TLDR Dev 2025-01-30
React Native Audio API 🔊, 0-downtime DB upgrades 🧑💻️, Sublime Text in 2025 💻
UX++ and DX++ with Sync Engines (9 minute read)
Managing remote state in web web apps is complex. Current methods, while improving developer workflow, create unnecessary loading states, stale data issues, and complexities around optimistic UI updates. A better solution is to use sync engines, such as Replicache, that provide a local data copy, allowing for instant UI updates, offline capability, and efficient data synchronization via a push/pull model similar to Git.
A Major Postgres Upgrade with Zero Downtime (18 minute read)
Instant, a Firebase alternative, successfully performed a zero-downtime major Postgres upgrade from version 13 to 16. Initial attempts using in-place upgrades and blue-green deployments failed due to downtime constraints and incompatibility with active replication slots. Instant used a manual process that involved creating a new Postgres 16 replica from scratch, carefully switching subscriptions, and writing a custom zero-downtime algorithm, which paused new transactions, waited for active ones to complete, and then switched writes to the new instance.
Next.js: an honest review (8 minute read)
Next.js 14 is fast and developer-friendly, but the App Router introduces complexity with use server, use client, and serialization quirks. While Vercel makes deployment easy and offers great DX, middleware limitations and accidental server code bundling can lead to unexpected issues.
Why I still like Sublime Text in 2025 (20 minute read)
The author, a long-time Sublime Text user since 2010, explains why they continue to prefer it in 2025. The main reasons include its speed and efficiency, great support for Language Server Protocol (LSP) features, and a powerful snippet system with scope-based functionality. It also has good built-in project workspace management that allows for customized settings, build systems, and file exclusion.
Why Did You Render (GitHub Repo)
This is a React library that helps developers identify unnecessary component re-renders by monkey-patching React to log potential optimization opportunities. It supports React Native and offers various configuration options for tracking components and hooks.
Mathesar (GitHub Repo)
Mathesar is an open-source, self-hosted web application that provides a spreadsheet-like interface for interacting with PostgreSQL databases. It allows for viewing, editing, querying, and collaborative work on data without needing to write code.
Introducing React Native Audio API (7 minute read)
The react-native-audio-api has been released. It brings Web Audio API functionality to React Native, with precise audio control and high-performance playback across iOS, Android, browsers, and desktop applications. The tool implements a modular routing system where audio flows from source nodes (like files or generators) through processing nodes (for effects) to output nodes (speakers/hardware).
Why DeepSeek had to be open-source (and why it won't defeat OpenAI) (5 minute read)
DeepSeek open-sourced its reasoning model R1, rivaling OpenAI's o1 in performance but at a fraction of the cost due to innovative training methods. This open-source approach, counterintuitive to standard business practices, was necessary for DeepSeek to overcome Western skepticism towards a Chinese AI company and gain market access. The increasing commoditization of LLMs that have similar performance across various models makes the higher cost of proprietary models like OpenAI's less appealing.
On DeepSeek and Export Controls (18 minute read)
The advancements of the Chinese AI company DeepSeek, while impressive, do not weaken the case for US export controls on chips to China. There are three dynamics in AI development: scaling laws, shifting the curve (cost reduction), and paradigm shifts in training methods. DeepSeek's progress is largely consistent with expected cost reductions and doesn't represent a fundamental breakthrough, but rather shows that strong export controls remain crucial for determining whether the future of AI will be unipolar (US-led) or bipolar (US and China).
Effective AI code suggestions: less is more (6 minute read)
An AI-powered code review tool initially attempted to prioritize code suggestions by severity (bugs, style, etc.), but this led to an overload of low-priority suggestions, drowning out important findings. A shift to a "laser focus" approach, instructing the model to only identify critical bugs and problems, improved results. This resulted in a 50% increase in suggestion acceptance rates and an 11% increase in overall impact.
The most important software engineering news in one daily email
Join 450,000 readers for
one daily email