5 comments

  • evantahler 1 hour ago
    I feel like asking the thing that you are measuring, and don’t trust, to measure itself might not produce the best measurements.
    • john_strinlai 1 hour ago
      "we investigated ourselves and found nothing wrong"
  • Retr0id 43 minutes ago
    What is "drift"? It seems to be one of those words that LLMs love to say but it doesn't really mean anything ("gap" is another one).
    • jldugger 8 minutes ago
      IDK how it applies to LLMs but the original meaning was a change in a distribution over time. Like if you had some model based app trained on American English, but slowly more and more American Spanish users adopt your app; training set distribution is drifting away from the actual usage distribution.

      In that situation, your model accuracy will look good on holdout sets but underperform in user's hands.

    • idle_zealot 40 minutes ago
      I believe it's businessspeak for "change." Gap is suittongue for "difference."
  • aleksiy123 2 hours ago
    Interesting approach, I've been particularly interested in tracking and being able to understand if adding skills or tweaking prompts is making things better or worse.

    Anyone know of any other similar tools that allow you to track across harnesses, while coding?

    Running evals as a solo dev is too cost restrictive I think.

  • wongarsu 1 hour ago
    See also https://marginlab.ai/trackers/claude-code-historical-perform... for a more conventional approach to track regressions

    This project is somewhat unconventional in its approach, but that might reveal issues that are masked in typical benchmark datasets

  • tejpalv 3 hours ago
    [dead]