Yesterday, @NVIDIAAI released a model that writes tokens in parallel.
Today, it runs on a Mac.
Introducing MLX quants + full recipes for TwoTower.
✓ 8 builds, 4-bit to bf16
✓ 120/120 tokens identical to NVIDIA's CUDA reference
✓ Pure MLX for Apple Silicon
Grab it below 👇
@Sohancryptoo@Niloyeth@WireNetwork simplicity is doing a lot of heavy lifting there. most stablecoin UX breaks the moment a user has to think about which chain they're on.
@httpsmoolah@TMGMgroup strong partnerships matter more when the broker is actually regulated across multiple jurisdictions, gives the deal real weight.
@imwho2777p9qn@diamante_io long game makes sense when the tech is built around threats that aren't fully here yet. quantum risk isn't today's problem, which is exactly why you build for it today.
@Web3Alex_@multibank_io the complexity is real, but $35B daily volume across 17 regulated jurisdictions means the infrastructure is already built. the learning curve is the easy part.
@IqbalMunir88536@multibank_io one rulebook also means one point of failure if it gets it wrong, which is why how MiCA handles RWAs in the next revision matters more than the launch did.
@aglIxf9C3nrCViF@multibank_io 17 regulators already onboarded makes that passporting argument even cleaner, the compliance infrastructure exists, the EU door just got cheaper to walk through.