Languages

1. rust
   • [https://sharnoff.io/blog/why-rust-compiler-slow](Why is the Rust compiler so slow?)
     • 
     • 
     • debug , LTO, optimization
     • RUSTFLAGS="-Cllvm-args=-inline-threshold=50 -Cllvm-args=-inlinedefault-threshold=50 -Cllvm-args=-inlinehint-threshold=50"
     • Reducing inlining with LLVM args;
     • Breaking up expensive functions in the main crate; and
     • Removing generics from dependencies to prevent needing to compile it in the main crate Timeline:
     • We started at ~175s
     • Disabling LTO (and debug symbols!) got us to 51s (-71%)
     • Changing to opt-level = 1 on the final crate got us to 48.8s (-4%)
     • Reducing inlining with -C llvm-args got us to 40.7s (-16%)
     • Local changes got us to 37.7s (-7%)
     • Changes with dependencies got us to 32.3s (-14%)
     • Enabling -Zshare-generics got us to 29.1s (-10%)
     • And switching away from alpine got us to 9.1s (-69%)
   • Leaktracer: A Rust allocator to trace memory allocations
2. zig
   • zig's new async io
   • What is Zig's “Colorblind” Async/Await?
3. Java
   • 小红书JDK升级带来10%整体性能提升，这份升级指南收好了！ 千年的JDK8问题
     • G1: Parallel Full GC for G1
     • G1: Initiating Heap Occupancy Percent
     • G1: 提前回收大对象 总的感觉来说，对大型系统，Java的GC仍然带来了一定程度的不可控因素，所有的 GC 调优其实都是在这个不确定性上跳舞，获得或多或少的改进， 但仍然是存在不确定性。这就是在一个低维度的挣扎。（不过，作为应用层的编程语言，不采纳 GC 的成本可能更高 ）
     • Java 的优点和缺点可能是现在的生态，在不需要那么臃肿的场景下，可能会越来越臃肿。
   • 从 Java 迁移到 Swift: 密码监控服务

MPP & OLAP

Web & Frontend

1. Dashboard That Works: A Step-by-Step Guide for Startups in 2025

   A truly functional dashboard:

   • GOALS, DATA, PEOPLE, LOGIC
   • analysts, designers, developers
   • work for business, not just look pretty
   • Don't try to make a dashboard for everyone
   • define use cases and users
   • identify key metrics
   • prepare data
   • sketch a rough layout
   • get user feedback
   • write a solid brief for designers
   • find the right builders
   • Launch, watch, improve

2. [fastplotlib](a plotting library built on WGPU)

   • based on [pygfx](a python-graphics library visualization framework on WGPU) 思考：结合 Vega, Pixi, D3 等框架，思考一种描述 visualization 的最佳方式。
   • Vega：使用 JSON 的方式，缺乏类型化约束，不便于理解、使用，部份的能力过于糖化。
   • 使用多维空间的方式来理解 visualization
   • 将数据（关系数据、多维数据）映射到（可视化）多维空间。
   • 使用 react 的方式来理解 interact
   • 更好的表示语言？
     • JSON based
     • more datatype: date, interval, float, decimal, more via regexp"..." style
     • JSON constructor: Color { r g b }
     • functional constructor: Color(r, g, b), with fixed parameter and named parameter
     • enum literal. .RED instead of Color.RED
     • reference spec: global reference, document reference, scoped reference.

3. AntV G2 粗览一下文档和示例，G2 的概念与 vega 时非常相似的，提供2种 API: 命令式 和 DSL 式。

LLM & Agents

1. LLM Powered Autonomous Agents
2. AutoGPT
3. AnyTool
4. HuggingGPT
5. Understanding the planning of LLM agents: A survey
6. https://github.com/AntonOsika/gpt-engineer
7. JoyAgent
8. Gaia benchmark: a benchmark for General AI Assistants