Azure Copilot Observability Agent is now generally available, and Microsoft is explicitly framing cloud ops as “agentic observability”: agents correlate logs, metrics, traces, topology, and operational context instead of only surfacing alerts. Sources: Microsoft blog | Brendan Burns on LinkedIn
Microsoft also shipped a fresh capacity signal for the agent stack today: Azure HDv2, HXv2, and ND MI455X v7 are being expanded for “new agent-driven workloads” and production-scale inference. Source: Microsoft expands Azure AI and HPC infrastructure with AMD
GitHub moved Copilot spend visibility forward again on Jul 20: users can now see AI credits used per billing cycle, and admins can manage AI credit pools for cost centers directly in the billing UI. Sources: per-billing-cycle usage | AI credit pools
GitHub Copilot app gained /security-review in public preview, bringing the CLI security scan into the desktop app; Visual Studio’s June update also added MCP-server trust validation and GA for the C++ modernization agent. Sources: Copilot app security review | Visual Studio June update
Microsoft Research’s Adoption and Impact of Command-Line AI Coding Agents is the clearest fresh field study this pass: Microsoft’s early-2026 rollout of Claude Code and GitHub Copilot CLI found adoption spread through social networks and adopters merged roughly 24% more PRs; the HN thread is active and worth skimming. Sources: arXiv | HN thread
microsoft/skills-for-fabric is a traction-backed repo to watch: the repo shows 834 stars, 230 forks, and latest release v0.3.8 on Jul 16, making it a real Microsoft skill bundle for Copilot CLI / Fabric workflows rather than a throwaway demo. Source: GitHub repo