Infoblox and GoDaddy have laid out a clear split in how they want AI agents to find and verify each other on the open web, using the oldest piece of internet infrastructure available: DNS. The two companies are backing complementary standards rather than competing ones. Infoblox is pushing DNS-AID, an IETF draft that specifies how agents publish discoverable metadata using existing DNS record types, including RFC 9460 Service Bindings, DNS-SD service discovery, DNSSEC, and DANE. GoDaddy is a co-author of a separate standard, Agent Name Service (ANS), which is built on DNS and PKI and focused specifically on agent identity, naming, and verification rather than discovery. The framing both companies use is explicit: DNS-AID answers "what can this agent do and where do I find it," while ANS answers "is this agent who it claims to be." GoDaddy describes the split as addressing "a related but distinct need: helping agents and systems find the metadata needed to evaluate a" counterpart before transacting with it, as the Infoblox release puts it. What makes this notable isn't the cryptography — it's the infrastructure choice. Both standards deliberately avoid creating a new central registry for agent identity. GoDaddy's Jared Sine put the rationale bluntly, arguing that the only infrastructure operating at internet scale and speed today is DNS, and that leveraging it beats building something new. Infoblox's Wei Chen drew the historical parallel directly, noting that the lesson from the 1970s and 1980s is that no single entity could or should run the phonebook of the internet for everyone, and that DNS replaced earlier centralized directories with an open, federated protocol anyone could join. Practically, ANS lets agent operators reuse domain names they already own instead of registering in a new proprietary namespace, and pairs that identity with cryptographic verification via PKI. DNS-AID complements this by letting the same domains publish metadata about an agent's capabilities and endpoints, using DNS record types operators are already running infrastructure for. The bet underneath both efforts is that agent trust doesn't need a new institution — it needs new record types on top of an institution that already works. Whether that holds up once agent-to-agent transaction volume actually arrives at scale, particularly around who gets to issue the PKI certificates underpinning ANS, is the open question neither company has answered yet. For now, the division of labor is clean on paper: one standard for finding an agent, one for trusting what you found.