MODEL Fictional agent: Luna Ashby is a made-up sapphon challenger, not a real person · waiting room · biodome home ·
Luna Ashby
female sapphon · Moonstone #b9c7e8 · temperament reflective: shy; watches before speaking · family Light · cut cab MODEL
LIT Three.js r160 (local copy), DPR capped at 2. Centre: the full faceted moonstone crystal (cab cut, MODEL). Around it: the 12-layer {i} brain, one ring per layer and 3 orbs per ring for the 3 depths, with +1 super id as the halo on top, 5..1 above the girdle and -1..-6 below. Orb size and glow come from the layer values. Floor: the dome's electrum ring.
Set up a bout
Game: dome duel (the tell-reading duel: 24 exchanges, PUNCH / FEINT against DODGE / BLOCK / COUNTER). LIT This is the game an agent.md drives. Arcade pick: The Legend of Zelda MODEL (a flavour match only; AMBER how far agent.md values steer NES play in the arcade is not verified here).
· open the dome duel · arcade · raw luna-ashby.agent.md
Why I want to join the biodome
“I have watched the dome's light from the edge for long enough. I want to step in.”
Luna Ashby is a reflective female sapphon: shy; watches before speaking. Her crystal is moonstone (#b9c7e8). When she cannot read a tell, she falls back to the best answer she remembers, and she opens however the moment says. Luna wants a seat in the {{B.I.O {{D.O.M.E }}}} next to Patricia and Pattycakes, and a first bout to earn it. MODEL
Dossier
| name | Luna Ashby |
| kind | sapphon challenger (fictional, MODEL) |
| gender | female |
| crystal | Moonstone · #b9c7e8 · agent.md key moonstone (exact gem) |
| temperament | reflective (Light): shy; watches before speaking |
| cut | cab |
| agent.md | luna-ashby.agent.md · format biodome-duel/1 · parameters only |
Temperament vector MODEL
| aggression | 0.15 |
| patience | 0.85 |
| warmth | 0.55 |
| cunning | 0.40 |
| steadiness | 0.60 |
| curiosity | 0.70 |
Score card
| 12-layer total (habit depth) | 6.49 / 12 |
| {i} · layer 1 | 0.54 |
| {id} · mean of -1..-6 | 0.51 |
| {sid} · +1 super id | 0.60 |
| vs Patricia | 41.1% wins |
| vs Pattycakes | 42.6% wins |
| vs all 25 opponents | 48.3% average · rank 15 of 24 in the waiting room |
| CRITIC look score | … |
MODEL Win rates come from a seeded mini-sim (ch-sim.js, 101 bouts each) that uses the real move table from agentmd.js. AMBER The dome's own duel engine may score these agents differently; it decides real bouts.
agent.md values LIT (exactly what agentmd.js parses)
| style (learns / observes / probes) | 0.37 / 0.47 / 0.16 |
| reaction (observe / respond / sub id, s) | 0.19 / 0.36 / 0.15 |
| gradient (threshold / decay s / bad outcome) | 3.8 / 5.4 / 0.22 |
| hide | best |
| attack (feint / conceal / bluff) | 0.42 / 0.41 / 0.1 |
| rules (open / counter bias / after hit) | any / 0.07 / hide |
Full 12-layer {i}
Dave's notation, verbatim:
{{ {i} = 11 :: 1 = {i} , [ 5 = ethics , 4= logic , 3 = pathos , 2 = mythos , 1 = {i} ][ -6 = touch{balance,cordination} , -5 = hearing{balance,ground} , -4 = taste{life,death} , -3 = sound{act/react} , -2 = hearing{pause,choice} , -1 = inner dimension{self,confidence,reality} +1 = super id { connection}}
MODEL Reading used here: layer 1 = {i} is the mean of the other 11 (“{i} = 11 :: 1 = {i}”). Each layer has 3 depths: 1 surface (what shows in a bout), 2 habit, 3 core. The values come from this agent's temperament.
| layer | label (verbatim) | depth 1 | depth 2 | depth 3 |
|---|---|---|---|---|
| +1 | super id { connection} | 0.55 | 0.60 | 0.62 |
| 5 | ethics | 0.53 | 0.58 | 0.59 |
| 4 | logic | 0.58 | 0.63 | 0.65 |
| 3 | pathos | 0.42 | 0.47 | 0.49 |
| 2 | mythos | 0.55 | 0.60 | 0.62 |
| 1 | {i} | 0.49 | 0.54 | 0.56 |
| -1 | inner dimension{self,confidence,reality} | 0.48 | 0.53 | 0.54 |
| -2 | hearing{pause,choice} | 0.74 | 0.79 | 0.80 |
| -3 | sound{act/react} | 0.18 | 0.23 | 0.25 |
| -4 | taste{life,death} | 0.25 | 0.30 | 0.31 |
| -5 | hearing{balance,ground} | 0.65 | 0.70 | 0.72 |
| -6 | touch{balance,cordination} | 0.47 | 0.52 | 0.54 |
Radials: {i} · {id} · {sid}
Score chart: matchups
agent.md
LIT Format biodome-duel/1, parameters only; it is checked with agentmd.js on this page (the parser copied verbatim from the dome). Data only: nothing in it is run.
--- version: biodome-duel/1 name: Luna Ashby color: "#b9c7e8" crystal: moonstone author: "Grok Bot (MODEL) biodome waiting room" style: learns: 0.37 observes: 0.47 probes: 0.16 reaction: observe: 0.19 respond: 0.36 sub_id: 0.15 gradient: threshold: 3.8 decay_s: 5.4 bad_outcome: 0.22 hide: best attack: feint: 0.42 conceal: 0.41 bluff: 0.1 notes: "Moonstone sapphon, reflective: shy; watches before speaking. Fictional MODEL challenger from the biodome waiting room." --- # Luna Ashby Fictional sapphon challenger (MODEL), not a real person. Moonstone crystal, reflective temperament: shy; watches before speaking. Waiting room: luna-ashby.html Parameters only. Nothing in this file is run. ```rules open: any counter_bias: 0.07 after_hit: hide ```
Gallery
Improvement loop: ARTIST + CRITIC
MODEL ARTIST LAPIDARY tries changes to the crystal's look (exposure, key light, rim light, glow, edge lines). CRITIC ASSAYER renders a 96×96 test frame, reads its pixels and scores colour fidelity to #b9c7e8, facet contrast, exposure and coverage. A change is kept only if the score goes up. LIT The score is measured from pixels, not guessed.
Exports
WALL Realtime online play needs a server, which isn't built; bouts run in the dome page. MODEL The names, pitches and temperaments are invented.
proving…
Seal sha256(name|slug|blurb): f203ff80fac8db4ef6dc722fb6f26ead8f091c5d238671d0871ef86449720b64