I think people compare bundles of code and tools instead of asking "how much capability does it actually let you extract without forcing you to own everything around it?" when it comes to harnesses/harness engineering And there's enough harnesses that you can spend all day naming them, but the tradeoff I always seem to run into is that the more freedom they give you, the more infrastructure you become responsible for maintaining, the less time you get to work on things you want or need to And the reverse is that the less freedom you have, the more you end up fighting the designers choices and questioning their sanity whist losing your own Amp manages the tradeoff by having a small set of primitives that are ridiculously well engineeried underneath so I can chat to puck, spin up orbs, attach runners and testing plugins and work on an insane number of things all at once without thinking twice And what helped me come to this realisation was my raw dog test where when I test a new harness, I don't give it anything and see how far I can get with it just out of the box, and I got ridiculously far with amp. The reason being the primitives are really well chosen. i.e Pi's primitives are built around composition so every new capability needed is an engineering project, which is fine when that's what you want to do. But Amp is built really well around delegation so you get a lot more work done and the whole experience feels like you're living in the future