Search the archive

Register with just your email address →

Codeium, Indeed, and many others, in 25+ talks and workshops

covering:

HOW TO CHOOSE THE RIGHT DATABASE

Common database use cases include text search, file storage, large-scale analytics, spatial data, and caching. This article goes over the right database to choose for each use case. For example, relational databases are best for ACID-compliant structured data, while NoSQL solutions like MongoDB are better for flexible schemas.

WHAT WE LEARNED COPYING ALL THE BEST CODE ASSISTANTS

Val Town's 2024 development involved incorporating features from leading AI code assistants like GitHub Copilot, Codeium, and ChatGPT. It initially focused on autocomplete and simple chat interfaces, then moved on to implement features similar to ChatGPT's function calling and Claude's code generation capabilities. Val Town faced challenges like slow generation speeds and unreliable diff functionality, but also had some unique wins, like automatic error detection and a streamlined hosting/API experience.

HITTING OKRS VS DOING YOUR JOB

In software engineering, OKRs often redundantly reflect the product roadmap and fail to communicate new initiatives or changes in focus. In marketing, OKRs highlight specific campaigns and targets for a given quarter, complementing rather than duplicating ongoing work. The difference is from the nature of the work: engineering involves continuous product development, while marketing executes more project-based campaigns. Engineering teams could learn something from marketing teams when setting OKRs.

HOW I PROGRAM WITH LLMS

This developer's LLM usage falls into three categories: autocomplete, improved search, and chat-driven programming. Chat-driven programming is great for well-defined tasks, allowing the LLM to generate code drafts. The developer then refines the code with his own code along with using the LLM's ability to quickly iterate on feedback and handle repetitive tasks.

HOW DO YOU DEAL WITH NON TECH PEERS WHO TRY TO SUGGEST THAT THEY MIGHT BE TO AND WANT TO TRY TO WRITE CODE USING AI TOOLS? (REDDIT THREAD)

A software engineer asked Reddit for advice on handling non-technical colleagues who want to use AI tools to write code, particularly for UI components. Many experienced developers in the Reddit thread suggested letting colleagues try, as long as they understood that the code will require thorough code review first. This approach allows non-technical staff to learn firsthand the complexities of software development, while still being able to help with simpler development tasks.

Start Building

BIPPY

Bippy is a toolkit that allows access to React's internal fiber tree by mimicking React DevTools. It lets you inspect and manipulate React internals without modifying React code, but carries the risk of breaking applications due to reliance on potentially unstable internal APIs.

PDF TAB RENAMER

This Chrome extension improves tab readability by renaming tabs displaying scientific papers from various sources (like arXiv, Google Scholar, and ACM) with more informative titles. It's useful for developers who usually have to read or refer to papers for their work.

THEY SQUANDERED THE HOLY GRAIL

Apple Intelligence has an impressive underlying secure computing architecture, but unfortunately, the released features are underwhelming and half-baked. They fall short of Apple's typical high standards. Apple has squandered the opportunity to deliver a truly groundbreaking product, focusing instead on underwhelming AI tools.

MY REVIEW OF THE COMING WAVE

Bill Gates recommends "The Coming Wave," written by Mustafa Suleyman. Suleyman is the current CEO of Microsoft AI and the co-founder and former head of applied AI at DeepMind. The book discusses three key risks: economic disruption, control problems with AI, and the empowerment of bad actors. There is an urgent need for containment strategies to mitigate these risks while still using AI for beneficial causes.

THE FUTURE OF HTMX

htmx aims to emulate jQuery's longevity and stability by prioritizing minimal API changes and focusing on bug fixes rather than frequent feature additions. Future development will emphasize stability, with quarterly releases containing primarily bug fixes and minor improvements. The project will concentrate on improving the hypermedia ecosystem surrounding htmx.

AMIDST THE NOISE AND HASTE, GOOGLE HAS SUCCESSFULLY PULLED A SPACEX

Google's long-term, vertically integrated approach to AI, from chip design (TPUs) to applications, mirrors SpaceX's success in space launch, creating a massive cost and performance advantage.

HOW DO YOU LIKE WHAT YOU'VE BUILT?

A simple question - "How do you like what you've built?" - prompted a colleague to identify and fix bugs and reconsider a problematic requirement during a code review.

ON LLMS AND CODE OPTIMIZATION

LLMs can suggest some improvements (like faster digit sum functions and parallelization), but they struggle with more nuanced algorithmic optimizations requiring human insight, such as strategically ordering operations to minimize computational cost.

THE EVOLUTION OF A STRUCTURAL CODE EDITOR

This post goes over the iterative development of a structural code editor for the EYG programming language, which evolved from a direct AST manipulation interface to a user-friendly system combining keyboard shortcuts with a visual, mouse- and touch-friendly UI.

BUILDING A POLITE & FAST WEB CRAWLER

A Mozilla engineer noticed that poorly-behaved bots, primarily from OpenAI and Amazon, caused 70% of the load on Diapsora's servers (an open-source social network). To address this, he built a “polite and fast” web crawler using Python and gevent, prioritizing respectful rate limiting by defaulting to one request per domain every five seconds and dynamically adjusting based on server responses. The crawler efficiently manages the queue by fetching robots.txt files in parallel and uses a sophisticated refetching strategy using HTTP headers to minimize redundant requests.

YOU DON'T NEED NEXT.JS

ComfyDeploy migrated its dashboard from Next.js to a plain React setup, which improved build times (from 3 minutes to 18 seconds) and hot reload speed (under 200ms). This decision was driven by escalating Vercel costs, testing difficulties with Next.js's server actions, and painfully slow local development. The migration incorporated TanStack Router and Rspack, forcing architectural improvements and cleaner code.

BACK TO BASICS: WHY WE CHOSE LONG POLLING OVER WEBSOCKETS

Inferable chose HTTP long polling over WebSockets for its real-time system due to scalability challenges with its Node.js/TypeScript backend and PostgreSQL database. Long polling, unlike short polling or WebSockets, keeps the connection open until data is available or a timeout is reached, offering a balance between immediacy and resource efficiency. This approach avoided complexities associated with WebSockets, such as adapting its observability stack, authentication mechanisms, and infrastructure configurations.

HOW AI-ASSISTED CODING WILL CHANGE SOFTWARE ENGINEERING: HARD TRUTHS

There are two developer types: "bootstrappers," who use AI for rapid prototyping, and "iterators," who use AI in daily workflows. While AI accelerates development a lot, it introduces challenges like the "70% problem," where achieving the final 30% of software quality requires real human expertise. As a result, AI makes the lives of experienced software engineers easier, but doesn't replace them.

REFLECTIONS

Sam Altman reflects on OpenAI's journey, the unexpected success of ChatGPT's launch, and the subsequent rapid growth. It was challenging and messy to build a company around groundbreaking technology, as shown by his own public firing and subsequent reinstatement. He's learned a lot and will be more confident moving forward.

free

LOGLAYER

LogLayer is a JavaScript logging library that provides a consistent logging experience across different logging providers. It offers a flexible API for adding tags, metadata, and errors to log messages, supporting multiple loggers simultaneously (e.g., DataDog and New Relic).

TAILWIND CSS MOTION

A Tailwind CSS plugin that lets you animate any element with inline styles. It also comes with a Chrome extension to let you play with animating your website directly.

CSS FLEXBOX PLAYGROUND

Experiment with different flex properties to understand how they affect layout. CSS code is generated in real time and controls are available for finetuned adjustments.

YOU WOULDN'T DOWNLOAD AN AI

This developer extracted an AI model from Microsoft's Seeing AI app. The app's model was found encrypted within the APK file as a TensorFlow Lite model. It was then verified using Netron, which led to the successful extraction of the AI model's weights and biases.

CAN LLMS WRITE BETTER CODE IF YOU KEEP ASKING THEM TO “WRITE BETTER CODE”?

An LLM was told to "write better code," which led to performance improvements in a Python program through multiple iterations. However, the improvements eventually plateaued, and the code became over-engineered. A second experiment used more precise prompt engineering, explicitly outlining optimization goals, which eventually led to a 100x speedup.

THANK YOU (2024 EDITION)

The CSS-Tricks website returned to consistent publishing in 2024, leading to 26 million unique pageviews, many significant site updates and content additions, and new goals.

WORLD'S FIRST MIDI SHELLCODE

This author gained remote code execution on a Yamaha PSR-E433 synthesizer via its MIDI interface by reverse engineering its firmware.

PRINTF DEBUGGING IS OK

Devs should take a pragmatic approach to debugging, and all techniques are valid (including just using printf) for efficiently finding and fixing bugs.

THE RECOIL REPOSITORY HAS BEEN ARCHIVED ON JAN 1, 2025. YOU HAD A GOOD RUN (REDDIT THREAD)

After both maintainers of Recoil, a React state management library created by Meta, were laid off, the repository was officially archived as no Meta engineers were willing to take up the work of maintaining it.

SCALING FROM A BILLION TO A MILLION TO ONE

Initially, this developer built an LLM monitoring tool using scalable technologies like Next.js, Vercel, and PlanetScale. However, the need for compliance and on-premise deployments forced a shift to AWS, requiring a simplification of the architecture and a reduction in resource demands. Ultimately, the need for a single-user, easily installable version led to further scaling down.

OPTIMIZING RUBY'S JSON, PART 4

The author, a recent maintainer of Ruby's `json` gem, made it the fastest JSON parser and generator in many benchmarks. These improvements weren't complex; they involved straightforward optimizations identified through profiling, showing that simple changes can yield substantial speed gains. The motivation stemmed from the widespread, albeit problematic, use of the faster-but-less-reliable `oj` gem, which caused compatibility issues and crashes due to its aggressive monkey patching and less robust coding practices.

FUTURE-PROOFING YOUR SOFTWARE ENGINEERING CAREER

LLMs help automate routine tasks, but these tools still require human oversight for complex system design, debugging, and understanding business needs. Engineers can future-proof their careers by focusing on areas where AI struggles, such as system architecture, domain expertise, and translating user needs into scalable solutions. Great engineers use LLMs to help them move faster and work more efficiently.

HOW I USE CLAUDE

Claude 3.6 led to this author using Claude for tasks like coding, advice, exploration, and writing. He breaks down his exact usage for each. For example, for code, Claude writes something for him 50% of the time and acts as a rubber ducky to throw ideas off of the other 50% of the time.

I AM RICH AND HAVE NO IDEA WHAT TO DO WITH MY LIFE

After selling his company for $60 million, the author (Loom's co-founder) found himself adrift, lacking purpose despite his wealth and freedom. He pursued various ventures, including exploring robotics and climbing a Himalayan peak. Currently in Hawaii studying physics, he's embracing a more humble approach, focusing on self-discovery and accepting ambiguity.

BENCHJS

BenchJS is a browser-based JavaScript benchmarking tool that allows users to write, run, compare, and share benchmarks using TypeScript or JavaScript, providing real-time metrics and collaborative features.

NITRO

Nitro is a cross-platform framework for building native modules in React Native using C++, Swift, or Kotlin. It provides statically compiled bindings for improved performance and type safety and is highly performant compared to alternatives.

REACT + AI STACK FOR 2025

The React + AI development stack for 2025 includes Next.js as a meta-framework, Tailwind CSS for styling, and Zustand for client-state management. Useful AI-powered tools include Visual Copilot, an AI-driven Figma plugin that converts designs into React code, and Cursor, an AI code editor. Bolt, a browser-based platform for building React apps from text descriptions, is also useful.

RULES FOR WRITING SOFTWARE TUTORIALS

Effective software tutorials have clear titles and introductions that explain the tutorial's goal and demonstrate the end result. Other recommendations include making code copy-pasteable, using long command-line flags, separating user-defined values from reusable logic, and keeping the code in a working state throughout the tutorial.

GENERATIVE AI IS NOT GOING TO BUILD YOUR ENGINEERING TEAM FOR YOU

Generative AI will not replace junior software engineers. While AI can generate code quickly, it cannot produce reliable, maintainable code, and the real challenge lies in managing complex software systems. Hiring junior engineers is an investment in the future.

THE MOST VISITED FRONT-END WEB AND MOBILE BLOG POSTS IN 2024

AWS' most popular blog posts in 2024 were about new AWS Amplify features, the Amplify AI Kit for building AI-powered apps, and AWS AppSync improvements.

THE 2025 AI ENGINEER READING LIST

A curated reading list of approximately 50 papers, models, and blog posts covering ten key areas in AI engineering, designed to be a practical resource for AI engineers of all levels.

IMPASSABLE CLOUDFLARE CHALLENGES ARE RUINING MY BROWSING EXPERIENCE (HACKER NEWS THREAD)

A user reports that Cloudflare's anti-bot measures are excessively blocking legitimate users, making large portions of the web inaccessible due to their heavily customized browser configurations and frequent VPN use.

ADVENT OF CODE 2024 IN PURE SQL

A database architect successfully completed Advent of Code 2024 using only SQL.

WHAT EVERY DEVELOPER SHOULD KNOW ABOUT OFFLINE-FIRST APPS

Offline-first app development prioritizes local data storage and synchronization with the server after the fact, so that data persists even without internet connectivity. This article goes into the various ways to implement offline-first architectures natively on the web using components like IndexedDB, the Cache API, and Service Workers. It also goes over strategies for conflict resolution, state management, error recovery, and testing.