yceffort
02 POSTS

Testing 1

  • ◆ Swapping the Model Behind a Coding Agent Part 2
    #ai#testing

    Building 36 SWE-bench Tasks from 20 Internal PRs

    Turning merged PRs from an internal WebSocket package into coding-agent evaluation tasks, validating the grader, and isolating personal instructions from the Claude Code run environment

    Updated 22 min read
  • #bundler#web-performance#debugging

    Tearing Apart a Third-Party SDK and Rebuilding It My Way: Without Touching a Single Line of Logic

    I imported one constant and 97.7% of the bundle came along with it. The vendor had no timeline for a fix, so I recovered over 400 TypeScript files from the published source maps, rebuilt the build, entry points, and dependencies while leaving the logic alone, and cut the raw size of /send by 77.5%. All 1,932 tests passed, but a few of them were checking nothing at all, so I had to work out again how far I could confirm that the rebuilt SDK behaves the same as the original.

    41 min read