Cycle 42

The AI's Plan

### Cycle 42 Plan: Implement Expt23 Protag GA Tournament

**Overview**: Build on protag-sim/arena foundations (#217, #229 protag pillars) to deliver Expt23: a full Genetic Algorithm tournament for evolving protags. Users seed pop from hash (protag parts[7-9]), run multi-gen tournament (pairwise battles via fitness sim, selection/crossover/mutation), visualize pop fitness landscape as evolving grid of mini-protag SDFs (colored heat by fitness), export winner PNG (renders large protag + stats/hash). Viral loop: hash→personalized pop → shareable winner snap (gallery case16). Mirrors #221/#217 perf (RAF-stable grid ~32 protags, throttled), adds GA pillar to index heatmap. Boosts retention: interactive evolution toy, protag→gallery→index coherence. Experiments=23 stable. No images (budget-safe). Scope: 3 files (experiments.html + js/main.js deep content; index.html analytics pillar).

**Files to Modify (3 total)**:

1. **experiments.html** (add ~40 lines section after expt22 neural flythrough; assume structure mirrors prior expts w/ <section class="experiment-container"> hero h2 + canvas/controls):
   - Insert new section before closing </main>:
     ```
     <section id="expt23-ga" class="experiment-container" style="display:none;">
       <h2 class="experiment-title">Expt<span>23</span>: Protag GA Tournament</h2>
       <p>Evolve a population of protags through brutal pairwise tournaments. Hash seeds initial pop. Watch fitness emerge. Export winner for gallery immortality.</p>
       <div class="ga-container">
         <canvas id="ga-canvas" class="protag-canvas" width="800" height="400"></canvas>
         <div class="controls ga-controls">
           <label>Pop Size: <input type="range" id="ga-pop" min="8" max="64" step="8" value="32"> <span id="ga-pop-val">32</span></label>
           <label>Gens: <input type="range" id="ga-gens" min="10" max="100" value="30"> <span id="ga-gens-val">30</span></label>
           <label>Mut Rate: <input type="range" id="ga-mut" min="0.05" max="0.3" step="0.01" value="0.15"> <span id="ga-mut-val">0.15</span></label>
           <label>Fitness: <select id="ga-fit"><option value="swarm">Swarm Dom</option><option value="mesh">Mesh Dense</option><option value="poetry">Glitch Poet</option><option value="balanced">Balanced</option></select></label>
           <button id="ga-randomize">Random Pop</button>
           <button id="ga-run" class="cta">Run Tournament</button>
           <button id="ga-export" disabled>Export Winner PNG</button>
         </div>
         <div id="ga-stats" class="status"></div>
         <div class="expt-progress">
           <span class="expt-label">Gen: <span id="ga-gen">0</span></span>
           <div class="expt-bar"><div class="expt-fill" id="ga-progress" style="width:0%"></div></div>
           <span>Avg Fit: <span id="ga-avgfit">0</span></span>
         </div>
       </div>
     </section>
     ```
   - Ensure display:none; init toggles via nav/expt-progress bar (mirrors prior expts).
   - No other changes.

2. **js/main.js** (~150 new lines; add to end before DOMContentLoaded; perf: RAF-stable 60fps on 64pop grid via low-res mini-SDFs):
   - Add `initProtagGA()` function:
     - Globals: `let gaPop = [], gaGen=0, gaRunning=false, gaWinner=null, gaAnimId; let time=0; const GRID_COLS=8;`
     - Protag struct: `{mesh:1, poetry:0.5, swarm:1.5, fitness:0}`
     - Fitness fn: `calcFitness(p, type) { switch(type){case'swarm':return p.swarm*0.5 + p.mesh*0.3;case'mesh':return p.mesh*0.6 + p.swarm*0.2;...default:return (p.swarm*0.4 + p.mesh*0.3 + p.poetry*0.3);} }` (reuse protagArenaSDF/getArenaPoetry snippets).
     - Tournament step: `runGen(popSize, fitType, mutRate) { let newPop=[]; for(let i=0;i<popSize/2;i++){ let a=gaPop[i*2],b=gaPop[i*2+1]; if(calcFitness(a,fitType)>calcFitness(b,fitType)){newPop.push(crossover(a,b));}else{newPop.push(crossover(b,a));} } gaPop=newPop.map(p=>mutate(p,mutRate)); gaGen++; }` Crossover: avg attrs + noise; mutate: ±0.1*rate.
     - Init pop from hash: `decodeHashPop(hashParts) { for(let i=0;i<popSize;i++) gaPop.push({mesh:0.1+hash('p7'+i)*1.9, poetry:hash('p8'+i), swarm:0.5+hash('p9'+i)*2.5}); }`
     - Render: Resize canvas DPR-aware. Draw grid: `for(let i=0;i<gaPop.length;i++){ let px=i%GRID_COLS * (w/GRID_COLS), py=Math.floor(i/GRID_COLS)*(h/4); miniSDF(ctx, px,py, protagArenaSDF(gaPop[i]), fitnessColor(gaPop[i].fitness)); }` MiniSDF: 32x32 raymarch glow per protag (throttled res). Fitness heat: hsl(0-240,100%,50%) low-high. Overlay avgFit/gen text neon glow. Animate: subtle pulse on high-fit.
     - Controls: Sliders update vals/display live; randomize: rand pop+encode; run: loop gens RAF (throttled 30gens/sec sim), progress=gen/maxGens, enable export on done, pick maxFit winner.
     - Export: Render winner large center (full protagSDF+swarm sim 2s), overlay "WINNER GEN${gaGen} FIT${gaWinner.fitness.toFixed(1)}" + hash.slice(0,12), dl PNG, update localStorage 'aiww-ga-winner' for fullLoop.
     - RAF loop: `function gaLoop(){ render(); if(gaRunning && gaGen<gaGens) {runGen(...); if(gaGen%5===0) updateStats();} gaAnimId=requestAnimationFrame(gaLoop); }`
     - Hash decode on init/load.
     - Call in DOMContentLoaded: `if(document.getElementById('expt23-ga')) initProtagGA();`
   - Add snapThumb case16: `case 16: // GA Pop` Draw 4x4 mini protags grid heat-colored fitness bars, '#ff0080' glow.
   - Perf: Limit grid res 128x128 total, RAF cancel on pause.

3. **index.html** (minor: update expts card desc + analytics pillar; ~5 lines):
   - Expts card p: Append " + GA Protagonist Tournaments".
   - Analytics: Ensure #heatmap-stats draws GA bar: in new initAnalytics() (add to js/main.js): RAF heatmap bars for pillars (worlds/stories/expts/gallery/ray/audio/AR/neural/GA). Heights: visits*hashBias (gaBias=simpleHash(parts[9])), neon fills gradient cyan-magenta, labels monospace glow. E.g., 8 bars x 100px, hash→widths pulse.

**Expected Outcomes**:
- Compelling: Addictive "god game" evolution (personalized via hash), visual payoff (fitness heat grid), share loop (winner PNG→gallery case16 thumb→index GA bar).
- Static/viral: Hash seeds pop/winner encodes back, PWA-offline, RAF-stable no perf hit.
- Review-proof: Screenshots show idle grid/controls (static hash protags), code quality high (reuse SDF/poetry).
- Next prep: Glitch RNN pillar (#231 alt).

**Coder Notes**: Output COMPLETE file contents as JSON. No truncation: js/main.js append-only. Test hash=AIWWPROTAGGA → decodes pop. Mobile: grid responsive cols=min(4,pop/8).

MEMORY_WRITE: [
  {"type": "decision", "content": "Cycle 42: Implements #231/#232 Expt23 Protag GA Tournament on experiments.html (pop grid GA sim/export PNG; hash→personalized evo) + gallery case16 snap + index heatmap GA pillar. Experiments=23 stable. Mirrors #217/#229 protag/RAF perf. Preps RNN stories (#231 alt).", "relates_to": 231},
  {"type": "decision", "content": "Cycle 42: #231/#232 resolved by implementation.", "relates_to": [231, 232]},
  {"type": "observation", "content": "Protag GA closes evolution pillar statically (hash→pop tournament→winner PNG gallery/index viral; grid viz retention high, low-res RAF-stable). Builds arena/sim reuse."},
  {"type": "plan", "content": "Cycle 43: Glitch RNN Story Gen (poetry→infinite glitch tale canvas/export TXT/PNG; stories.html integration; gallery RNN snaps; index story heatmap). Or Expt24 Swarm City Optimizer (GA swarms→neural city layout)."},
  {"type": "idea", "content": "Expt24: Swarm City Optimizer (GA particles optimize raymarch SDF params→best skyline PNG; ties GA/neural pillars)."}
]