{
  "$schema": "https://raw.githubusercontent.com/guitaripod/wwdc-sessions/master/schema/session.schema.json",
  "schemaVersion": 1,
  "id": "wwdc2026-328",
  "event": "wwdc2026",
  "eventName": "WWDC26",
  "year": 2026,
  "eventContentId": 328,
  "title": "Explore numerical computing in Swift with MLX",
  "description": "Bring NumPy-style computing natively to Swift with MLX Swift. Discover how to eliminate cross-language friction in your machine learning workflows by handling image processing, tensor operations, and neural network training in a single, type-safe environment. Explore the APIs that let you leverage GPU acceleration while enjoying the compiler, tooling, and debugging experience you already know.",
  "url": "https://developer.apple.com/videos/play/wwdc2026/328",
  "sosumiURL": "https://sosumi.ai/videos/play/wwdc2026/328",
  "type": "Video",
  "publishedAt": "2026-06-08T14:00:00-07:00",
  "platforms": [
    "macOS"
  ],
  "keywords": [],
  "topics": [
    "AI & Machine Learning",
    "Swift"
  ],
  "primaryTopic": "Swift",
  "duration": 871,
  "hasTranscript": true,
  "transcriptWordCount": 2168,
  "availableLanguages": [
    "eng",
    "fra",
    "jpn",
    "kor",
    "por",
    "spa",
    "zho"
  ],
  "media": {
    "hls": "https://devstreaming-cdn.apple.com/videos/wwdc/2026/328/5/51d0ab0a-f401-4514-9f04-6b211897d3e8/cmaf.m3u8",
    "downloadHLS": "https://devstreaming-cdn.apple.com/videos/wwdc/2026/328/5/51d0ab0a-f401-4514-9f04-6b211897d3e8/cmaf-download.m3u8",
    "downloadSD": "https://devstreaming-cdn.apple.com/videos/wwdc/2026/328/5/51d0ab0a-f401-4514-9f04-6b211897d3e8/downloads/wwdc2026-328_sd.mp4",
    "downloadHD": "https://devstreaming-cdn.apple.com/videos/wwdc/2026/328/5/51d0ab0a-f401-4514-9f04-6b211897d3e8/downloads/wwdc2026-328_hd.mp4",
    "duration": 871
  },
  "codeSnippets": [
    {
      "title": "Power iteration with MLX Swift arrays",
      "language": "swift",
      "startTimeSeconds": 184,
      "code": "import MLX\nlet n = 100\nlet steps = 10\nlet B = MLXRandom.normal([n, n])\nvar v = MLXRandom.normal([n])\n\n// get symmetric matrix A = Bᵀ + B\nlet A = B.T + B\n\n// Power iteration → top eigenvector of A.\n//   v ← A v / ‖A v‖\nfor _ in 0 ..< steps {\n    let Av = matmul(A, v)\n    v = Av / norm(Av)\n    eval(v)\n}\n\n// recover the eigenvalue.\n//   λ = vᵀ A v\nlet lambda = matmul(matmul(v.T, A), v)\n\nprint(lambda)"
    },
    {
      "title": "Mandelbrot set in plain Swift (scalar)",
      "language": "swift",
      "startTimeSeconds": 309,
      "code": "// Plain Swift, scalar-at-a-time\nvar counts = Array2D<Int>(width: w, height: h)\n\nfor y in 0 ..< h {\n    for x in 0 ..< w {\n        let c = Complex(xMin + Float(x) * xStep, yMin + Float(y) * yStep)\n        var z = Complex<Float>.zero\n        var limit = maxIterations\n        for i in 0 ..< maxIterations {\n            z = z * z + c\n            if z.lengthSquared > radiusSquared {\n                limit = i\n                break\n            }\n        }\n        counts[x, y] = limit\n    }\n}"
    },
    {
      "title": "Mandelbrot set in MLX Swift (array)",
      "language": "swift",
      "startTimeSeconds": 327,
      "code": "// Compute the Mandelbrot set on a grid of complex numbers\nimport MLX\n\nlet x = linspace(Float(-2.0), 0.5, count: w)\nlet y = linspace(Float(-1.25), 1.25, count: h).reshaped(h, 1)\nlet c = x + y.asImaginary()\n\nvar z = MLXArray.zeros(like: c)\nvar counts = MLXArray.zeros(c.shape, dtype: .int16)\n\nfor _ in 0 ..< maxIterations {\n    z = z * z + c                       // iterate z ← z² + c\n    counts = counts + (abs(z) .< 2)     // count bounded iterations\n}"
    },
    {
      "title": "Jacobi iteration with conv2d",
      "language": "swift",
      "startTimeSeconds": 447,
      "code": "// Jacobi iteration: average the four neighbors\n\n// Convolution weights\nlet kernel = MLXArray(converting: [\n    0,    0.25, 0,\n    0.25, 0,    0.25,\n    0,    0.25, 0,\n]).reshaped(1, 3, 3, 1)\n\n// Initial value\nvar temperature = heatSources\n\n// Run this in a loop until convergence\nlet next = conv2d(temperature, kernel, padding: 1)\ntemperature = which(heatMask, heatSources, next)"
    },
    {
      "title": "Successive Over-Relaxation (SOR)",
      "language": "swift",
      "startTimeSeconds": 557,
      "code": "// Successive Over-Relaxation: blend the previous and next state\nlet ω: Float = 2.0 / (1.0 + sin(Float.pi / Float(max(M, N))))\n\nlet redMask   = checkerboard(rows: M, cols: N, phase: 0)\nlet blackMask = checkerboard(rows: M, cols: N, phase: 1)\n\n// Update red cells using black neighbors\nlet sorRed  = ω * conv2d(temperature, kernel, padding: 1) + (1 - ω) * temperature\ntemperature = which(redMask, sorRed, temperature)\ntemperature = which(heatMask, heatSources, temperature)\n\n// Update black cells using (now-updated) red neighbors\nlet sorBlack = ω * conv2d(temperature, kernel, padding: 1) + (1 - ω) * temperature\ntemperature  = which(blackMask, sorBlack, temperature)\ntemperature  = which(heatMask, heatSources, temperature)"
    },
    {
      "title": "Curve fitting with automatic differentiation",
      "language": "swift",
      "startTimeSeconds": 673,
      "code": "// Define a loss, then optimize it with autodiff\n// x, y: data points as MLXArrays\nfunc f(_ θ: MLXArray) -> MLXArray {\n    θ[0] + θ[1] * x + θ[2] * x ** 2\n}\n\nfunc loss(_ θ: MLXArray) -> MLXArray {\n    mean((f(θ) - y) ** 2)\n}\n\nvar θ = zeros([numParams])\nlet gradLoss = grad(loss)\n\nfor _ in 0 ..< steps {\n    let g = gradLoss(θ)         // ∇L(θ)\n    θ = θ - learningRate * g    // parameter update\n    eval(θ)                     // force evaluation\n}"
    }
  ],
  "resources": [
    {
      "title": "MLX Swift LM on GitHub",
      "url": "https://github.com/ml-explore/mlx-swift-lm",
      "type": "documentation",
      "description": null
    },
    {
      "title": "MLX Swift Examples",
      "url": "https://github.com/ml-explore/mlx-swift-examples",
      "type": "samplecode",
      "description": "MLX Swift Examples"
    },
    {
      "title": "MLX Examples",
      "url": "https://github.com/ml-explore/mlx-examples",
      "type": "samplecode",
      "description": "This repo contains a variety of standalone examples using the MLX framework."
    },
    {
      "title": "MLX Swift",
      "url": "https://github.com/ml-explore/mlx-swift",
      "type": "documentation",
      "description": "MLX Swift is a Swift API for MLX."
    },
    {
      "title": "MLX LM - Python API",
      "url": "https://github.com/ml-explore/mlx-lm",
      "type": "documentation",
      "description": "MLX LM is a Python package for generating text and fine-tuning large language models on Apple silicon with MLX."
    },
    {
      "title": "MLX Explore - Python API",
      "url": "https://github.com/ml-explore/mlx",
      "type": "documentation",
      "description": "MLX is an array framework for machine learning on Apple silicon, brought to you by Apple machine learning research."
    },
    {
      "title": "MLX Framework",
      "url": "https://mlx-framework.org",
      "type": "documentation",
      "description": null
    },
    {
      "title": "MLX",
      "url": "https://ml-explore.github.io/mlx/",
      "type": "documentation",
      "description": "MLX is an array framework for machine learning on Apple silicon, brought to you by Apple machine learning research."
    }
  ],
  "path": "sessions/wwdc2026/328-explore-numerical-computing-in-swift-with-mlx/",
  "files": {
    "metadata": "sessions/wwdc2026/328-explore-numerical-computing-in-swift-with-mlx/metadata.json",
    "readme": "sessions/wwdc2026/328-explore-numerical-computing-in-swift-with-mlx/README.md",
    "transcript": "sessions/wwdc2026/328-explore-numerical-computing-in-swift-with-mlx/transcript.md",
    "transcriptJSON": "sessions/wwdc2026/328-explore-numerical-computing-in-swift-with-mlx/transcript.json"
  }
}