# user-controlled meta-apps over the web
Layers are independent apps that run on top of existing websites, with their own domains, permissions, identity, data, UI, communities, and economic models.
A layer can transform, annotate, filter, augment, audit, explain, automate, or socially recontextualize the page underneath — without the host site's permission or participation.
Websites monopolize interpretation. The page you are served is the only lens you are given; annotation, filtering, audit, automation, and alternative communities all require the site's cooperation, which is exactly backwards. And from the owner's side: a static site that needs one form is forced to become a backend operation overnight.
The browser is a substrate, not a destination. Users should be able to install interpretation the way they install software — scoped, permissioned, revocable, and economically independent of the page underneath.
the canonical journey
The motivating user is not exotic. It's everyone who ever published an HTML page and then needed the most basic feature on the web.
A static site: files on a CDN. Cheap, fast, cacheable, secure by omission — the best deployment story computing has ever produced.
The most basic feature: a sign-up form. A contact form. A file upload. Table stakes since 1995.
One <form> tag summons a backend, a database, spam filtering, email deliverability, auth, sessions, a storage bucket, a privacy policy, a build pipeline, and a bill. The feature took an hour; the apparatus takes the rest of your life.
A forms SaaS, a serverless function, a full-stack framework — each rents you a backend in a different shape. The page still ends up welded to infrastructure it never wanted.
The form was never the page's job. Sign-up, contact, and uploads are layers: independent apps with their own identity, data, permissions, and lifecycle, composed onto the static page instead of welded behind it. The site stays files; the capability is installed, scoped, and revocable.
Coupled with a personal data server that takes AYNIL — all you need is log — seriously: sign-ups and submissions land as events appended to a user-owned oplog, uploads as blobs in a content-addressed store. The site never grows a backend; it gains edges into ledgers. The store sketch already exists as POSEVEN, under desh.
mechanics
A stable addressing scheme over page structure, so a layer can bind to meaning ("the price column", "the comment thread") rather than brittle selectors.
A declaration of which origins a layer touches, what it reads, what it injects, and what leaves the machine — reviewable before install, enforceable after.
Grants are per-layer, per-origin, user-owned. A layer's community identity is its own, not borrowed from the host site.
Layers carry their own monetization and governance. The host page is the terrain, not the landlord.
verbs
Re-render hostile interfaces into usable ones.
Attach durable, shareable margin notes to any page region.
Watch what a page does and keep the receipts.
Drive repetitive flows under user-owned rules.
Bring your community's view of a page with you.
artifacts
the load-bearing hard part; binding to meaning across page changes
scopes, capabilities, and disclosure format
a captured-reference track literally named for this investigation — permissionless web-layer composition: identity, composition, subscriptions, cross-site interaction — with Solid, federation, and local-first/CRDT sources filed alongside, from the research run that seeded Rhett. The founding question already had a reading list
one layer, one site, full grant lifecycle
edges
The lab's founding question: who gets to interpret the web? An old idea-backlog entry called it "metaverse layers to extend web pages beyond the URL model" — the name was already Layers. The canonical journey that motivates it: I had a static website, and now I need the most basic feature — a sign-up form, a contact form, file uploads. Then the nightmare begins.