Parse notes
This guide explores profiling techniques to pinpoint high CPU usage in Node.js applications. It covers using built-in profilers, Chrome DevTools, the Node Inspector API, and the perf tool. The guide also mentions continuous profiling for ongoing monitoring and explores memory leak detection.
A poor Largest Contentful Paint (LCP) score can often be caused by backend issues. While tools like Page Speed Insights and Google Lighthouse are helpful for frontend analysis, tracing user journeys and examining spans help uncover more issues. By doing so, developers can pinpoint problematic database queries or other backend operations that contribute to slow LCP scores.
Teams should focus on fewer tasks and allow for more downtime, which counterintuitively leads to faster feature delivery and increased value. When working on too many things at once, teams may actually deliver things slower because there are too many things that are in a work-in-progress state. Smaller, faster completions keep momentum going.
This author switched teams to AI/ML Training Infra due to the fast growth and opportunities in the AI space. They were also motivated by curiosity about the technical challenges in AI compute efficiency and the opportunity to work with talented engineers.
json.bash is a command-line tool and bash library that creates JSON.
fast-json-stringify is a library that converts JavaScript objects to JSON strings. It's faster than JSON.stringify, especially for small payloads.
A collection of Vue tools that help make development easier. This includes a type-check command line tool, TypeScript plugins, VSCode extensions, and more.
Property-based testing is a technique that originated from academic research. It has seen widespread adoption in various programming languages. However, most current libraries lack advanced features like stateful and parallel testing, which were established as state-of-the-art fifteen years ago. This lack of progress is due to the perception that pure function testing is more valuable, the need for specialized training for stateful modeling, and the success of closed-source solutions in industry adoption.
Choosing the right embedding model for search and RAG applications requires considering your specific needs and evaluating different models on your own data. The MTEB Leaderboard is a helpful starting point, but validation on your own use case is necessary for the best results.
Canva's experimentation analysis platform started with a data scientist-focused Python package and a command-line interface. It then transitioned to an interface that could be used by all employees and a separate microservice for analysis components. This allowed it to manage experiments much easier in-house.
Lots of people have pivoted into AI/ML development by doing courses on the side, transitioning within their company, and/or getting into data engineering.
One million particles rendered with the GPGPU technique using WebGL, with source code and explanations provided.
Palico AI is an LLM development framework for fast iteration.
Projects fail due to resource scarcity, multitasking, and siloed teams.