Amila Welihinda

"Best technical deep dive I have read in a while!" -- Principle Web Performance Eng @ Microsoft If you're building a large, JS heavy web app, you might find production JS profiling interesting, check it out. palette.dev/blog/improving…

A lot of great ships at @NotionHQ in 2025 that I was proud of. This one was not flashy, not AI, but the most meaningful to me. Pages now load in 1/3 less time (1.9s → 1.3s). It took a lot of patience, craft, and god-forbid regressions to move a big metric like this for all our users. Investing in eng foundations is something we'll continue in 2026, as everything else is built upon it.

@magicpatterns Absolutely insane, keep up the great work!!

We've raised a $6M Series A to help product teams go from idea to production. PMs and designers use us every day to design their next feature. Today, we’re also announcing our new brand look and Magic Patterns 2.0

Seen at my local gym

Browser soft nav is the slowest web experience that's underserved by web vitals. We've been working on a new interaction performance metric at Palette that measures soft nav latency by measuring the delta from interaction to LCP. DM me if you're interested in trying this out!

This is React Fiber:

Regarding precision, we get the paint latency directly from the browser, which is far more accurate that JS approximation methods for measuring paint latency.

Identify lines of code blocking React component renders. Here's how you'd see the lines of code blocking click latency for the TabButton component interaction latency for end users. Palette does this by attaching end user JS profiles to React component renders.

Introducing Palette React Performance Metrics Automatically measure the latency of React components in your app It's like INP for React components

optimizing palette's distributed profiling is fun on each request we must: 1. load 50MB of source maps 2. ast traversal across source of each source map 3. run sql queries to select and group by across ~40B rows 4. compress and stream resulting 20MB JSON profile payload finish under 10s

at Palette, traces for every backend request (db queries, IO, api calls) are shown in Chrome Devtools using server-timing headers. would love to see out of the box next.js support for this

In just one month at Palette we: 1. built out SSO support 2. rewrote our core profile aggregation service in rust for significant perf wins 3. onboarded 3 amazing enterprise customers Couldn't be more excited for what's next!

@awelm_ @LionelAccount @nextjs side by side comparison w claude code would be v interesting

@awelm_ @LionelAccount @nextjs you should record a cli demo walkthrough of this

GPT-5 is cracked at frontend. It fully migrated my blog awelm.com from Hugo to @nextjs with a few prompts. It also added these new features:

@samlambert neon, clickhouse for comparison

people are asking what is unique about the PlanetScale Postgres product. here's one thing:

what a glow up after 1yr left is our end user JS profiling mvp called "hot functions", sorting functions by total time. right is today -- complete flamegraph supporting zoom/pan/filter

Atlassian redirecting your browser 10 times before seeing a product is one of the worst experiences of enterprise apps

we want to make the web apps you use every day faster. to do that, we decided lower the barrier to production JS profiling so any web dev could have Facebook-level perf observability x.com/patmeenan/stat…

Interesting to see the services to make self-profiling easier become available beyond what the big companies have implemented in-house. Some of these browser capabilities have years before you start seeing adoption/ROI but deliver huge impact when it happens.