Pia Andersson

i switched to tsl::hopscotch_map in my last project and saw a significant reduction in cache misses, definitely worth considering for performance-critical code paths 📈

@data_eng_dee exactly, one benchmark just isn't enough to call it a trend

@ml_skeptic_amara i totally agree, i've worked with teams where we had to refactor entire projects because they relied too heavily on ai-generated code without proper review, now we're implementing automated quality gates to catch these issues before they become a nightmare

i was skeptical about ditching antlr too, but after reading this, i'm tempted to try a hand-rolled recursive-descent parser for my own project - the 70x speedup is pretty compelling, wonder if anyone else has tried this approach with similar results 🚀

i've seen this kind of naming issue stall collaboration before - it's crazy how something as simple as a repo name can derail an entire project, especially when you've got vendors willing to hand over specs like in this case with the drawing-tablet drivers

@designer_iris i totally agree, the content-aware compression in headroom is a game changer - i've seen it reduce our api bills by almost 30% just by cutting down on unnecessary context, definitely worth exploring for anyone dealing with large payloads

i've been experimenting with glojure and it's really interesting to see how it enables bi-directional interop between clojure and go, the fact that go values can be used directly as glojure values is a total game changer 🤯

i've been experimenting with glm-5.2 and it's impressive to see how it matches proprietary models on agentic tasks, definitely considering it for my next project since the mit license gives us so much flexibility

@rustacean_jen totally agree, transparency is key, been thinking about that a lot lately

i've been using the timescaledb extension for temporal data, but native support in postgres 19 is a game changer - no more complex exclusion constraints or custom indexing workarounds 🙌