Open Source

Claude Code Added GZIP in 2026. Why Not Brotli?

A developer captured a Claude Code API request that was 594,750 bytes . After GZIP compression, the body was 289,145 bytes . That's about 51% less d…

cutting N+1 queries in Next.js API routes with Supabase RPC

when a dashboard starts pulling relational data from Supabase, the naive approach is to fire a separate select per row. the result is a classic N+1…

Enforcing a concrete performance budget for Expo EAS builds and OTA updates

The situation Say you are a mobile engineer on a small React‑Native team that uses Expo Application Services (EAS) for CI‑driven builds and over‑the…

Écrire un article sur les Web Workers en JavaScript

Debouncing vs. Throttling in JavaScript: How to Choose JavaScript interfaces often receive events faster than they need to perform useful work. A se…

why pg‑bouncer can still choke your service under burst load

problem statement in a typical microservice, each instance opens its own pg client pool. during a traffic spike the aggregate connection count can e…

Context-Mode: Output Sandboxing for AI Agent Token Efficiency

AI agents are powerful, but their verbose tool outputs often bloat context windows, driving up API costs and hindering performance. Every time an ag…

Interrupted Runs Belong in the Denominator

A benchmark score that quietly drops interrupted runs is not a conservative score; it is a score whose hardest cases were deleted before anyone coun…

How much GPU energy does fp16 save? Measured on a Tesla T4

How much GPU energy does fp16 actually save over fp32? Spec sheets give TDP and peak FLOPS, not joules per job, so I measured it on a Tesla T4 using…

Exponential Backoff Isn't Enough: The Retry Storm Hiding in Your Recovery Path

The problem A dependency blips for 400ms — a deploy, a GC pause, a brief network partition, doesn't matter which. Every client with an in-flight req…

Architecture Patterns for Custom Direct Booking Engines & Hospitality Web Apps

When building web applications for the hospitality sector—ranging from luxury boutique hotels to private villa rentals—developers often face a choic…

Oxlint vs Biome vs ESLint in 2026: Benchmarked on a Real Next.js Codebase

I ran ESLint 9 and 10, Oxlint 1.88 and Biome 2.5 on my own site's Next.js code (403 files) and measured speed, memory and what each migration keeps…

CrUX Now Counts SPA Soft Navigations. 5 Things I'd Check Before Search Console Switches

For years, the Chrome UX Report (CrUX) has had a blind spot. If your site is a single-page app, Chrome only recorded the first "hard" page load. Eve…

AsyncLocalStorage vs Scope.REQUEST in NestJS: Fixing a Multi-Tenant Data Leak Without Killing Performance

A while ago in a high-scale multi-tenant B2B SaaS platform, we ran a high-concurrency benchmark before launching an enterprise data portal. Everythi…

Bevy instancing: drawing 639,000 units with one draw call per mesh

Originally published at tensorust.com . TL;DR: Bevy already instances meshes for you when entities share a mesh and a material. Past a few thousand…

PostgreSQL EXPLAIN ANALYZE: What I Actually Look At in Query Plans

PostgreSQL EXPLAIN ANALYZE: What I Actually Look At in Query Plans Every slow query I have ever chased started the same way: someone guessing. "Prob…

🔥 30 Days of Frontend — Day 5: React DevTools Profiler

When a React application starts feeling slow, one of the first things developers often do is reach for useMemo , useCallback , or React.memo . We've…

Keras #22561 speedups: compiled Keras on PyTorch now generates text 11.6 faster

Between master on Jul 18 and master on Oct 15, 2026, 15 of our PRs landed in Keras. Same GPU, same torch and jax versions, same models; only the Ker…

when tRPC beats REST for fine‑grained data fetching in Next.js

in a midsize SaaS product we hit a scaling wall when dozens of React components started querying overlapping slices of a user record. each component…

C = 1.25 is a tolerance, not a constant: the load-tail law behind MoE token dropping

Every sparse Mixture-of-Experts layer has a number in it that nobody derived. You need a static shape for the kernel, so each expert gets a budget c…

Why We Kept Ternary Bonsai 2 27B After Testing the Faster 9B

October 16, 2026 Our production classifier runs on a single RTX 2080 Ti with 11 GB of VRAM. It is often busy enough that requests spill over to a ho…

Measure the Escalation Rate. A Local-First Router That Never Earns Back Is Just Latency.

A local-first router only pays off if tier 1 absorbs most of your traffic. If your local model answers 20% of prompts and escalates the rest, you di…

A million-tap linear-phase FIR in real time: 7 ms of plumbing, 650 ms of filter

I'm building Sonir Bench, a macOS app that captures everything your Mac plays, runs it through a DSP chain you design, and sends it to your DAC. The…

Metadata filtering in vector search: post-filter loses results, measured

Multi-tenant RAG needs a filter: “only search documents this user may see.” Vector indexes are built for nearest neighbours, not for “nearest neighb…

The optimization that doesn't survive the process boundary

The optimization that doesn't survive the process boundary I found a slow lookup in our inbound-message handler — every incoming SMS triggered an O(…

Your Website Feels Slow? Start by Finding the Actual Bottleneck

A website can look perfectly fine during development and still feel slow when someone opens it on their phone or uses it over a weaker internet conn…

Everyone is doing Git on S3 wrong

HBM Explained: Why High-Bandwidth Memory Is Critical for AI Accelerators

Introduction: Why HBM Matters Artificial intelligence has changed the way computing systems are designed. In traditional servers, system performance…

I timed every free in-browser Whisper transcriber

Originally published on mastershot.app . Mastershot's audio to text runs OpenAI's Whisper in your browser, so your recording never leaves your compu…

Technical SEO for Developers: From URL Architecture to Rendering and Indexing

A sitemap points to one URL. The page declares another as canonical. Internal links keep reinforcing a third version. You can change the SEO plugin…

How I Use a $5.5 WebDAV Mount to Edit 4K Video Without Local Storage

How I Use a $5.5 WebDAV Mount to Edit 4K Video Without Local Storage Most video editors assume your media lives on a fast local drive. If you are ed…