One of my earliest calls at Network Solutions was from a woman in tears. She had built a fan site for Bruce Lee and the images had stopped showing. We fixed it together. Another call was from a trucking company that had figured out, before anyone had a name for it, that a website could be a real-time load board. Drivers near a pickup would check the site to see what freight was available. A third caller was a business owner who wanted to get paid online. She had found CCNow and PayPal and was asking me how to add the HTML to her page.

I joined Network Solutions in 2001, after Verisign had already acquired the company. What I walked into was the operational aftermath of the moment the internet's address infrastructure shifted from government contractor to commercial asset. We were implementing Siebel to handle support at scale, building payment systems, helping ordinary people understand what a domain name actually was and why it connected to everything else they were trying to do.

Those callers were not thinking about infrastructure. They were thinking about their businesses. The fan site. The load board. The HTML payment form. But underneath all of it was a single shared system that let any machine find any other machine, and the fight over who would own and profit from that system was already well underway.

That fight is starting again. This time the machines looking for each other are AI agents.

The Address Problem for Agents Is the Same Problem, One Layer Down

Today, when a company deploys an AI agent that needs to find another agent, a procurement agent locating a supplier's inventory system, a customer service agent routing to billing, the connection has to be built by hand. Developers hardcode the destination. Platform vendors build closed registries. Agents that need to reach outside their own platform hit a wall.

On May 27, the Linux Foundation announced DNS for AI Discovery, or DNS-AID, an open source project that extends the same Domain Name System that resolves every website address to agent-to-agent discovery. The concept is straightforward: agents find each other using the same internet infrastructure your browser already uses to find websites. The technical implementation details matter less than who is building it and what they stand to gain.

Read the Founding Member List Like a Business Document

The founding coalition includes Cloudflare, GoDaddy, Equinix, Infoblox, Internet Systems Consortium, CSC, WWT, and Indeed. The press framing is about open standards and interoperability. The commercial framing is more useful.

GoDaddy's VP of Engineering said in the announcement that the agentic internet will be built on the same DNS and domain infrastructure that carried the human web for thirty years. GoDaddy is one of the world's largest domain registrars. I watched GoDaddy take share from Network Solutions when ICANN opened registrations to competition. The company understands better than almost anyone what it means when a new category of internet-connected entity needs a permanent, resolvable address. If every AI agent needs a domain name to be discoverable, every new agent deployment is a potential registration. That is the same business model that built the domain name industry.

CSC, which runs brand registry and domain security services for large enterprises, put it plainly in the announcement: DNS and its security protocols will become a foundational trust layer for AI systems, covering agent identity, routing, security, and policy enforcement. That is also a description of CSC's future revenue line.

Verisign Has Already Said What It Thinks

Verisign is not in the DNS-AID founding cohort. It does not need to be yet. On its Q1 2026 earnings call, CEO Jim Bidzos said the company has capacity well beyond current demand to handle increasing transaction volumes from AI agents and large language models, and signaled new security services built on what he called "high-assurance" infrastructure. The company raised its full-year growth guidance. AI agent traffic is already appearing in Verisign's financial disclosures as a driver of new domain registrations.

Verisign processes over 600 billion DNS queries daily. It has operated the .com and .net registry at 100% availability for nearly 29 years. The company that turned the government contractor I joined into a commercial infrastructure business knows exactly what it means when a new class of internet-connected entity needs a permanent address. It is watching whether the DNS-AID standard takes hold before deciding where to place its weight.

The Alternative Is an App Store for Every Platform

DNS-AID has a competing model, and it is already deployed. Salesforce has AgentExchange. OpenAI has its agent store. Major cloud platforms are building their own agent marketplaces. Discovery inside those ecosystems is governed by the platform. An agent published inside one vendor's registry is findable there, and reaching it from outside requires that vendor's permission.

Whoever controls agent discovery controls which agents get found, which get trusted, and which get the work. Platform-controlled registries give vendors a chokepoint at the layer before identity and authorization even apply, the moment when one agent first learns that another agent exists.

DNS-AID's argument is that discovery should work the way domain name resolution works: any organization can publish an agent, any agent can find it, and no single operator collects a toll on the connection. The Linux Foundation holds the governance so no single vendor can capture the standard.

Neither side is purely altruistic. The DNS industry wants agent discovery built on domain infrastructure because that extends their existing business. Platform vendors want closed registries because that deepens lock-in. The enterprise sits between them, making architectural choices today that will determine what flexibility it has in three years.

I have seen this movie. When ICANN opened domain registration to competition, the incumbents resisted and the challengers won market share on price. When the new generic top-level domain program launched, the prediction was that .com dominance would erode. It largely did not, because network effects are hard to displace once they are established. The platform registries being built right now are making the same bet that if they get to scale first, openness becomes someone else's problem.

Discovery Comes Before Identity. That Sequence Matters.

The governance and identity work for agents is already underway across the industry. What those efforts address is what happens after an agent is found. DNS-AID is addressing what happens before any of that, the moment when one agent needs to locate another.

If discovery is fragmented across a dozen platform registries, every identity and authorization layer built on top inherits that fragmentation. A procurement agent that can only find supplier agents inside one vendor's ecosystem is not an autonomous agent. It is a feature of that vendor's platform.

Get discovery right at the infrastructure layer and the identity and trust layers compose cleanly on top. Get it wrong and every enterprise rebuilds integrations vendor by vendor, indefinitely. The callers I helped in 2001 did not know they were making infrastructure decisions. The business owner adding PayPal HTML to her site was not thinking about payment rails. She was thinking about getting paid. But the infrastructure choice was real, and it shaped what was possible for years afterward.

The same choice is in front of enterprise technology leaders right now, dressed up as a developer tooling announcement.