Vercel Speed Insights now measures soft navigations
Client-side route changes in single-page apps now show up in Core Web Vitals, not just full page loads, though soft-navigation events add to your Speed Insights Plus bill.
Tagged
Everyone knows to eager load. The queries that actually take sites down are the ones hiding in an accessor, a Blade partial, or a policy. There is a one-line switch that finds all of them.
Most slow queries are one index away from being fast. The reason people do not add it is that nobody ever explained what the index is actually doing.
Transformers.js and WebGPU put real models on the client. The demo is easy; knowing which features belong there is the part worth thinking about.
Every performance sprint I have been on made the site fast for about four months. The teams whose sites stayed fast did something different, and it was not technical.
WordPress will happily let you write a query that scans a million meta rows. Here is which arguments are expensive, why, and what to do instead.
Client-side route changes in single-page apps now show up in Core Web Vitals, not just full page loads, though soft-navigation events add to your Speed Insights Plus bill.
The inference server's new default KV cache mode grows context capacity by half on some GPUs, but adds up to 18% decode latency.
The Vercel Toolbar automatically flags layout shifts on a deployment, names the elements involved, and lets you replay the shift to see what moved.
ngp, a beta compiler frontend built on the Rust toolkit Oxc, replaces Angular's reliance on a TypeScript API that version 7.0 removed. Early benchmarks show cold builds two to three times faster and memory cut nearly in half.
A new setting stops large IN (subquery) lookups from dying with an out-of-memory error. The release also speeds up GROUP BY on LowCardinality columns by 30%.
UX CSS compiles design tokens straight to class names, validated at compile time. LiveComponent updates also cut dashboard render times by up to 46% in the project's own benchmarks.
coolify-rs is a project of mine: one Rust binary in place of Coolify’s PHP, Laravel, Horizon and Redis, serving the same REST API against the same Postgres database.
A rewritten JSON API ships as a normal standard-library package in Go 1.27, with stricter decoding and unmarshalling the Go team says is up to 10x faster.
The terminal's latest release finishes its Kitty graphics support, adds an SSH wrapper, and patches ten security bugs including one letting a clicked link run arbitrary commands on macOS.
TorchTPU, built by Google with Meta's PyTorch team, drops the torch_xla dependency so PyTorch code runs on TPUs unmodified. Hugging Face reports decoding up to 21.5 times faster on some models.
An unread response body used to force a fresh TCP connection on every request. Go's net/http.Transport now drains it automatically, and the Go team's own benchmark shows requests running nearly five times faster.
Real User Monitoring now reports the phases behind each Core Web Vital, so you can see whether a slow LCP is a server, network or render problem.
d1-3B and d1-omni-600M are open-weight "decision models" that skip text generation entirely, returning a calibrated answer in one forward pass in as little as 8ms.
The 'use cache' model closes its remaining performance gaps, Turbopack's disk cache shrinks by a fifth, and the Rust React Compiler cuts memory use by 30%.
The Rust-based dataframe library adds out-of-core spilling, first-class SQL support and a native Map type, and silently drops row-order guarantees unless you opt back in.
Chrome 155 decodes .jxl images without a flag, using a new Rust-based decoder. Firefox is due to follow on 13 October, and Safari has supported the format since 2023.
A block of JSON tells the browser which links to prerender. On a server-rendered site it closes most of the perceived gap with a single-page app.