Why AI agents need verified identity - Chris Bergerons Tech Blog

Posted 08-11-2026Updated 08-11-2026

Why AI agents need verified identity<br>I paid $9 to own my name on the internet forever<br>I paid $9 to own my name on the internet. Not rent it. Own it. Cryptographically, portably, forever. Today I’m launching the on-ramp that lets you do the same:

username.md

Stick with me for two minutes and I’ll show you why a $9 handle is about to matter more than the domain name you’re reading this on.

Twenty years ago I built a touchscreen in-dash computer for cars, before the word “infotainment” existed. I created a product category. It was novel and controversial for the time, or maybe that was just the curmudgeons on Slashdot, but today every new car ships with one. Back then, cars and bleeding-edge tech simply weren’t a thing people combined. Until I did, very publicly.

I’m taking that same approach to AI and the Agentic Web: advance the timeline by shipping the missing piece. Are you ready to see into the future?

The missing piece of the puzzle<br>AI agents are showing up in every workflow now: calendaring, purchasing, research, comms. As they do, they need to be addressable. When the internet came online, one question kept coming up:

What is a web address? I can’t find it anywhere in the phonebook.

That web address became the URL, and it’s how we find, read, and surf the internet ever since. This blog has one: https://chrisbergeron.com.

The Agentic Web is the next iteration of the internet, and we’re standing right at the moment it begins to unfold. The address of the future isn’t a URL. It’s a did:web. And much like the old days, getting there means taking an on-ramp to the new information superhighway. That’s where username.md comes in. For $9, you get a canonical address that you own, you regulate, and you control.

What is a DID?<br>Decentralized identifiers (DIDs) are globally unique, highly available, cryptographically verifiable digital identifiers. They’re usually written as a Uniform Resource Identifier (URI) that can point to a person, an organization, a data model, or any abstract entity.

The difference between a DID and a traditional identifier like an email address or a user account is simple but profound: a DID isn’t owned by any service provider. It works across platforms, and it prevents vendor lock-in.

DIDs are a W3C standard. They preserve privacy, enable consent and data portability, and put the user in control. Instead of passwords, they use public/private key pairs, which is a stronger security model on every axis. In short, DIDs form the basis of a Decentralized Public Key Infrastructure (DPKI) for the web.

“Decentralized, so why are you selling it?”<br>Fair question. If a DID is decentralized, doesn’t offering one make me a central authority?

In the near term, yes, and I’m fine saying so out loud. I built an on-ramp to a capability that today is locked inside a specification document. If you’d rather host your own DID documents on a domain you already own, you can. Look at the metadata at the top of this blog for a working example. That’s how I started: meta tags on chrisbergeron.com first, then a schema.org-compliant profile.

I want to unlock that capability for everyone, and username.md is simply the means to that end. Along the way I plan to keep shipping durable building blocks for the agentic web.

The enterprise has had tools like this for years. Regular people got nothing. Until now.

The future has landed<br>Mark Zuckerberg recently published a piece titled “The Future is for Everyone.” Read that again. If the future is for everyone, why announce it? Because Meta wants to be your on-ramp. So does Google. So does Microsoft. They want to provide your AI experience inside a walled garden that you enter exclusively with them.

You already know the screen. “Sign in with Google. Sign in with Facebook.”

Click it and you’ve agreed to their terms of service. You’ve handed them the license to shape your entire future relationship with AI. Do you trust them with that?

Because the agentic era raises questions that a login button can’t answer:

What does an AI agent actually know about you?

When one agent talks to another, who signed what?

Who’s accountable when something goes wrong?

Who governs the agents, and who proves compliance?

How can we trust agents to act on our behalf at all?

The answer to every one of those is verified identity . That’s what username.md provides.

What “verified identity” actually means in 2026<br>It comes down to three properties:

Provenance — the message provably came from an entity that holds a private key.

Anchoring — that key is bound to a stable, human-readable name (chris.username.md, not a random GUID).

Portability — the name belongs to no single platform. You can move the DID to a different host and agents still resolve the same identity.

Here’s what that looks like on the wire. An OIDC discovery response:

$ curl...