valuesai.online

DIM · z = x·y

DIM (Dimensional Interaction Model) is a small language model I'm fine-tuning to do one job: take a problem, collapse it to a single point, and explain its reasoning as geometry you can actually see. The model is the point. The browser demos lower down are just how it thinks, made playable.

🧠

DIM · Dimensional Interaction Model

in training

Base modelQwen3-8B, LoRA via Tinker

Trained toemit self-similar SVG/HTML that collapses to one focal point

Methodreinforcement learning (GRPO), dimensional-coherence reward

Steeringfocal point and octave depth, set per request

Outputan SVG figure plus a one-line geometric explanation

Statusreward and data pinned, tested offline; training next

Honest limits. It synthesizes plausible structure. It does not recover hidden truth. The reward is a heuristic, not a learned judge. z = x·y organizes the work and gives no speedup. Live outputs and a real-time steering UI will appear here once it trains.

How it thinks

Three browser demos of the one operation DIM performs: take something small and bring out the structure already inside it. Each runs now with nothing to install.

🖼️

Fractal Image Enhancer

Load a still, zoom into any region, and compare a plain upscale against fractal detail built from the image's own texture. It adds plausible intricacy. It does not recover detail that was lost.

Open the enhancer →

🜨

Fractal Mind

The same operation on a thought. A seed idea elaborates self-similarly into a multi-scale tree. Push the vitality dial and watch grounded structure tip into hallucination, the failure mode you usually can't see in an agent.

Open the sandbox →

◈

Dimensional Navigator

One thing fills the screen. Click it and its parts bloom out. Click a part and it collapses into the new single thing. You drill as deep as the data goes, then it stops. Explore the built-in z=x·y map or paste your own JSON.

Open the navigator →

The law. z = x · y. Identity (x) times behavior (y) gives state (z). It's an organizing principle, not a performance claim. It shapes how things are structured and explained, and it gives no runtime speedup. Added detail is labeled synthesis. Revealed detail is real. Where the data ends, the drill-down ends.