Cloud Editor @InfoQ, Microsoft Azure MVP, Crew Member @AzureLowLands, Traveller, Runner, #Seahawks, #SeaUsRise, #SeaKraken, Metal Music, Opinions are my own!
AWS replaced fat-tree data center networks with random graph theory. Now the default for most new builds globally.
https://t.co/4WtQ7rHHkz
#cloud#networking#aws
Part 7 of APIM for AI Workloads is live: Azure API Management as MCP Gateway.
Three capabilities: expose REST APIs as MCP tools, proxy external MCP servers, govern A2A traffic.
https://t.co/AQeKTJ8Z34
#azure#apimanagement#mcp#agenticai
Part 6 of APIM for AI Workloads is out: semantic caching with Azure API Management.
APIM checks vector similarity against Azure Managed Redis before touching your AI backend. Cache hit = zero tokens consumed.
https://t.co/ra42wyeUb5
#azure#apimanagement#ai
Post 2 of my Logic Apps Agent Loop series is live.
Trigger → instructions → model → tools — the four building blocks of every agent loop, with a working demo and run history screenshots.
https://t.co/jrFbbUt3EB
#azurelogicapps#agenticai#azure
A trailing slash bypassed AWS HTTP API's Lambda authorizer. Same bypass enabled an unauthenticated wire transfer. Root cause: path normalization mismatch between routing and auth layers. Same bug class hit gRPC-Go via CVE-2026-33186.
https://t.co/Y5WK3GE24b
#aws#security#cloud
Google Cloud suspended Railway's production account without notice. 8-hour outage, 3M users, all deployments down.
https://t.co/bLqJLswAef
#cloud#sre#gcp
Azure Logic Apps just got a lot more interesting.
The agent loop lets a connected LLM reason, pick tools, act, and iterate, no hardcoded steps, no fixed path.
Post 1 of 7 in my new series on Logic Apps agentic workflows:
https://t.co/JrK4snq5Ss
#AzureLogicApps#AgenticAI
Microsoft shipped Azure Linux 4.0, its first general-purpose server Linux distribution. Fedora-based, built for Azure VMs. Last of the three major hyperscalers to ship a server distro.
https://t.co/mbXoHXU0Z5
#azure#linux#cloud
Azure Logic Apps now lets agents generate and execute Python/JS/C#/PowerShell in Hyper-V sandboxes, inline within integration workflows.
https://t.co/XHsqveBQiv
#azure#ipaas#aiagents
Part 5 of my APIM for AI series: load balancing and circuit breaking.
PTU primary → PAYG overflow → auto failover. Plus the three-state circuit breaker that protects your AI backends.
https://t.co/Kuouj7b25a
#azure#ai
Part 4 of my APIM for AI series: the token metric policy.
Emit token counts to App Insights per consumer the foundation for AI cross-charging and FinOps dashboards.
https://t.co/Gn22Cz1OJ2
#azure#ai#finops
Cloudflare and Stripe built a protocol for AI agents to create accounts, buy domains, and deploy to production autonomously. $100/month default spending cap.
https://t.co/HHAuevRujO
#cloud#aiagents#cloudflare
Anthropic published a detailed postmortem on six weeks of "Claude is getting dumber" complaints. The cause wasn't the model. https://t.co/nARwASOwsD #llm#devtools#postmortem#anthropic
Commercial or open source AI gateway?
New companion post to my APIM for AI series five decision factors, a flowchart, and why regulated EU orgs almost always land on commercial.
https://t.co/SwipKkXTYc
#azure#ai#llm
AWS WorkSpaces now lets AI agents operate legacy desktop apps through computer vision. No APIs needed, no app modifications. Agent authenticates via IAM and drives the desktop like a human.
https://t.co/AGsyVL0hgw
#cloud#aws#aiagents
New on InfoQ: Google announced GKE Agent Sandbox and Hypercluster at Cloud Next '26.
gVisor isolation for secure agent execution at 300 sandboxes/sec + a million chips from one control plane.
https://t.co/dGyWGMn3wt
#kubernetes#gke#ai
Part 3 of my APIM for AI series: the token limit policy.
Enforce TPM limits per consumer at the gateway — before requests hit your AI backend. Counter-key strategies + sizing guidance included.
https://t.co/IIlRd3Q2SI
#azure#ai#llm
AI is eliminating junior developer roles faster than the industry can adapt. New post connecting Stanford, McKinsey, and the Russinovich/Hanselman CACM paper — what the productivity data doesn't show, and what needs to change. https://t.co/bIn3bsQZ7x
#ai#techtalent