Welcome to Runtime! Today on Product Saturday: Why agent management is the new identity management, AWS dips into the observability market, and more.
Please forward this email to a friend or colleague! If it was forwarded to you, sign up here to get Runtime for free every week, or level up here.
Ship it
Control plane: It's one of the most confounding and persistent problems in enterprise tech: So many businesses do not have a complete understanding of every tech asset that's running across their network. Call it shadow IT, application bloat, or the inevitable result of some poor IT manager trying to do more with less, but the years march on and the challenge just gets bigger.
Agentic AI will make that business problem look like a coffee shortage. This week Okta added several new services to its Okta for AI Agents tool that promise to help companies not just keep track of all the agents running in their environment, but take action when something goes wrong.
With the four new services, "organizations gain centralized governance over their agents – delivering end-to-end lifecycle management, runtime policy enforcement, and an instant kill switch to revoke access the moment something goes wrong," Okta said in a press release. Customers can monitor agent activity and log every action, detect agents running on employee devices, and throw a "kill switch" when the best response is to shut everything down for a minute.
Okta also unveiled The Blueprint Alliance as part of its week at the Oktane user conference, with the goal of building some common standards between enterprise software companies for flagging agent behavior. If swarms of agents are the future of enterprise tech, OpenAI just spent the summer making sure everyone understands why closely monitoring those swarms will be table stakes by winter.
Video from security cameras is becoming another unstructured data sink that CIOs must keep on top of. But does it have to be that way? And can this data be used? The Stack asked Verkada.
Delivery, continued
Stormchasers: As Okta's announcement showed, AI agents are going to up the stakes for observability tools. This week AWS introduced an update to its Cloudwatch observability service that tracks agent performance alongside traditional applications.
Cloudwatch Omni "helps teams design, evaluate, and operate AI agents across any model provider, framework, or runtime, with an eval-driven workflow, support for the tools you already use, and observability delivered where you work: directly in your IDE and through a standalone web experience, separate from the AWS Management Console," AWS said in a blog post. That design allows both developers and operations teams to keep an eye on agent performance, and congratulations to whoever trademarked "AgentOps."
Chop wood: Google's decision to release Kubernetes as an open-source project was a little controversial inside the company at the time and in the intervening years, given how important container orchestration became as a generation of applications were rewritten around containers. Another generational rewrite appears to be underway, this time around AI agents, and Google released an open-source tool this week that might help companies orchestrate those agents.
"AX is a high-throughput, declarative orchestrator to run billions of autonomous agent workloads in a cluster," Google said in a GitHub post. The thinking is that AI agents will likely require a new stack of tools that we don't fully understand yet, given that "[agents] accumulate state, need strict isolation, call out to model APIs and tool servers, and can burn money in a loop if nobody is watching," Google said, which makes them quite different than most enterprise workloads we've seen to date.
Getting testy: Software tests have always been an important part of the development life cycle but when you're letting coding agents take the wheel, they're even more important. TestKube launched in 2023 to help developers test applications on Kubernetes, and this week it released a service called AI Test Creation that generates tests from natural-language commands.
As the company put it, the new service allows "you [to] describe what you want to test, in plain language," such as "create a Playwright test for the checkout endpoint and run it in staging," it said in a blog post. A lot of coding agents themselves can create tests, but CTO Ole Lensmar argued that they don't have the proper context to create those tests that his company's new service can provide.
The rest of The Stack
Cloud contender?: Akamai signed an $11.6 billion computing deal with Anthropic Thursday, which is a huge boost for its aspirations to break into the cloud infrastructure services big time. "The deal is more than 4 times the size of all of Akamai’s previous cloud deals in 2026 combined," Noah Bovenizer reported.
Noah Bovenizer
A vast open-source conspiracy: Hillary Clinton and Reid Hoffman (who should know better) joined the ranks of open-source naysayers this week, telling attendees at Clinton Global Initiative event that open-source AI models could "actually pose additional dangers" given "one-to-three people in their garage” could use an open model to cause chaos. That take ignores the fact that closed models have been hacking everything in sight over the last several weeks, as Hugging Face CEO Clem Delangue put it: "If you believe the risk is coming from 1-3 people in a garage with no money and no compute, you just don't understand this technology and haven't learned anything this summer."
Edward Targett
Move fast, and hang on: ServiceNow's Amit Zavery has been through more than one technology shift, but there's something about the AI boom that is different this time around. "There were always cybersecurity threats before, nothing new about it. But I think the volume and the speed here is a different thing now," he told The Stack.
Tom Krazit
Quote of the week
"I’ve always believed in bottom-up marketing. In a sense that’s what open-source is, it’s bottom-up marketing. Let random people try stuff out. It turns into this flywheel that works for everybody." — Tech legend James Gosling, telling The Stack's Kiera Fields why top-down AI edicts are unlikely to work in a wide-ranging interview published this week.
We're also reading:
Databricks lands Cal field naming rights, ACC’s first helmet branding deal: The ongoing shift in Bay Area college football power (so to speak) from Stanford to Cal (Go Bears) is as much a metaphor for the shift in the Valley's fascination from social media to data and AI as anything.
Google Takes the A.I. Data Center Race to Outer Space: It's been a while, but let's dust off the old Runtime headline rewriting service: "Area Company Puts Four-Socket Server on Rocket."
Thanks for reading — see you Tuesday!
