DR McCullochHOMESHOPABOUT
PILEARYTIX Β· WEAVEA FAITHFUL RECONSTRUCTIONINDEX BUILT 2025-11-24

The Loom of
the Abundance

A machine that answers you in the salvaged voices of the dead & refuses to repeat you back. Rather than writing a sentence of its own, it retrieves: for every line you speak it searches a corpus of 5,782,731 short fragments of recorded film speech, scores the strangers competing to answer you and weaves the winner into the thread, and you can watch it choose.

CORPUS5,782,731dialogue fragments
EMBEDDING384-dimall-MiniLM-L6-v2
INDEXFAISSIndexFlatIP
SOURCEarchive.orgpublic-domain film Β· window 1980
VOICESsevenone loom

Scroll down, speak a line and watch the dead compete for it.

I

The Loom

One line goes in and seven voices reach for it.

Who answers

The engine draws the fragments closest to your line then re-ranks them, and you can watch the scores. Toggle the era to see the anti-echo rule change which stranger wins.

SCORING ERA

The shortlist rises

The loom is idle, waiting for a line to draw candidates from the corpus.

The woven thread

Each reply is a real stranger’s line, retrieved and toned.

The thread is bare, so speak a line to begin weaving.

II

Why it will not echo you

The anti-echo rule, from RESPONSE_APPROPRIATENESS_FIX.md

An earlier version of the engine had a flaw that was almost a confession. Ask it “What happened here?” and it would find, as the closest match in 384-dimensional space, the fragment “What’s happened here?” and return it. The nearest voice to your voice is your own voice, very slightly delayed. A mirror like that answers nothing.

The fix adds a scoring component (response-appropriateness) that punishes the echo. High word-overlap scores down to 0.1, while a statement that actually answers a question scores up to 0.9. So the machine can no longer parrot you. It has to reach past the mirror for a stranger who really answers, and that refusal to hand you back your own words is the whole design. Flip the era in the loom above and you can watch the mirror win, then watch the rule strike it down.

YOU What happened here?
OLD SYSTEM “What's happened here?”echo Β· 0.1
FIXED SYSTEM “someone would say, bailed out. And there’s a question of”woven Β· answers
YOU I heard a funny story
OLD SYSTEM “a similar echo”echo Β· 0.1
FIXED SYSTEM “what happened. Nobody told you…”woven Β· answers
YOU What secrets lie hidden?
OLD SYSTEM “all secrets would automatically be exposed”echo Β· 0.1
FIXED SYSTEM “Oh! Well, I never kept any secrets…”woven Β· answers

The appropriateness rule

conditionwhat it detectsscore
Exact matchexact echo of your line0.1
High word overlapmore than 70% shared words0.1
Medium word overlap50–70% shared words0.3
High semantic similarityvector proximity above 0.750.1–0.4
Question β†’ statementa statement answering a question0.9
High new-word ratiomore than 60% new words0.8
Defaulta moderate, ordinary response0.6
LEGACY Β· pre-fix
final = (0.5Β·semantic + 0.3Β·flow + 0.2Β·novelty) Γ— (0.5 + 0.5Β·character)

The mirror could win here, because semantic proximity dominated and nothing forbade the echo.

ANTI-ECHO Β· current
final = 0.30Β·semantic + 0.25Β·flow + 0.15Β·novelty + 0.20Β·character + 0.15Β·appropriateness

The appropriateness term breaks the mirror. These are the five weights config.py declares and the loom above runs them exactly. The docs also name a sixth relevance term (0.10) not carried by the shipped config; it is left out here.

III

The seven voices

Character definitions, verbatim from character_manager.py

No character writes a word. Each is only a set of tone-weights, a bias in how the corpus is read. The same prompt pulls a different stranger through each of these seven filters, so the whole of a character’s personality here comes down to which tones it leans toward.

The Detective

A noir detective character who speaks in dramatic, mysterious tones with formal undertones.

analyticalcynicalobservantworld-weary
mysterious0.4
dramatic0.3
serious0.2
formal0.1

The Comedian

A witty, light-hearted character who prefers humorous and playful dialogue.

wittyenergeticoptimisticentertaining
humorous0.5
playful0.3
casual0.2

The Scholar

An intellectual character who speaks formally and seriously.

intelligentthoughtfulpreciseknowledgeable
formal0.6
serious0.4

The Wanderer

A casual, philosophical character who speaks in thoughtful but relaxed tones.

contemplativeeasygoingcuriousintrospective
casual0.4
mysterious0.3
playful0.3

The Hero

A dramatic, serious character who speaks with conviction and intensity.

bravedeterminedpassionateintense
dramatic0.6
serious0.4

The Mystic

An enigmatic character who speaks in mysterious and cryptic ways.

enigmaticspiritualwiseotherworldly
mysterious0.5
formal0.3
dramatic0.2

The Friend

A warm, approachable character who speaks casually and humorously.

friendlysupportiverelatablewarm
casual0.5
playful0.3
humorous0.2
IV

The specimen archive

The salvaged seam every woven line is drawn from

These are real lines, harvested verbatim from the subtitle and speech tracks of public-domain films. The full corpus is multilingual and thick with ASR noise and civic-meeting filler; this is a hand-curated English seam of the eerie ones, and every line the loom weaves is drawn from here. Filter by the tone the engine detects in each.

linefragmentdetected tone

Colophon & honest reconstruction

What this is. A faithful client-side reconstruction of the pilearytix-weave dialogue engine, part of my Pilearytix of the Abundance, a Celtic cyber-shamanism narrative RPG. The seven characters, their tone-weights, the tone lexicon, the tone-compatibility matrix, the scoring weights of both eras and the anti-echo rule are reproduced verbatim from the source (character_manager.py Β· tone_analyser.py Β· config.py Β· RESPONSE_APPROPRIATENESS_FIX.md).

What is honest about the numbers. Tone detection, character compatibility, flow, novelty and appropriateness run the real rules, in your browser. The semantic stage of the real engine is a FAISS search over 384-dimensional MiniLM embeddings of 5.78 million lines, far too large to ship. Here it is approximated by an honest lexical cosine over the curated seam. The drama is true even where the vector space is smaller. No live backend is faked, and no film titles or timestamps are invented because the flat corpus carries none, so each fragment is shown only with its detected tone and its index in the shipped seam.

Underneath the framing, the original project is a prosaic and beautiful thing: an Internet Archive subtitle harvester that rips dialogue tracks off old films, timestamps them and embeds them for semantic search. The Celtic cyber-shamanism sits on top of that, and underneath it the uncanny part is plainly true: a machine answers you in the found voices of strangers pulled from the archive, and it will not say your own words back to you.

DR McCulloch Β· the weave