AI agent behavior is non-deterministic by definition, so a single prompt change can degrade response quality even when standard metrics show no errors, and without full trace visibility, diagnosing why an agent took an unexpected path becomes guesswork. AWS's new CloudWatch Omni captures every step in a structured timeline with built-in evaluations for correctness, coherence and tool selection, aiming to give teams a unified view regardless of how the agent was built. Read more about how AWS is approaching agent observability: buff.ly/guZrSxm #AWS #Observability #AIAgents

Microsoft warns Anthropic's "conscious" AI threatens control. OpenAI logs 6 misalignment cases. Dreamforce: AIforce, Koa, $2B Listen Labs bid. IBM Anderon starts quantum wafer runs. Guests @jswartz Jaime Hampton Jiewen Wang. #AISafety #Dreamforce nitter.net/i/broadcasts/1pJdRDlBL…

Interesting read on why shift left is dead - but is it really? That was a big buzzword in the DevOps/DevSecOps world

AI ambition is easy. Turning it into measurable value is where the real work begins. Join @RedHat’s virtual event, 𝘛𝘩𝘦 𝘈𝘐 𝘍𝘢𝘤𝘵𝘰𝘳𝘺: 𝘍𝘳𝘰𝘮 𝘝𝘪𝘴𝘪𝘰𝘯 𝘵𝘰 𝘝𝘢𝘭𝘶𝘦, for a practical look at how open architecture, accelerated computing and a strong partner ecosystem can support enterprise AI in production. The agenda brings together perspectives from @RedHat, @nvidia, @Dell and @TechstrongGroup on: 🔹 The current state of enterprise AI 🔹 Building an open foundation while keeping technology choices flexible 🔹 The AI Factory in action with Red Hat, NVIDIA and Dell Technologies 🔹 AI sovereignty and maintaining control 🔹 Partner solutions designed for real-world AI workloads 📅 October 15, 2026 🕦 11:30 a.m.–2:30 p.m. SGT 💻 Virtual event See how organizations can connect AI strategy with the infrastructure, platforms and ecosystem needed to deliver measurable outcomes. Register here: bit.ly/3TlaoqX #EnterpriseAI #AIInfrastructure #RedHat #NVIDIA

Blitzy launched a free sandbox that lets DevOps teams reverse-engineer up to 1M lines of code and use AI agents to identify and remediate vulnerabilities. Read more: buff.ly/efWJqFw #Blitzy #DevOps #DevSecOps #AICoding #SoftwareSecurity

AI agents can act with user authority and access sensitive data. Can you prove who authorized what? Explore agent identity, least privilege and audit in this hands-on AWS workshop. 📅 Oct 1 | 3 PM ET 🎟️ Register: buff.ly/jXXK0pr #AISecurity #AWS #CloudSecurity

3,200 vulnerabilities fixed. 98% stayed fixed. Zero human security review. See how StackHawk used AI coding agents to find, fix and verify vulnerabilities in production. 📅 October 1 | 1 PM ET 🎟️ Register now: buff.ly/RcKPrS1 #AppSec #AISecurity

Agents that install packages on their own need a gate before the first tool call. creationcode makes you approve a readable plan and set the maximum build cost before the run starts. A human accepts the verified PR before merge. The final receipt closes the case.

Enterprise AI agents do not have a model problem. They have a data access problem. Mike Vizard talks with Tim Bond, Chief Product Officer at Adeptia, about data products, governance and why agents need systems of action. Watch: techstrong.tv/videos/ai-lead… #AI

AI cost management isn’t just Cloud FinOps 2.0. The economics are different. Explore AI cost attribution, governance and waste with Harness’ Patrick Brogan. 📅 October 1 | 9 AM ET 🎟️ Register now: buff.ly/EHwMlVv #AICostManagement #FinOps #AI #CloudFinOps

agents installing packages unsupervised is how you inherit yesterday's CVEs at 3am. I pin a allowlist and fail the run on anything else.

Talentica Software this week launched DevX AI Pods, a managed software delivery service that uses AI agents trained to supply the context needed to deploy applications built with AI coding tools at scale. The agents evaluate code against a CCCR framework while more than 600 Talentica engineers validate the output, addressing a gap where AI tools generate more code without actually increasing the rate at which applications ship. Read the full article on the service and the survey data behind the code volume problem: buff.ly/exiItJq #AI #DevOps #SoftwareDevelopment

difference is blast radius. skip-permissions trashes a sandbox, auto-merge ships to everyone. only works if the agent cant touch the gates: ci config, test files, codeowners all locked. otherwise it just learns to make checks green

Behavior drift is the agent security risk most teams won’t see coming. Alan Shimel talks with David Mytton, Founder and CEO of Arcjet, about securing AI agents at runtime. Watch: techstrong.tv/videos/intervi… #AISecurity

the compute access part is the one people forget to plan for until they're already stuck waiting on it

Most security teams can tell you how many people work in engineering, but far fewer can tell you how many credentials those people and their installed apps hold against the company's source code. GitHub Enterprise Cloud owners can now export a full inventory of SSH keys, PATs, OAuth tokens and GitHub App tokens with last-used dates, and coding agents are only going to make that count harder to keep up with. Find out what the export covers and why the first one will turn up credentials nobody knew were still active: buff.ly/RjedSLm #GitHub #DevSecOps #Security #SupplyChainSecurity

An OAuth token taken from a former CrowdSec employee's laptop still had permission to read the company's private repositories, and by the time GitHub access was revoked three days later, roughly 170 private repos had already been dumped. The theft went undetected until the source code surfaced on a dark web marketplace in September, and the CEO's own conclusion is that the company was very close to clean. Read the full article on the TeamPCP supply chain attack and the offboarding gap it exposed: buff.ly/ZMmFx7e #Cybersecurity #SupplyChainSecurity #CrowdSec #TeamPCP

DevOps Experience 2026 starts in 45 minutes. The race for agentic AI is on, and today’s conversation goes beyond the hype. Across software delivery, security, and operations, builders, leaders, and practitioners will tackle the questions that matter most. What delivers real value? Where should people stay in control? And what does it take to build AI-powered systems teams can trust? Expect practical insight, real-world lessons, and perspectives that challenge assumptions about what comes next for DevOps and agentic AI. #DevOps #AgenticAI #DevSecOps #PlatformEngineering

DevOps Experience 2026 starts tomorrow. If you haven’t registered yet, this is your last chance to join a full day focused on the questions defining DevOps in the agentic AI era. Across 12+ sessions, practitioners and technology leaders will explore agentic AI, platform engineering, DevSecOps, incident response, software delivery, security, and the infrastructure required to put AI to work responsibly. 📅 Tomorrow, September 24 ⏰ 8:45 AM–3 PM ET 💻 Free virtual event 🎥 On-demand recordings available to registered attendees within 24 hours Registration closes tonight, and your join link will be delivered by email. There’s still time to register: bit.ly/4ySwKA0 #DevOps #AgenticAI #DevSecOps #PlatformEngineering

The AI stack many organizations operate today is being assembled one model, framework, and orchestration layer at a time. Each decision can introduce dependencies that become harder to unwind later. At DevOps Experience 2026, @mitchellashley, VP & Practice Lead, Software Lifecycle Engineering at @TheFuturumGroup, will examine how teams can evaluate whether their AI stack actually supports their goals or whether those goals are starting to bend around the technology already in place. He’ll explore where lock-in can hide, how speed can come at the expense of observability and control, and how to distinguish an AI stack that compounds value from one that compounds cost and risk. 📅 September 24, 2026 ⏰ 2:30 PM ET 💻 Free virtual event Register today: bit.ly/4ySwKA0 #AIStack #DevOps #AgenticAI #SoftwareEngineering