News for hackers. Stuff that matters.
Re:Is Opus 5.5 nerfed? Thinking patterns changed on November 3 by brainwad on Monday November 09, 2026 @01:53PM (#50429855)
> You cant experiment on human populations tool use the same way you run AI/ML experiments.

Google and Facebook have proven otherwise. Unless you are very, very lucky you are always part of multiple running experiments on their products, and nobody really cares.

Re:Is Opus 5.5 nerfed? Thinking patterns changed on November 3 by _blk on Monday November 09, 2026 @02:12PM (#50430170)
If it's mostly indiscernible from baseline that's perfectly fine and allows for incremental improvement. Think infinitesimal changes in light of multiple dimensions. When you have separate development and stable versions there's always the risk of packing too many large changes and no one actually wanting to upgrade (given a choice) because the product is already good enough. Pro and con for both approaches but IMO, you need mostly small changes and occasional big steps when small ones can't get you there.