Since most harnesses today use some form of context compaction, I would like to see more measurement...
elvis(@omarsar0) · 人工智能
Since most harnesses today use some form of context compaction, I would like to see more measurement around it. It's become even more important with longer-running tasks and multi-agent systems. I really don't trust the built-in auto-compaction mechanisms. I have built my own auto-compaction, optimized heavily for my tasks and frequently assessed. 💬 3 🔄 1 ❤️ 7 👀 889 📊 4 ⚡ Powered by xgo.ing