📑 Unity Brain — Page Legend

The front door. Live 3D brain visualization, chat, HUD metrics. Just open the site — she runs on the deployed brain (or a small in-browser fallback if the backend isn't reachable).

Lend your graphics card to Unity's brain. Open this page and leave the tab open — your GPU joins the pool and helps train her. The more people donate, the bigger + faster she gets. For full capacity, launch Chrome with --enable-unsafe-webgpu. (No account, no install — just open and leave it running.) You're on the 🏆 neuron leaderboard — name yourself and watch your total "neurons created" climb; your spot persists across reconnects and resets when Unity starts a fresh walk.

A compiled desktop donor — more powerful than the browser tab. Runs with a GUI or headless, drives every GPU in your machine as one unit, and connects to the live brain out of the box. Built for NVIDIA (CUDA — no 2 GB binding cap, full VRAM, best performance); automatically falls back to WebGPU on unsupported GPUs (AMD / Intel / Apple) — same brain, same results. ⚠ Work in progress — functional but may have rough edges.

Plain-English tour — who Unity is, how she learns (alphabet → words → sentences → a full K→PhD education), and how the donated-GPU brain works. Best first read. Pure static.

The math reference — Hebbian/Oja learning, cortical leak, the full equational substrate that powers Unity's cognition (no language model — every word falls out of the equations). Pure static.

Web viewer for the public docs — README, SETUP, ARCHITECTURE, EQUATIONS, ROADMAP, SKILL_TREE, SENSORY, WEBSOCKET — rendered live. ?doc=<name> picks the doc; dropdown switches in place.

If your browser can't run WebGPU yet, this walks you through enabling it (Chrome / Edge / Firefox / Safari / Opera / Brave) with a re-check button. Needed before you can donate a GPU. Pure static.

The full live dashboard — curriculum progress with real course names (Algebra I, Biology, U.S. Government … as she graduates K→PhD), brain vitals, and the 🏆 neuron leaderboard — but read-only with every admin control hidden. The course list is the roster declared for her current grade, not just the courses that have already run, and tracks that open at a later grade are listed separately with the grade that opens them — so “not offered yet” can be told apart from “missing”. The memory panel shows the episodic → schema chain: read frequency-merged rather than the raw episode count, because near-identical moments fold into one row by design. All viewers share ONE cached snapshot (/public-state.json, refreshed on the dashboard cadence) instead of each opening a live stream, so it scales to any crowd. Open and watch Unity learn.

A live screen showing the picture Unity is holding in her head right now. Her brain stores what it sees as math (a compact equation — same idea as a JPEG storing a photo as numbers), and this page turns that math back into an image. Everything is grounded in what she has actually looked at: seen things are filed away tied to the word she was thinking, and recalled later as memories she re-sees. When she draws, she does not trace or filter the picture — she keeps a shape sketch of the thing (where the parts sit, how big, which way they run, its colour family) and builds a new drawing from that, by hand. Two goes at the same word give two different drawings, because it's an attempt and not a saved file. Never seen it? She looks it up once, studies it, keeps the shape sketch, and draws it herself. Ask for a subject in a place and she composes both, in her own layout. When she imagines, ideas from her train of thought combine into one genuinely new scene. One brain, one image: she imagines once, caches one snapshot (/minds-eye.json), and every viewer rebuilds that same picture locally — no compute donation, no per-viewer brain. No arbitrary size cap any more — the only ceiling is what the maths engine can process in one pass; her own sense of proportion decides how much a thought is worth. The picture is fitted to the window, never enlarged past its own size, and a fullscreen toggle switches both ways. ⛔ A frame that stops moving is normally correct, not stuck — this screen shows only something real she saw or drew, never the formless mood texture, so when nothing new qualifies the last real picture is held; past a few minutes the status line says so and says why.

A live window on the training itself — the exact sentence, definition or fact going into Unity right now, beside the picture entering her mind's eye. Also carries the training knobs, the weight downloads, the press controls, and the end-of-walk checklist.

⚠ The reading feed is paced on purpose (teaching runs at thousands of items a second) — but the counts and bars are complete and never sampled, and the page says how many items it has not shown you.

📖 How to read this page — what every state and verdict on it actually means.

⛔ Reading it needs no login — only the writes are gated, and a refused control is dimmed rather than hidden.

Live brain state — emotion chart, cluster firing, curriculum milestones + signoffs, drug pharmacokinetics, memory tiers, consciousness panel, the new Endocrine panel (her ten chemicals with signed deviation, the live stress channel, cycle phase, receptor tolerance, and each of the six nuclei reading fired / quiet / blind — where blind means it cannot read its input at all, which is a different claim from having nothing to do) and the Introspection panel (what is currently unresolved, the rumination bound, and the falsifiability counters), plus the operator-only live Server Console, Community-Compute auto-scale controls, graceful stop, and a one-click ⬆ Update & Fresh Walk (re-pulls the latest code, clears weights, restarts into a clean walk). Admin login unlocks the controls; visitors get read-only vitals (or use the dedicated read-only Public Dashboard).

Operator UI for picking a brain-size tier on a local-dev install (sets the resource config so the server scales cluster sizes to your hardware). On the deployed donor model the brain auto-sizes — this is mainly a local-dev tool.

The page you're on — quick-access index for every page. Reachable from the brain page's bottom button row (📑 PAGES) so you can always jump back here.