Explore numerical computing in Swift with MLX

A Podkey summary of Explore numerical computing in Swift with MLX, from WWDC 2026.Today’s roundup is really about one idea showing up in a bunch of useful ways: MLX Swift lets you write array-based code that stays clean, runs fast, and scales from little experiments to serious GPU work. The big themes are lazy evaluation, automatic differentiation, and a cross-language setup that makes Swift feel a lot less isolated than people sometimes assume. And the examples are nice and concrete, from Mandelbrot rendering to heat diffusion to fitting a simple curve without hand-deriving anything.Why lazy evaluation mattersArray computing and the GPUMandelbrot as the clean speed demoHeat diffusion and why Jacobi is slowWhy SOR improves thingsAutodiff for curve fittingA shared API across languagesOpen source and the growing ecosystemThis podcast was created with Podkey. Make your own at https://podkey.fm