reduces network congestion and improves AI factory efficiency, tokens per watt, and cost across Kubernetes, Slurm, and Slinky environments.

Q2 2026 server sales surged 52% year over year to $166.32 billion.

Microsoft's refreshed 12-inch Surface Pro and 13-inch Surface Laptop bring Snapdragon X2 Plus processors, local AI capabilities, and enterprise management tools.