TLDR Web Dev

TLDR WEB DEV 2024-09-25

5 sections17 entrieslinks blockcomplete parse

Parse notes

  • Subscription/account footer truncated before parsefooter_truncated

01ARTICLES & TUTORIALS

REDESIGN CASE STUDY 2024

This article goes over a complete frontend redesign of a website that involved creating a new header, a unique hero section featuring personalized hats drawn by the author's daughter, and a circular navigation for the hats section. It also incorporated interactive elements using Framer Motion and scroll-driven animations, along with the usage of CSS features like container queries, :has(), offset-path, and more.

ishadeed.com19 min read

POSTGRES WRITE-AHEAD LOGS

Postgres Write-Ahead Logs (WAL) are needed for logical replication in Postgres. WAL works by storing each state change as a command in an append-only file before the change is actually made to the database, allowing for recovery from the last checkpoint in case of a crash. WAL offers various configurable parameters like `wal_level`, `fsync`, `wal_buffers`, and `checkpoint_flush_after` to optimize performance and control data retention.

artie.com8 min read

WHAT I'VE LEARNED IN THE PAST YEAR SPENT BUILDING AN AI VIDEO EDITOR

This developer, while building a generative video platform, used Microsoft Promptflow and Temporal to build generative workflows but encountered some challenges with unpredictable LLMs and embeddings. He found that embeddings might not be the best solution for every search problem and traditional search methods may be needed in some cases.

makeartwithpython.com13 min read

02OPINIONS & ADVICE

MOVE FAST AND ABANDON THINGS

This former Mac game developer reflects on his experience of rapid prototyping and abandoning unfinished projects, arguing that this approach, while seemingly chaotic, was ultimately effective for him. He quickly tested game ideas as it allowed for the identification of truly compelling concepts. This "move fast and abandon things" philosophy was applicable even in his later career as a software engineer at Apple, where he found himself a "blue-collar" programmer amidst "white-collar" professionals.

engineersneedart.com11 min read

WHAT I TELL PEOPLE NEW TO ON-CALL

New on-call engineers should not fix every issue personally, but rather assess the situation and ensure it's addressed. They should also shadow experienced engineers to learn best practices and prioritize their well-being by not sacrificing personal health for non-critical issues.

ntietz.com9 min read

SOFTWARE DEVELOPMENT TOPICS I'VE CHANGED MY MIND ON AFTER 6 YEARS IN THE INDUSTRY

After six years in the industry, this developer now values typed languages for team collaboration, sees standups as helpful for onboarding, and believes sprint retrospectives are valuable when focused on actual course correction. Software architecture is more important than he once thought and practicality is always better than dogmatically adhering “best practices.”

chriskiehl.com3 min read

03LAUNCHES & TOOLS

Sponsored

HOW TO ADD AN ONBOARDING FLOW FOR YOUR APPLICATION WITH CLERK

As part of your onboarding flow, you may want to collect extra information from your user and use it to alter your application state. Follow this quick walkthrough example using Next.js and TypeScript [11] to add custom claims to your session token, read session data,

go.clerk.comSponsor

JSON FOR YOU

JSON For You is a tool that visualizes and processes JSON data with features like graph and table views, comparisons, validation, nested parsing, jq support, CSV import/export, and more.

github.comGitHub repository

04MISCELLANEOUS

ON IMPACTFUL AI RESEARCH

Impactful AI research focuses on open-source work through artifacts like models, systems, and benchmarks. Students should prioritize investing in projects over papers, selecting timely problems with large headroom and "fanout," and thinking two steps ahead while iterating fast.

github.com16 min read

THE DATA PIPELINE IS THE NEW SECRET SAUCE

In AI infrastructure for companies, data pipelines are the most important part of creating and managing proprietary datasets to train and fine-tune LLMs. There are various inference hosting models to use, including hosted APIs, edge computing, on-premise data centers, and third-party cloud hosting. Generally, enterprises will start with off-the-shelf cloud solutions and progress towards specialized infrastructure for specific use cases.

heavybit.com17 min read