Curving the Browser: A Topology Lens for Safer AI Navigation

#4
by issdandavis - opened

Browsers already enforce trust boundaries. They just mostly present them as flat binary states.

  • secure or not
  • same-origin or cross-origin
  • allowed or denied

SCBE's browser work asks a different question:

What if the browser exposed trust as a geometry instead of a checkbox?

Why this is grounded in real web standards

The W3C Secure Contexts specification exists because sensitive features should only be available when minimum standards of authentication and confidentiality are met.

Official spec: https://www.w3.org/TR/secure-contexts/

That spec also makes a point that matters for browser mediation: secure delivery is necessary, but it is not the whole story. Embedding context and feature exposure still matter.

RFC 6454's web origin concept is another part of the picture: origins are not a fuzzy feeling. Scheme, host, and port boundaries matter.

RFC 6454: https://www.rfc-editor.org/rfc/rfc6454

What exists in the repo

The current AetherBrowser lane already contains pieces of this model:

The core idea is simple:

  • analyze a page as a topology
  • represent trust movement in bounded hyperbolic space
  • allow more than allow or deny

Why preview matters

The best part of this model is not the math vocabulary. It is the operational distinction between preview and commitment.

For risky pages, you may want:

  • metadata only
  • summary-only preview
  • constrained reading
  • blocked execution

That is different from fully entering the page and letting scripts, forms, or agent actions commit.

In SCBE terms, that is the phase-read / phase-tunnel idea:

inspect a projection of risky space without granting full action rights inside that space.

What is implemented versus proposed

Implemented in the repo:

  • topology payload generation
  • page-analysis topology summaries
  • phase-tunnel code
  • red-zone test coverage in the AetherBrowser lane

Still proposed:

  • real-world browsing comparisons
  • user studies
  • comparisons against conventional enterprise browser isolation

So the careful claim is not that the curved browser has already beaten standard browser UX.

The careful claim is that the repo now contains a concrete mediation model worth testing.

Search visibility note

I’m posting this publicly, but Google itself is explicit that making a page eligible for Search does not guarantee that it will be crawled, indexed, or served immediately.

Useful official references:

That means the right goal is search eligibility and public accessibility first, then indexing verification later.

Sign up or log in to comment