Martin Michalek

Martin Michalek

@machal.bsky.social

Web Performance at PageSpeed.ONE FrontKon Conference. Prague based. 🇨🇿

246 followers239 following183 posts

Martin Michalek reposted

If ads are slowing down your site, DevTools can now identify the scripts responsible → goo.gle/4i2xbRL In Chrome DevTools, open Application > Ads to check experimental CrUX ad metrics (including CPU and Network usage), browse a table of ad scripts, and highlight every ad on the page.

Bonus: If you haven’t visited our website in a while, take a look—we’ve been working hard on it over the past few weeks. There’s plenty of new stuff to explore. → pagespeed.one/en/?utm_sour... Keep your websites fast! #WebPerformance #WebPerf #PageSpeed #PageSpeedONE

pagespeed.oneWebsite Speed Monitoring | PageSpeed.ONEA slow website harms conversions and SEO. PageSpeed.ONE monitoring alerts you to issues before they impact your metrics. Try it for free.

#FrontKon was 10 days ago, and I still don’t know how to sum it up. 650 developers, almost 30 talks. Actually, I don’t even know where to start. It was quite painful this year, but it ultimately led to a powerful, positive catharsis. Eventually, the right words will come. 😅

Webdesign these days. Two modals. And another one below them, blurred out. Just so I know it’s not going to be easy for me! Meetup dot com, love you! 😅

Working on website speed? Help me, please. I’m putting together small study on why it’s sometimes difficult to get faster websites approved I’m interested in your real-world experiences. It takes 3–5 minutes to complete. Every response helps. 🙏 pagespeed.one/en/perfnow #WebPerf #PageSpeed

pagespeed.oneSurvey: How do you get web performance prioritized? | PageSpeed.ONEA voluntary survey on how web performance gets prioritized at work. For people from performance.now() and from elsewhere.

Martin Michalek reposted

Why is speed so hard to sell? Martin will look at where familiar arguments fail, what WebPerf can learn from SEO, and why ownership and incentives matter as much as metrics. You will leave with practical tips to turn performance into prioritized work. @machal.bsky.social at perfnow.nl

A five-minute speech I spent two weeks preparing. Because putting together a 30-minute talk isn't a problem. But preparing a short, punchy, funny intro for your own conference... and keeping it under five minutes? That’s a nightmare! 😅 So, hopefully, Mr. Kout and I pulled it off at #FrontKon.

I'm working on my English. You know… I'm old, no talent for languages… 😅 I've tried everything possible. One thing really works for me – @italkioffical.bsky.social with live teachers, with iTalki Plus, AI lessons based on data from live lessons. What a great idea and execution!

I've continued my investigation into how "Delay JS execution" fakes Lighthouse scores. New piece on my personal blog: – which WP plugins ship it (WP Rocket, LiteSpeed Delayed, Perfmatters…) – what their own docs admit – how to spot it in the source. michalek.blog/blog/delay-a...

michalek.blogDelaying all JavaScript is not optimisation, it just moves the problem · Martin MichálekOne checkbox in a caching plugin can lift a Lighthouse score by tens of points without making a site faster for anyone. Here is which WordPress plugins ship that checkbox, what their own documentation...

Martin Michalek reposted

#AI is turning the developer world upside down. Discussion is essential, and talks offer only one perspective. That’s why we’re adding two panel discussions to #FrontKon: 🤖 Agentic development in practice 👩‍💻 The present and future of work in IT Looking forward! www.frontkon.tech

An interesting move. As part of the #CrUX data suite, Google will also measure technical and UX metrics related to ads. Finally, an effort to decipher just how much ads weigh down websites (including Google’s own).

Software development is in the (AI) crisis. Questions coming into my mind: 1) When LLMs write code, what does it mean to be the developer? 2) Should devs write code? 3) Are devs becoming (agents) managers? 4) What is happening with focus, deep work and learning? Do you see another questions?

View full thread

Multiple peaks can reveal distinct groups of user experiences. They may point to different technical conditions—for example, cached vs. uncached visits. One metric value can’t show you these patterns. A histogram can.

The 75th percentile is useful—but it hides part of the story. If LCP p75 is 2.10 seconds, 25% of user experiences are slower than that. The histogram shows the users behind the number.

Martin Michalek reposted

We turned off a plugin that was "speeding up" a client's site. — Lighthouse: 95 → 60 — Core Web Vitals: unchanged It wasn't optimising the site. It was optimising the test. We've found and documented third case of "hacking" Lighthouse score. My article: pagespeed.one/en/blog/ligh...

pagespeed.oneThe Illusion of Fast Websites: Gaming the Lighthouse Score | PageSpeed.ONEPlugins like WP-Optimize, WP Rocket, or Website Speedy can boost your Lighthouse score by deferring JavaScript. However, they do not enhance user speed – and sometimes, even testing tools catch on.

We turned off a plugin that was "speeding up" a client's site. — Lighthouse: 95 → 60 — Core Web Vitals: unchanged It wasn't optimising the site. It was optimising the test. We've found and documented third case of "hacking" Lighthouse score. My article: pagespeed.one/en/blog/ligh...

pagespeed.oneThe Illusion of Fast Websites: Gaming the Lighthouse Score | PageSpeed.ONEPlugins like WP-Optimize, WP Rocket, or Website Speedy can boost your Lighthouse score by deferring JavaScript. However, they do not enhance user speed – and sometimes, even testing tools catch on.

Unpopular opinion: Think about Elon Musk what you want, but… His apps are getting really good in last months. Looking at X (comparing with Bluesky) or Cursor.

We know at least three well-known optimization plugins that game the Lighthouse score. Site owners think their website got faster. It didn't. Only the number did. Real users see zero improvement. Come on. Don't buy this. Measure your site speed with data from real users. #WebPerf #CoreWebVitals