The data to watch for $PRL. If the @prlnet work @WeinsteinOmri has teased gets matmuls fast and inference overhead down $PRL net energy expenditure should go up dramatically. This is because if inference is faster on PRL kernels than existing stacks it is a no brainer for inference providers to move over. To not do so would leave both inference revenue and mining revenue on the table. Inference providers moving to PRL stack would increase blockchain difficulty and may put a rapidly increasing floor under price (assuming useless mining cost or energy cost somewhat function as floors). To be explicit, we still don't have a full picture on the demand side. Just supply. It likely won't play out exactly as said above but still worth watching.