Search the archive

that helps your team deliver its best work. Register now

HOW WE IMPROVED AVAILABILITY THROUGH ITERATIVE SIMPLIFICATION

GitHub goes over its monitoring and observability stack, having implemented a comprehensive approach to improving availability through iterative simplification. The company utilizes a suite of tools, including Datadog, Splunk, Scientist, and Flipper, to monitor performance, analyze data, experiment with code changes, and control the rollout of new features. With these, GitHub has been able to identify and resolve performance issues proactively.

COMPONENT COMPOSITION IS GREAT BTW

Conditional rendering within JSX can lead to complex logic and cognitive overload, especially when dealing with multiple states. Using early returns is a great alternative to conditional rendering, which improves code readability, maintainability, and type inference.

REARCHITECTING: REDIS TO SQLITE

Wafris, an open-source web application firewall company, transitioned its Rails middleware client from using Redis to SQLite as the backing datastore. This was because Redis deployments had high network latency and poor performance in cloud environments. SQLite offered a speed advantage, especially in the "hot path" of rule evaluation, which involves frequent reads of large datasets.

THE PYTHON PACKAGE INDEX SHOULD GET RID OF ITS TRAINING WHEELS

The Python Package Index (PyPI) of prebuilt binaries has grown exponentially, leading to massive storage and bandwidth costs. This author proposes a three-pronged solution: 1) leverage advancements in systems programming to create repeatable builds, allowing for easier reconstruction of binaries, 2) make prebuilt binaries optional, reducing PyPI's storage burden, and 3) establish secondary caches for popular packages to offload bandwidth pressure.

THE DAILY SCRUM

The daily scrum meeting feels like micromanagement and creates unnecessary pressure on developers. Since software development is inherently non-linear, and progress can be slow or non-existent for days, making daily check-ins feel like a constant judgment.

WHY I STILL BLOG AFTER 15 YEARS

This developer initially started his blog to chronicle his game development journey but has since expanded its scope to cover his interests, including programming, 3D printing, and writing. He finds that writing helps him clarify his thoughts and improve his communication skills.

Schedule a demo to learn more

LLAMA 3.2: REVOLUTIONIZING EDGE AI AND VISION WITH OPEN, CUSTOMIZABLE MODELS

Meta's Llama 3.2 features smaller vision and text-only models for edge and mobile devices. These models are great for tasks like summarization, instruction following, and image understanding, even outperforming closed models like Claude 3 Haiku.

UNITOOLS

Unitools is a suite of tools to simplify code sharing across Next.js and Expo (React Native) projects. It takes a framework-based approach, providing a unified API within a base library that abstracts platform-specific details, while also offering platform-specific libraries for compatibility and performance optimization. Unitools currently has packages for links, images, routers, and font plugins.

AI SDK 3.4

Vercel's AI SDK 3.4 introduces new features including language model middleware, data stream protocol, structured outputs, multi-step generations, tracing improvements, mock models for testing, and provider updates.

WHERE DEVELOPERS FEEL AI CODING TOOLS ARE WORKING—AND WHERE THEY'RE MISSING THE MARK

While most developers are using AI coding tools, their trust in the accuracy of these tools is not universal. Developers see benefits in increased productivity, faster learning, and improved efficiency, but they express doubts about AI's ability to handle complex tasks.

CONWAY'S LAW: THE HIDDEN FORCE SHAPING YOUR SOFTWARE ARCHITECTURE

Conway's Law states that software architecture often reflects the communication structure of the organization that created it. This means teams tend to create software components aligned with their own structure, even if a different architecture would be more efficient. Engineering managers and developers should focus on designing their teams well, which inherently leads to better software architecture.

AI TOOLS FOR SOFTWARE ENGINEERS, BUT WITHOUT THE HYPE

In this podcast, Simon Willison (co-creator of Django) shares his experience using LLMs to increase productivity, dispelling common misconceptions about their use in software engineering and offering practical tips and tricks to use them effectively.

MONITOR YOUR LOCAL AND REAL-USER CORE WEB VITALS PERFORMANCE IN DEVTOOLS

Chrome DevTools now provides real-time local Core Web Vitals performance metrics alongside real-user data.

WINAMP

Winamp, a multimedia player, has been open-sourced.

MIRA MURATI EXITS OPENAI

After six and a half years, Mira Murati, the CTO of OpenAI, is leaving the company.

AVOIDING IF-ELSE HELL: THE FUNCTIONAL STYLE

This article teaches how to refactor complex, nested if-else statements in code using functional programming techniques like guard clauses, decision tables, and function composition.

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.

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.

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.

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.

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.

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.”

and update with the user's onboarding state. Read the blog

UPDATED PRODUCTION-READY GEMINI MODELS, REDUCED 1.5 PRO PRICING, INCREASED RATE LIMITS, AND MORE

Google has released two updated production-ready Gemini models: Gemini-1.5-Pro-002 and Gemini-1.5-Flash-002. These models offer significant improvements in performance, including faster output, lower latency, and increased rate limits. The pricing for Gemini-1.5-Pro has been reduced by 50% for prompts under 128K tokens.

ANNOUNCING TYPEDSQL: MAKE YOUR RAW SQL QUERIES TYPE-SAFE WITH PRISMA ORM

Prisma ORM v5.19.0 introduces TypedSQL, a new feature that brings type safety to raw SQL queries. It allows developers to use raw SQL for complex or performance-critical queries while maintaining the same type-safe environment of their workflow.

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.

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.

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.

GOOGLE CACHE IS FULLY DEAD

Google has disabled its cache feature, which was previously used to access older versions of websites - users can now only access archived websites through the Wayback Machine.

NANO-SPAWN

Nano-spawn is a tiny, dependency-free process execution library that offers a promise-based interface and various features for managing subprocesses.

A SHORTCUT TO EDIT LONG SHELL COMMANDS IN YOUR $EDITOR

Using `ctrl + x` followed by `ctrl + e` in bash or adding a keybinding for `edit-command-line` in zsh allows you to edit long shell commands in your preferred editor.

GOOGLE WINS CHALLENGE AGAINST $1.7 BILLION EU ANTITRUST FINE

Google won a legal challenge against a €1.49 billion ($1.66 billion) antitrust fine from the European Union.

Discover the joys of Stytch 😍

WHEN POSTGRES INDEXING WENT WRONG

When a concurrent Postgres index creation silently failed, the app that used the database faced performance degradation, almost leading to a production outage. This failure was exacerbated by the use of partitioned tables, resulting in inconsistent index usage across different partitions. To avoid such issues, monitor concurrent index creation, validate indexes regularly, and create partition indexes consistently.

IMPLEMENTING FEATURE FLAGGING WITH THE NEXT.JS APP ROUTER

Use environment variables to control feature visibility, validate them, and then access them within server and client components using custom functions and a feature provider. This blog post explains how to implement basic feature flagging with the Next.js App Router. It demonstrates a practical application of feature flagging by controlling the visibility of sidebar routes based on feature flags.

OPTIMIZING FOR HIGH LATENCY ENVIRONMENTS

Chrome's User Experience Report provides Round-Trip-Time (RTT) data, which reveals the network conditions of users. Many users have slower network connections, so web applications should be able to handle high-latency environments well. Some strategies to handle these include reducing transfer size, using a CDN, upgrading to HTTP/2 and TLS 1.3, and adopting HTTP/3 (QUIC) if possible.

THE INTELLIGENCE AGE

Sam Altman describes a new “Intelligence Age” driven by new AI advancements. This new era promises massive improvements in various aspects of life, including healthcare, education, and even solving global problems like climate change. While AI's potential for prosperity is immense, there is still a need to navigate risks, like those related to labor markets.

SERIOUSLY, STOP USING RSA

RSA is a widely used public-key cryptosystem which is arguably fundamentally flawed and should be abandoned. Its security relies on carefully chosen parameters, which developers often fail to select properly, leading to vulnerabilities. These vulnerabilities stem from subtle mathematical properties that average developers are unlikely to grasp, making RSA inherently fragile and prone to attacks.

MY LEARNINGS FROM 7 FAILED TECH INTERVIEWS

Zach Wilson, an ex-Airbnb staff data engineer, goes through seven tech interviews he failed over the last decade. He shares learnings from his time. For example, job hopping early in your career is great for growth, but not as great after you hit your mid-career.

[13] or get a free taster on Youtube

TS-BLANK-SPACE

ts-blank-space is a TypeScript-to-JavaScript compiler that focuses on speed by removing type annotations and replacing them with whitespace. It utilizes the original TypeScript parser and is written in pure TypeScript.

SIDEKICK

Sidekick is a tool that simplifies the deployment applications on VPS', making the process of hosting side projects faster and more affordable by automating the setup and deployment process. It enables zero-downtime deployments and provides features like high availability, load balancing, and automatic SSL certificates.

MICROJS

A library of JavaScript packages with small footprints.