Bitcoin built the chassis.
BTX changed the engine.

An SHA-256 ASIC has one job. It is purpose-built down to the silicon for one specific operation, and the day it stops being profitable to do that operation it becomes scrap metal. The capital invested in the chip cannot be redeployed to anything else, because the chip cannot do anything else. The depreciation profile is brutal, and when the network re-rates difficulty, the marginal facility hits zero margin without warning.

A GPU has many jobs. It mines BTX when block reward arithmetic favours mining. It serves inference when an application calls it. It trains when there is a training run to do. It backfills idle cycles into one of those when the others are slack. Apple-Silicon Metal does the same for a different shape of operator — a household, an SMB, an inference rig on a desk. The same dollar of silicon shows up across multiple revenue streams, with its residual value tied to the asset class hyperscalers are chasing, not to a single chain's difficulty curve.

That is the operator-side argument for BTX. We made the household and small-operator version in Mining pools, AI agents, and the Mac mini that earns. Dual-use silicon is what makes that argument economically defensible — and what makes single-purpose ASIC capacity look, for the first time, like a stranded asset waiting to happen.

Every blockchain whose signatures are classical — Bitcoin, Ethereum, every fork of either — sits on a cryptographic substrate that a sufficiently capable quantum computer will eventually break. Knowing when is unsettled. Knowing that the only response is a coordinated migration to post-quantum schemes is not. That migration has to touch every wallet, every exchange integration, every custodian's key management, and every long-lived treasury. Anything held in keys that are lost, inactive, or otherwise unable to participate becomes permanently exposed the moment the threat materialises.

BTX removed the problem at the root. Every spend uses ML-DSA-44 (FIPS 204) — the routine NIST-standardised lattice signature. Recovery paths use the more conservative hash-based SLH-DSA-128s (FIPS 205). Both have been required from genesis. There is no migration window to live through, no key-rotation event to coordinate, no policy debate about how to compensate keys that cannot upgrade. The chain ships with the answer baked in.

For a settlement layer that intends to host bank stablecoins, regulated payment activity, and machine-driven autonomous treasuries — the topics covered in When the block reward is a matrix multiply and Machine-coded banking policies on BTX — this is the difference between a chain you can underwrite for 30 years and a chain whose entire underwriting has a deferred but inevitable hard-fork event embedded in it.

An SHA-256 mining operator turns a kilowatt-hour into a probability of finding a block and a discarded hash. The block reward, less power and overhead, is the only revenue line. When difficulty rises or the price of the asset falls, that single revenue line determines whether the facility runs.

A BTX-capable operator running the same kilowatt-hour through a GPU has at least two revenue lines. There is the BTX block reward, denominated in matmul work that the network already values. And there is the alternative use of the same cycles to serve inference or train a model — work that hyperscale tenants pay for directly. The operator can choose which line to push toward at any moment, can backfill idle inference cycles into mining automatically, or can split the same GPU across both via a co-scheduler. The capital is the same; the optionality is materially larger.

That optionality has knock-on effects all the way up the stack. Financing terms for a dual-use facility look different from a single-purpose mining shed because the residual value of the silicon is anchored to the AI hardware market, not to a single chain's difficulty. Tenant negotiations look different because the operator has a credible BATNA. And the long-tail of small operators — household-class Mac minis, on-prem workstations, SMB inference rigs — can participate at all, which on a hashing chain they fundamentally cannot.

The Energy Magazine piece is right about what the mining industry built. The megawatt-scale, substation-fed, fibre-connected, regulator-known facility is the physical platform the AI economy now stands on, and it took a decade of mining operators to build the playbook. The chassis is sound and the credit is deserved.

The engine is the part that has aged. SHA-256 produces nothing the AI economy can use. Classical signatures sit over a deferred hard-fork problem. ASIC capital cannot be redeployed when difficulty re-rates. None of those constraints are inherent to power-secured compute; they are inherent to the choices made when the original engine was designed. BTX keeps the chassis and changes the engine: matrix multiplication for the work, post-quantum signatures for the keys, dual-use GPUs for the silicon, lattice-based privacy for the ledger. Same megawatts. Different output.