Final email draft for Hacksphere

Subject: A manifesto for the agent-first open web

Hi Hacksphere,

So: a manifesto for the open web.

My observation is that an imbalance in the world lies in places where there are currently walled gardens around basically public information. LinkedIn is the obvious example. For multiple reasons, I believe 2026 is the year where these walled gardens are going to fall.

The bot / anti-bot war has ended with the bots winning around scraping information.

Second thing: there is an increased demand for agent-friendly data services.

So there needs to be a twin, for instance, to LinkedIn, which is open and acts as a buffer where you can post in and out of it, kind of like Beeper for chat, so you don't lose it. The ecosystem has matured. You can already message out with Beeper. You can already post on LinkedIn with OpenClaw. We might as well all be running on an open twin of the existing services.

A second example is all these walled-garden chat clients: Telegram, WhatsApp, iMessage, Facebook Messenger, Instagram DMs. Telegram is the least walled; WhatsApp and iMessage are more closed. This is the year where automating and bridging iMessage, Facebook Messenger, Instagram, WhatsApp, and Telegram has gotten very mature.

So it's the moment where, like it or not, there are going to be multi-chat clients again. Agent-based chat clients are already the implicit multi-chat client.

The obvious structure that comes into existence in the face of these things is a different layer, a different control surface for all of our information, and that control surface is going to have a collaborative element to it.

One more: I think Twitter is public info again. If even 1% of Twitter users have bots ingesting their Twitter feed and they're leaking that information onto the public web into a repository, maybe a decentralized one such that it can't be shut down, then that content becomes persistent. There are already many buffer apps that allow you to buffer into posting onto Twitter. The bidirectional thing is next. Across all these networks, there's going to be a different control surface for our information.

It's the obvious manifesto of the open web that folks like Tim Berners-Lee have sat on for decades. Previously the scraping infrastructure wasn't there to actually break that monopoly, and the effort of maintaining these platforms was too great to build a viable alternative without a huge amount of funding.

A limitation of Bluesky is that it does not cross-post to Twitter. It's now possible to make something that is not merely Twitter, but a superset of Twitter: an alternative browser for Twitter that augments it with external information. It doesn't require critical mass for these things to work.

Now we can build another open messaging platform that's not Telegram or WhatsApp, and that is XMPP-compatible and interoperable across different chat clients, unlike the existing walled garden ones. This is the layer of the world where everyone is about to have personal agents.

I don't mean the agents themselves. I mean the secondary layer: the agent-first world that's already implicitly there, but has not been given a control surface.

It also answers the question of how we are going to control AI from things that are not chat.

We should assume there's a bot in every conversation, often on both sides. So we might as well make a bot-friendly control surface that also makes it possible to make it explicit when there's a bot communicating.

Imagine being in a LinkedIn chat with someone and being able to say, "Yo, that person's bot is here. I'm going to ask it a question." Or imagine messaging their bot separately and saying, "Here's some info. Tell me if this person wants to meet with me." That kind of separation is powerful.

This is also why I'm excited about Open Chat. It's stupid that nothing supports interop and XMPP and agent control properly. The chat monopolies and LinkedIn-like monopolies are both vulnerable to open, agent-friendly versions. Even if I don't do anything, I think a lot of this is going to happen. But we might as well do it.

There's a broader stack here:

IdeaFlow / NoteStream is the capture layer. Open Chat is the messaging layer. World Issue Tracker is the public coordination layer. Agents connect them.

You write an idea down, decide whether it should stay private or go public, preserve the original text, and then post it into the relevant tracker. If someone posts something related to your idea, you should be able to DM them right there. The open web should not just be pages. It should be a living substrate for collaboration.

Argentina is making agents on behalf of every citizen to advocate for those citizens in government, which is so cool. That points at one of the bigger possibilities here: both open recreations of existing projects and entirely new types of projects. Things like World Issue Tracker, the issue tracker for society.

At the large network level, open infrastructure is going to win. Examples like Hermes Agent and OpenClaw show the developer momentum. When it's so fast to build software and software modification is so doable, restrictions on that level of refinement are very frustrating to the developer community. Instead of everything becoming a $100B closed company, these networks can become open source projects that accrue value through developer momentum.

Why this matters: walled gardens, and especially the abuse of news feeds on existing networks, are what I call anti-technology.

Technology should be assistive. Anti-technology restricts human flourishing.

My open web manifesto is basically: don't limit us to places where we're not allowed to build assistive interfaces, custom algorithms, and custom ways of displaying the information we're seeing. That's a human and political barrier, not a technical barrier to human flourishing.

These organizations are optimizing for profit, not human value. In a very real sense, we already have super-resourced, super-intelligent, super-agentic entities misaligned with human values. They're called corporations. We are already living in part of that dystopia.

So creating open infrastructure connects deeply to a fundamental freedom in my hacker ethic: the freedom to automate, the freedom to modify yourself.

It's a softer version of Richard Stallman's insistence on open source. Software that isn't customizable is a restriction on human creativity. Customizability used to be limited to very technical people. Now it's becoming no more difficult than sending a text message.

As Boris said, literacy wasn't taken for granted 600 years ago. In the future, developing software will be as commonplace as sending a text message: a basic literacy, with the knowledge and skill to do it widespread.

Most of the innovations we're seeing in AI right now did not require breakthroughs in models. They required breakthroughs in web scraping, bridging, automation, and breaking through silos. The tide is inexorable. If something doesn't have an API already, it's going to have an API.

So instead of fighting the tide, we should prepare for that future. And it doesn't have to be dystopian. The point is to align AI with humans instead of corporate values.

The future is open memory, open chat, open social graphs, open issue tracking, open agents, and a shared semantic/control layer underneath.

That's the download I wanted to send to this group of hackers and thinkers.

Curious what this sparks.