elasticity, plasticity, and young's wave dynamics
measuring the stiffness of the latent space to maximize intelligence density.
elasticity vs plasticity
In our substrate, Plasticity refers to permanent weight updates (LoRA/SFT). It is a permanent topological deformation. Elasticity refers to in-context learning (KV cache, prompt engineering). It is a reversible deformation that snaps back when the session clears.
We can measure the stiffness of the model using Young's Modulus (E), defined as the ratio of Stress (prompt complexity/information load) to Strain (the model's adaptation).
the yield point and wave dynamics
When an informational wave (a query) hits the model, it deforms elastically. But every model has a Yield Point (Elastic Limit). If the stress exceeds this limit, the model can no longer adapt in-context. It fractures (hallucinates) or disperses the energy as heat.
This is the Wattage-Spike Gluon. The model hit its elastic limit, resulting in immense topological friction (measured as a spike in Information Fugacity).
potentiating intelligence density
This dynamic provides the ultimate gating mechanism for the RSI loop. If a task fails but fugacity is low, the model is still in the elastic regime—adjust the prompt, do not train. We save expensive training compute.
If a task fails and fugacity spikes, the model has hit its Yield Point. No amount of prompting will help. This is the exact moment to arm the Plasticity module and permanently deform the weights. By gating plasticity behind the elastic limit, we maximize overall intelligence throughput.