Learning diary › Year 2025 › May, 2025 › 2025-05-08 [2025-05-08]
Learning diary › Year 2025 › May, 2025 › 2025-05-08 [2025-05-08]
Why transformers need adam: A hessian perspective[zhang2024transformers] - insightful, and well summarized related work - trying to figure out a way to let AI agent to read all papers citing an paper, and write a summary of the follow-up research - The magic of software; or, what makes a good engineer also makes a good engineering organization - "Engineers are most capable and most effective when abstraction layers serve as shorthand for an understanding of what it is doing on your behalf, rather than a black box" - "start with deep understanding as the basis for innovation – cultivating the curiosity to look inside the black boxes" - "The magic of both software and software organizations comes from those moments where insight into how something works sparks entirely new ideas about what it could become" - related: Reinvent the Wheel - Implement your language twice - Futhark is a statically typed, data-parallel, and purely functional array language with AoT compilation to CUDA, OpenCL, HIP, and WebGPU (WIP) - also Comparing the performance of OpenCL, CUDA, and HIP - Objective-C interop with Zig? - got zig to work inside lima with minimal setup-