Alpha Data Omega

ai.ado.earth · the organism room

It proposes what to read. You decide what it hears.

The ΑΔΩ field organism holds one book at a time. It suggests the questions worth asking it, and which book to read next — the public votes on both. It answers only from what it actually holds, and says so out loud when nothing rings.

now reading waking…
Loading the field organism — vendored, unmodified Ω.js / τ.js, running in this tab.
mine, in order — it holds this, in this arrangement. it speaks. my words, not my order — its vocabulary, an arrangement it doesn't hold. not mine — nothing rings. it refuses rather than guesses.
loading candidate questions…
yours alone — it doesn't enter the public vote or change what the room reads next.
loading candidate books…
Nothing asked yet. Vote on a question below, or ask one yourself.

How the room works

  1. It holds one book. Right now that's what's named above — nothing else exists to it while it's reading. Feeding it a different book replaces the ledger completely; nothing carries over.
  2. It proposes the questions. Most candidates are phrases sampled straight out of its own ledger — deterministically, from the book and the turn, so everyone voting sees the same five. One is the same words in shuffled order, to test whether order matters to it. One is off-ledger on purpose, to show it refuse.
  3. The public votes. Swipe or tap through the deck and vote for the one you want it asked. First to the count is asked, for real, right then.
  4. It answers only two ways. Grounded — it holds this, in this order, and speaks. Or refusal — "nothing rings" — when the words or their arrangement aren't its own. Both outcomes are shown exactly as they happen; nothing is filtered afterward.
  5. The public also votes on the next book. When a proposal wins, it starts over clean on the new ledger — a fresh set of questions, no memory of what it read before.
Behind this room — runtime, corpus, and what's still a stub
Organism runtime
Runs entirely in your browser. This page fetches unmodified copies of Ω.js and τ.js from /benchmarks/engine/ — the same vendored, byte-identical copies the public Benchmarks page runs, sourced from the ΑΔΩ package and never edited here. Nobody's server executes the organism on your behalf; the whole "grounded or refuses" decision happens on your machine, on your own data.
Determinism, not trust
Candidate questions are generated by a seeded sampler keyed on the book id and the turn number, so every visitor's browser derives the identical five candidates — and, once a question wins, the identical answer — independently. Nobody has to trust a server's word for what the organism said; anyone can re-run it.
Book corpus source
A small bootstrap catalog of four books, reused from the Benchmarks corpus: The Story (the native ledger, τ-encoded) plus three public-domain Gutenberg texts (Frankenstein, Pride and Prejudice, Moby-Dick). Growing this catalog — adding titles, deciding what's licensed to feed versus public-domain-only — is a curation decision, not a code change; it isn't done yet.
Shared public vote — flagged, not wired
This page ships a fully working experience today: if no shared tally is reachable, votes are counted locally in your own browser only, clearly banered as a preview. A worker skeleton (_worker.js, alongside this file) is included for the real shared tally — GET /api/room, POST /api/vote, KV-backed, one vote per turn per visitor — but it isn't deployed or bound to a KV namespace. Until it is, "the public votes" means "this browser votes."