<?xml version="1.0" encoding="utf-8"?>
<feed xmlns="http://www.w3.org/2005/Atom">
  <title>eitam ring - systems notes</title>
  <subtitle>Go, GPUs, and pipelines that don't allocate</subtitle>
  <link href="https://eitamring.github.io/feed.xml" rel="self"/>
  <link href="https://eitamring.github.io/"/>
  <id>https://eitamring.github.io/</id>
  <updated>2026-07-11T00:00:00Z</updated>
  <author><name>Eitam Ring</name></author>
  <entry>
    <title>Ten weekends of CUDA in Go: 30 functions became 106</title>
    <link href="https://eitamring.github.io/posts/gocudrv-ten-weekends.html"/>
    <id>https://eitamring.github.io/posts/gocudrv-ten-weekends.html</id>
    <published>2026-07-11T00:00:00Z</published>
    <updated>2026-07-11T00:00:00Z</updated>
    <summary>A progress update on gocudrv, a cgo-free CUDA driver API binding for Go: the thread-pinned executor, embedded PTX, and why the CPU timer was off by 18x.</summary>
  </entry>
  <entry>
    <title>ClassMesh v1: classifying a million log lines for the price of electricity</title>
    <link href="https://eitamring.github.io/posts/classmesh-v1.html"/>
    <id>https://eitamring.github.io/posts/classmesh-v1.html</id>
    <published>2026-07-09T00:00:00Z</published>
    <updated>2026-07-09T00:00:00Z</updated>
    <summary>A confidence-gated classification cascade in Go. Deterministic rules first, models next, expensive calls only for what's left. Zero allocations on the hot path.</summary>
  </entry>
</feed>
