---
title: "Protocol Productivity"
publishedAt: "2026-05-05"
summary: ""
---

I am a big believer in the unified inbox thesis. Not a startup's idea of a unified inbox, where they've scraped together a bunch of apps into a single interface I didn't ask for. That just does not feel right. I mean something more fundamental. The protocol layer.

Protocols already exist for most of this. That is the part I keep coming back to. Email has SMTP and IMAP. Calendar has CalDAV. Contacts have CardDAV. News has RSS. Text messaging has RCS. Social has ActivityPub and ATproto. The infrastructure is there. The platforms just choose not to use it. Most of the closed platforms you use every day could run on one of these. They decided not to. That is not a technical constraint, it is a business decision.

I do not think everything needs to be open source. But I do think everything should have an open standard. Choosing not to use one is still a protocol decision, it just happens to be the wrong one for everyone except the platform. You feel it every time you have to manually copy something out of a messaging app and your clipboard fills with garbage, because there is no other way to get the content out.

The most effective way to build software is via building blocks that enable and encourage others to build. Protocols are what make that possible. When everything speaks a common language, every tool becomes a module. You can swap it out, build on top of it, hand it to an agent without a custom integration. Open ecosystems will outperform closed ones. The easier it is to compose and build on top of something, the more leverage everyone gains.

Protocols are also what make agents actually useful in your daily life. But there is another problem I think about more. Even where agents do have access, I still cannot point at something directly. I have to describe it. That is not really an agent working with your data, that is just a search box with more steps.

I wrote about this before in the context of [URLs](https://www.nklswbr.com/blog/urls): the hyperlink is the actual system of record. If a specific state exists in your app, it should probably have its own URL. That is true for people navigating the web. For agents it becomes a hard requirement. Access is one thing, addressability is another. Most platforms give you neither.

The thing I am most hopeful about is clients. Not new platforms, clients built on the infrastructure that already exists. The more good ones there are, the less defensible the closed alternatives become. And maybe along the way someone figures out what a new protocol looks like if it were designed from the start for this. We still need to figure out what a protocol looks like when agents are as much a participant as you are.