Jonathan Teo made his name as an early Silicon Valley investor -- Snapchat and Instagram among his calls. Lately, under the handle qubixt, he's been posting about BTX, a proof-of-work chain built around post-quantum cryptography and AI-relevant mining.
The technical hook is MatMul mining: instead of the single-purpose ASIC hash lotteries that secure chains like Bitcoin, BTX's proof-of-work runs matrix-multiplication computation, the same math underneath modern AI models. That means the GPUs, Apple Silicon and CPUs securing BTX aren't stranded hardware -- they can run open AI models or numerical workloads whenever they're not mining.
In an Instagram post yesterday, Teo described a new revision of BTX's work algorithm, v4.6, as pushing the chain into "its own tier," calling it the only work chain that requires AI-exact computation as proof, with each mining episode converging compute, memory and connectivity into a single deterministic result across more than two quadrillion operations. That's Teo's own framing, shared as early alpha -- BTX's official documentation at btx.dev still shows v0.33.1 as current and is expected to catch up once the new algorithm formally ships. What the site does confirm today: the MatMul design, a capped 21 million supply, and production post-quantum signatures already live for ordinary transfers.
Teo's broader thesis: every major capital cycle produces its own hedge assets, and permanent AI compute demand isn't guaranteed -- which is exactly the uncertainty a scarce, independently-issued asset can hedge against. He's positioning BTX as the hedge for the AI capex cycle specifically, a way to gain exposure to AI infrastructure demand without betting on any single company's outcome.
BTX is already supported inside bonuz's self-custodial wallet: no seed phrase, sign in with email or social, and roughly 185,000 BTX already held by bonuz users. It's also directly mineable and accessible through the browser-based PQ wallet at btx.dev.
