Cisco chips in; GitHub puts Agents in Actions
Today on Product Saturday: Cisco unveils a new networking chip for massive AI clusters, GitHub automates software-development busywork, and the quote of the week.
Today on Product Saturday: Cisco unveils a new networking chip for massive AI clusters, GitHub automates software-development busywork, and the quote of the week.

Welcome to Runtime! Today on Product Saturday: Cisco unveils a new networking chip for massive AI clusters, GitHub automates software-development busywork, and the quote of the week.
Please forward this email to a friend or colleague! If it was forwarded to you, sign up here to get Runtime each week, and if you value independent enterprise tech journalism, click the button below and become a Runtime supporter today.
Network effects: It was a rough week for Cisco after the reality of the memory-chip shortage hit home on Wall Street, but on Monday it introduced its latest high-performance networking chip for the top end of the AI infrastructure market. When it arrives in the second half of this year, the Cisco Silicon One G300 will deliver 102.4 Tbps of networking capacity, matching Broadcom's currently available Tomahawk 6.
Beyond the throughput, Cisco also added new features that help improve network utilization and get jobs done faster as the AI infrastructure market becomes more concerned about inference, as opposed to training. "What we’re starting to see now is a shift toward agentic AI workloads, and more adoption within enterprise service providers and a broader customer base," Cisco's Kevin Wolterweber told Silicon Angle.
They're with me: AI-produced coding slop is overwhelming open-source projects with half-baked or downright terrible code that has to be examined and verified just like any other contribution, and it threatens to destroy any lasting sense of community in open source. This week HashiCorp co-founder Mitchell Hashimoto released a new "community trust management system" called Vouch that could help.
"Historically, the effort required to understand a codebase, implement a change, and submit that change for review was high enough that it naturally filtered out many low quality contributions from unqualified people," Hashimoto wrote in a GitHub post introducing the concept. Obviously, AI coding tools have changed a few things of late, and Vouch allows open-source maintainers to create lists of trusted contributors that can be shared with other projects.
Flow state: While those AI coding tools take over the world, there are still a lot of tasks in the software development process that could be automated. Developers and engineering managers need to triage issues, document changes, and make sure tests are working properly, and GitHub released more details this week about a tool in GitHub Actions that could let agents do all that busywork.
GitHub Agentic Workflows allow customers to create "automated, intent-driven repository workflows that run in GitHub Actions, authored in plain Markdown and executed with coding agents," the company said in a blog post. Available for now as a technical preview, the product works with Anthropic's Claude Code, OpenAI's Codex, and Microsoft's GitHub Copilot coding agents.
Man in the box: Netbox Labs sprung out of Digital Ocean as the commercial vehicle for the open-source Netbox project, and it has raised $55.2 million (according to Crunchbase) to help network engineers automate the care and feeding of their networks. AI agents are doing more and more enterprise automation work these days, and this week the company announced that its Netbox Copilot agent is generally available.
Netbox Copilot is an "interactive AI agent for exploring, managing, and automating network and infrastructure operations – all through natural language, embedded directly in NetBox," the company said in a blog post. "I actually think the majority of the work, like the grunt work that engineers in these teams are doing these days, will start to get picked up by autonomous agents over time," co-founder and CEO Kris Beevers told Network World.
In the shadows: One surefire way to tell if an enterprise tech trend is catching on is if IT workers are implementing it on their own rather than waiting for direction from the suits, which is informally known as "shadow IT." The downside of shadow IT is that the central organization is still responsible for controlling spending and addressing security concerns, and this week Okta introduced a new tool for helping those leaders deal with "shadow AI."
Agent Discovery is a new feature in Okta's Identity Security Posture Management service that was designed to help customers "uncover hidden identity risks and misconfigurations of unknown and known agents, and map agents’ potential blast radius," the company said in a press release. If any "shadow agents" are discovered, administrators can assign them to a human manager and enforce security policies.

Speeding up code isn't enough. GitLab delivers AI across the entire software lifecycle. Learn more at GitLab.com.
During the rise of enterprise SaaS, individual departments inside companies seized control of spending on their tech tools because they were best positioned to know good software from bad and because it was so easy to get up and running. So far the AI buildout has flipped that arrangement on its head and returned power to the bosses, but according to new research from Alteryx IT leaders expect that lines of business will take over 11% of AI workflows over the next three years, and "shadow AI" might accelerate that trend.
"There are a ton of really, really, really small projects with just one or two maintainers that are also propping up a significant portion of the world's digital infrastructure, and if one of them raised a red flag and said, 'hey, I need help,' it might be that nobody listens because they aren't the second largest open-source project in the world." — Minimus's head of developer advocacy, Kat Cosgrove, pointing out that Kubernetes enjoys a lot of advantages compared to the rest of its open-source compatriots when resources are needed to make big changes.
The FTC sent letters to competitors of Microsoft this week looking for evidence of anticompetitive practices, according to Bloomberg, and all those companies probably had really nice things to say.
Salesforce sent emails informing employees who left its company kickoff event Tuesday after CEO Marc Benioff made several tone-deaf jokes about ICE that "your absence was noted" during subsequent presentations and gave them a 250-character field to explain themselves, according to Business Insider.

Speeding up code isn't enough. GitLab delivers AI across the entire software lifecycle. Learn more at GitLab.com.
Thanks for reading — see you Tuesday!