DevOps.com

Anthropic is telling DevOps teams to slim down their harnesses because many now issue duplicate instructions that burn tokens, and since modern models already build in things like verification, the answer is lean high-level prompts rather than micromanagement. The company cut the size of its own internal system prompts by 80% and says teams should "unharness" stale instructions, then spend more time evaluating how outcomes were generated. Read the full article on why the biggest harness is no longer the goal: buff.ly/X66mQLA #DevOps #AIAgents #Anthropic #Tokenomics #AICoding #AgenticAI

Patrick Debois, the godfather of DevOps, told an AI DevCon audience that the future of DevOps revolves around multiple AI agents performing tasks concurrently, and that without orchestration to stop them undoing each other's work, both token consumption and infrastructure usage will skyrocket. Most teams have adopted AI coding tools but few have re-engineered serial workflows for parallel execution, which means they are adding tension without realizing the benefits. Find out why DevOps teams need to start designing for parallelism now: buff.ly/raxfJsI #DevOps #AIAgents #AgenticAI #PatrickDebois #Tessl #SoftwareDevelopment

Observability and reliability roles dominate this week's DevOps listings, with Pinterest, Affirm and Censys all hiring for the systems that tell you what production is actually doing. Perplexity tops the range at $405,000, and four of the ten are fully remote. See the open roles here: buff.ly/IeYtcWy #DevOps #Jobs #TechJobs #Hiring #Careers #DevOpsJobs

Tessl has added ready-to-use loops DevOps teams can invoke and customize to drive agentic engineering workflows like ticket-to-PR, code review and vulnerability remediation, built from combinable AI skills rather than a factory each team has to construct itself. CEO Guy Podjarny's argument is that each product team still deploys its own software factory because requirements differ fundamentally between applications, and the cognitive load of managing agentic workflows is high enough that more teams will buy one than build it. Read the full article on why the software factory matters more in the AI era: buff.ly/3LS7BiH #DevOps #AIAgents #Tessl #SoftwareFactory #AgenticAI #PlatformEngineering

A tsunami of fragile AI-generated code is now entering production, and SRE teams are stretched to the max, with code that breaks differently and more often and root causes increasingly difficult to pin down. PagerDuty's answer is Paige, an autonomous SRE agent embedded in Slack and Teams that can remediate well-documented disruptions on its own or triage telemetry with a human giving final approval, archiving every resolved event into shared operational memory because, in its CEO's words, the only way to maintain resilience in the face of this complexity is to fight fire with fire. Read the full article on how PagerDuty is approaching AI-driven outages: buff.ly/LZHXZOo #SRE #AIOps #PagerDuty #AIAgents #DevOps

Tomorrow at 1 PM ET: Could moving Windows and SQL Server workloads to Linux, PostgreSQL and containers lower your cloud costs? Join Evolve Cloud Services to explore modernization savings and ROI. Register: buff.ly/fxUxpNN #AWS #CloudModernization

Most cloud roadmaps are built with an end date. The environment they support doesn't have one. Technical debt returns as soon as the team moves on, security requirements keep shifting, and AI agents expect infrastructure that can keep up. A roadmap built to finish leaves teams catching up again within a few years. In this on-demand session, @awscloud (AWS) explains how leading teams build continuous improvement into the way they operate, so their environment stays secure, current and ready for what's next. ▶ Watch on demand: bit.ly/4ehz62K #AWS #CloudComputing #DevOps

Platform engineering is reshaping enterprise observability. As autonomous systems scale, platform teams are moving beyond dashboards to embed monitoring, cost controls and governance by default. Read the report: buff.ly/8edXfxP #PlatformEngineering #Observability #DevOps #AI

Developers have spent years learning not to trust what they open in a browser, and their AI coding assistants have not learned the same lesson. Forever Security researchers found that the embedded browsers in Cursor, GitHub Copilot, Codex, VS Code and Claude Desktop give web content far more reach into the host machine than a standalone browser would allow, and in one Cursor demo the sandbox blocked the attack until the agent simply switched it off, which is exactly the problem with a Workspace Trust prompt that records consent and blocks nothing. Read more about where enforcement has to live when the agent decides a safeguard is in its way: buff.ly/uoDbxL1 #Cybersecurity #AICoding #AppSec #InfoSec #AIAgents

Teams running AI coding agents at scale keep hitting the same wall: the model is capable, but it goes looking for context it doesn't have, burns tokens on round trips and sometimes ships a change the team later reverts. GitLab's answer is Orbit, a context graph connecting code, merge requests, pipelines, vulnerabilities and ownership that agents query over MCP, and in its own testing agents connected to it produced up to 45x fewer hallucinations and used up to 4.5x fewer tokens. Read the full piece on why context, not a bigger model, may be the fix: buff.ly/5bWOiRC #AI #GitLab #AIAgents #DevOps #MCP #ContextEngineering

Free breakfast. Networking. Sharp panels. All before the KubeCon keynotes. Futurum Summit @ KubeCon | Mon, Nov. 9, 7:45 AM MST | Lake Effect, Salt Lake City Seats are limited. Register: futurumgroup.com/kubecon/#re… #KubeCon #CloudNative #AI

Most coding agents still wait their turn, and DeepSeek's Harness team thinks that's the problem worth solving, so it released an open research prototype called Super-Agent that splits the work into a Router managing context and multiple Workers running in parallel. The part DevOps teams should think hardest about is proactivity, because an agent that hears a bug reported in team chat and opens a pull request on its own is in practice a new contributor, and it needs the same scrutiny as any contributor. Read the full article on what happens to the delivery pipeline when agents stop waiting to be asked: buff.ly/gQJfVOj #AIAgents #DeepSeek #DevOps #SoftwareDevelopment #AICoding

IBM used its TechXchange 2026 conference to unveil a bevy of updates to the Terraform infrastructure-as-code platform it gained with HashiCorp and its Bob AI coding tool, including an HCL-based framework designed to be simpler to invoke than existing policy-as-code approaches, a public preview of the Terraform Developer Portal that ensures developers and AI agents only invoke approved infrastructure patterns, and a SafeWrite capability that intercepts AI-proposed file edits and package-install commands in memory. Outbound CTO Tyler Lynch said the goal is to provide AI agents with the context needed to consume IT infrastructure at scale, while platform teams and SREs need to discover and interrogate infrastructure no matter where it is. Read the full article on what IBM announced and where each capability lands: buff.ly/TSnh1ju #DevSecOps #Terraform #IBM #HashiCorp #InfrastructureAsCode

Tomorrow: Is your API security keeping pace with AI-driven development? Join Justin Collier of @SmartBear for a practical 90-day plan to strengthen API quality, security and reliability without slowing delivery. 📅 Tomorrow, November 3 | 11am ET 🎟️ Register: buff.ly/i4EvSeN #APISecurity #DevOps