https://macworks.dev/tags/reverse-engineering/Recent content in Reverse Engineering on MacWorksHugoenhttps://macworks.dev/docs/today/hackernews-2026-04-19/Mon, 01 Jan 0001 00:00:00 +0000https://macworks.dev/docs/today/hackernews-2026-04-19/<h1 id="hacker-news--2026-04-19">Hacker News — 2026-04-19<a class="anchor" href="#hacker-news--2026-04-19">#</a></h1> <h2 id="top-story">Top Story<a class="anchor" href="#top-story">#</a></h2> <p><a href="https://abacusnoir.com/2026/04/18/zero-copy-gpu-inference-from-webassembly-on-apple-silicon/">Zero-Copy GPU Inference from WebAssembly on Apple Silicon</a> On Apple Silicon, you can share a WebAssembly module&rsquo;s linear memory directly with the GPU—meaning zero copies, no serialization, and no intermediate buffers. By composing <code>mmap</code>, Metal buffers, and Wasmtime&rsquo;s custom memory allocator, the author ran a 1B parameter Llama model entirely from a Wasm guest with zero-copy overhead. This is pure, hardware-sympathetic engineering, proving that sandboxed runtimes don&rsquo;t have to ruin performance if you just leverage the underlying physics of the chip.</p>