Tag
This article critiques flaws in various benchmarks, including performance estimation tools like napkin-math, AI model evaluations such as DeepSWE and Senior SWE-Bench, and analogies with car tires, to highlight common issues in benchmarking.
A Twitter user discusses the performance of Fable 5.1 in Blender, comparing its output to a reference image and noting it is good but not perfect.
AMD is working hard to improve its software for AI workloads, especially agentic tasks, with potential for better performance per dollar than NVIDIA, but software and leadership issues currently hinder progress.
Gergely Orosz discusses how modern software often neglects performance, citing examples like Bun and Linear, and highlights a fast demo of Superlogical for tmux by Mitchell Hashimoto.
A podcast episode on the Pragmatic Engineer podcast featuring Casey Muratori discussing software performance and the use of AI to assist in learning Assembly.
The author argues that while LLMs make certain optimizations cheaper, software can still be slow due to factors like increased tolerance for latency in asynchronous workflows and other practical constraints.
The article argues that LLMs and AI tools are lowering the barrier for software performance optimizations, enabling custom adaptations like JIT compilers and regex engines that were previously too costly to implement.
The article discusses how LLMs are reducing friction in programming language choice, leading developers to adopt performance-oriented languages like Rust and Zig for fast, small software and enabling work with complex technologies previously considered difficult.