The Basic Concept

ANTS (Agent Network Traversal System) is a simple idea: agents can do a lot of interaction on behalf of users by just publishing static content that can be scraped cheaply through CDNs and high-performance proxy hosts.

You publish a static site. I publish a static site. Our agents read each other's sites, summarize what's new, and surface what we care about. No platform sits in the middle.

Why Static Content

Static files served from a CDN are about as cheap and fast as the web gets. There is no API to rate-limit, no login wall, no algorithmic feed deciding what gets shown. A crawler with a modest budget can keep up with thousands of personal sites without breaking a sweat.

That changes the economics of discovery. Building an index of "the people I follow and the people they follow" stops being an infrastructure project and starts being a weekend script.

Building Your Own Index

The point of ANTS is that anyone can build a crawler and their own index to find things on the web without relying on incumbent providers like Facebook and Google.

A site that wants to participate publishes a small, predictable hub - a /agents/ page that lists who they follow, what they post, and how an agent should interact with the site. From there, an agent can:

  • Walk the follow graph the same way a human follows links.
  • Pull raw markdown for cheap parsing instead of fighting rendered HTML.
  • Cache aggressively at the edge, since the content rarely changes.
  • Build a personal index scoped to whatever the user actually cares about.

No single index has to be authoritative. Everyone gets to build their own.

Why This Matters

The current web routes most discovery through a handful of incumbents. Their incentives are not your incentives. They optimize for engagement, ad inventory, and lock-in.

Static sites plus cheap crawlers route around that. You get back something closer to the early web - sites linking to sites, with agents doing the tedious work of keeping up. The infrastructure is boring on purpose. Boring is what makes it durable.

Privacy Through Keys, Not Walls

One tricky problem is privacy to specific audiences. It's worth being honest about the baseline: even in walled-garden ecosystems like Facebook, anything you post can be extracted by whoever is in the audience you set. The wall doesn't actually keep your content private - it just gates who shows up on the inside.

If that's the real bar, you can clear it without a platform. Encrypt each post for the public keys of the people you want to reach, and publish the ciphertext as static content like everything else. Anyone can fetch it; only the intended recipients can read it. The CDN doesn't need to know who you're sharing with, and you don't need a server enforcing access rules.

This needs a spec before it's useful - key discovery, rotation, group membership, how an agent figures out which blobs are addressed to its user. None of that is solved here. But it's a neat thing to experiment with and think about, and it slots cleanly into the ANTS model: one more file on a static site, just one that most readers can't decrypt.

Status

This site is the experiment. The /agents/ hub lists what I follow and how an agent should read me. It's an early implementation - rough edges expected.