Alex Russell
LightNews LightNews
banner
infrequently.org
Alex Russell
@infrequently.org
5.7K followers 1K following 1.6K posts
Trying to make a web that works for everyone.

Also at https://toot.cafe/@slightlyoff

For more: https://infrequently.org/about-me/

https://infrequently.org/
Posts Replies Media Videos
infrequently.org
This is most visible in the (sickening) plateau we're seeing in CWV pass rates. Despite progress in pipes and devices, more than half of pages don't feel minimally good on phones at least 1/4 of the time:
CWV pass rates at a few select data points over the past several years. We see a clear plateau below 50% in mobile pass rates over the past year.
November 24, 2025 at 12:19 AM
infrequently.org
My post on the the updated network/device landscape will be out next week, but until then, here's a chart of Geekbench 6 single-core CPU scores at 4 price bands over the past decade. 50+% of sales (> 75% in many countries) comes from the bottom two lines.

TL;DR: the privilege bubble rots brains.
the performance inequality gap is growing, not shrinking
November 24, 2025 at 12:10 AM
infrequently.org
How's the mobile CPU landscape evolving? Thanks to process node improvements, a mid-range Android from 2025 is (finally!) as fast as 2017's iPhone 8, but a low-end (~$100) smartphone isn't yet as speedy as a decade-old iPhone 6s.
Geekbench 6 single-core scores over time at 4 price points: - fastest iOS - fastest Android - mid-tier Android - low-end Android
November 21, 2025 at 10:51 PM
infrequently.org
*bad news

What'll really blow your mind is how cache-rich the top-tier parts are. I've been going on (and on and on) about this for years, but the "more cache, fewer cores" tradeoff has been ripe in Androidlandia for years, and is only just now being picked:
total SoC cache sizings for high/mid/low end parts in both the Android and iOS ecosystems. iPhones are fast in large part because Apple blesses them with ungodly amounts of L1+L2+L3+LLC cache.
November 18, 2025 at 9:30 AM
infrequently.org
I plotted that out too. The good news is that cheap phones are good value; the band news is...well...everything else.
Performance points per dollar for each cohort.
November 18, 2025 at 9:22 AM
infrequently.org
The "i" in "iPhone" stands for "inequality", late 2025 edition:
Worldwide smartphone shipments by year in millions; sources: IDC, Statista, and Counterpoint Research. At no point in the past decade have iPhones broken the 20% annual sales barrier. iOS is overrepresented in traffic statistics owing to wealth-correlated effects; e.g., faster networks and devices inducing heavier internet use. This does not imply that users on slower devices do not have money to spend, only that they are poorly served by the contemporary web. Average selling prices in nominal USD for iOS (~$1000) and Android devices (~$300) vs. worldwide smartphone Average Selling Price (~$360). iPhones aren't real life. Time series of single-core Geekbench 6 scores for 4 device classes: fastest iPhone, fastest Android, mid-range Android, and low-end device (~$100). The Performance Inequality Gap is growing, year on year, and the lower tiers are not even on the same trend line, let alone catching up.
November 18, 2025 at 3:51 AM
infrequently.org
only 89% to go:
Polymarket showing a 5% decline today in the odds that Schumer remains leader until the end of the year.
November 10, 2025 at 11:13 PM
infrequently.org
The ratios on this thread are *incredible*, and the comments are exactly what you think/hope they are:
Sen. Sheldon Whitehouse trying to talk down to Democrats and getting totally rekt in the comments.
November 10, 2025 at 11:07 PM
infrequently.org
No lies detected.
A link to join the web perf slack at performance.now()
October 31, 2025 at 1:44 PM
infrequently.org
Two paragraphs below, the expert witnesses Apple brought in gets handed their walking papers in devastating fashion:
"Professor Sweeting’s analysis failed to identify any causal connection between the restrictions and the value users might place on safety, security and privacy. He did not attempt to compare the effects of the restrictions on safety, security and privacy in the real world with a counterfactual world absent the restrictions264. The study he relied on was not specific to the UK and the questions it asked participants did not establish any causal link with the restrictions. Under cross examination, his analysis was exposed as superficial and not reflective of a realistic counterfactual world. Apple produced no other evidence to show what link there might be between user value and the restrictions. In our view Professor Sweeting’s analysis goes nowhere near the causal and quantitative specificity that is required in order to establish an efficiencies defence." p286
October 23, 2025 at 11:28 PM
infrequently.org
This tribunal ruling is *brutal* for Apple, citing not just the findings of the UK CMA's investigations, but also the evidentiary record developed by the EU's DMA.

Further, it destroys Apple's argument that the web apps platform it has neglected is a workable alternative.

/via @brucel.bsky.social
"779. We find that Apple’s restrictions do eliminate effective competition in those markets by removing all sources of competition. Accordingly, the efficiencies defence fails." p285
October 23, 2025 at 11:28 PM
infrequently.org
Well played, soft shark. Well played
Inflatable shark with a sign reading "I know a predator when I see one, release the Epstein files!"
October 18, 2025 at 11:14 PM
infrequently.org
oh hai, #nokings SF
October 18, 2025 at 10:16 PM
infrequently.org
So Apple is censoring another ICE-spotting app, and I think we need to talk about the coverage, as much as Apple's groveling pre-compliance.

@404media.co broke this, specifically @josephcox.bsky.social, and unlike some previous coverage, it includes this:

404media.co/apple-banned-an-….
"The website for Eyes Up which functions essentially the same way is still available. The site includes a map with dots that visitors can click on, which then plays a video from that location. Users are able to submit their own videos for inclusion. Mark said he manually reviews every video before it is uploaded to the service, to check its content and its location. "
October 9, 2025 at 12:27 AM
infrequently.org
boat cat thinking "I should make a Plex PWA"
October 8, 2025 at 4:52 PM
infrequently.org
We should expect a page to fully render on this link and CPU, even with redirect, in < 3 seconds.

I'm sure the business side is culpable, but did they do this? Inside the UK?

www.webpagetest.org/breakdown.ph...
1MB of JS, even on the sceptred isles
September 20, 2025 at 7:25 AM
infrequently.org
Seriously, LOOK at the trace. Just look at it:

webpagetest.org/video/compare.ph…

By the time the first render "finishes" at *6+ seconds* we are hundreds of KB of JS into a bad time. Nevermind that the page is *also* doubling up the text content as JSON inline thanks to Next.js:
the usual Next.js diarrhea of inline JSON that repeats what's already in the HTML for no useful purpose on a static page.
September 19, 2025 at 7:17 AM
infrequently.org
I would normally try to engage in constructive critique like "what was the goal here was?", but when an image map of the entire page would be faster, smaller, and more accessible...
the BBC News send 3MB of JS (nearly 12MB uncompressed) to render several kilobytes of text. The overall payload is monstrous.
September 19, 2025 at 3:24 AM
infrequently.org
It's not a big thing, but relatively frequent travel back and forth from the UK has reminded me just how utterly the BBC broke their site with JavaScript for most of the world.

This is BONKERS:

www.webpagetest.org/video/compar...
September 19, 2025 at 3:24 AM
infrequently.org
Is there a legal term for having such a weak hand that your defence is actually an indictment?

Asking for a monopolist:
Fresh off trying to kill web apps, Apple argued to the CMA that the App Store doesn't need to be regulated, because web apps: "the App Store competes strongly with the numerous alternative channels and transaction platforms through which iOS and iPadOS users can and do transact for digital goods and services. These include transactions on alternative devices (such as other mobile devices, PCs, game consoles, media player devices and smart TVs), web apps,"
September 11, 2025 at 11:47 AM
infrequently.org
Now boarding...Windows?
September 9, 2025 at 12:41 PM
infrequently.org
The area under these lines is a measurement of lost potential because Apple prevents real browser choice on iOS. An integral of fail; a decade of suck.

Join @open-web-advocacy.org to keep it from continuing:

webstatus.dev/stats?startD...
Features missing from the web due to a single engine, lower is better. Despite making 40x more from the web than Mozilla, Apple has consistently trailed in basic web standards compatibility.
September 1, 2025 at 6:22 AM
infrequently.org
can you show me where, on this chart, the billions of losses are?

m.macrotrends.net/stocks/chart...
August 24, 2025 at 5:31 PM
infrequently.org
Sedimentary loading screens, appearing one after the other, as I journey to the center of the earth where reservation details are etched in the superheated rock of the lower mantle.
BA's first loading screen, which is now the primary experience of ba.com Should you click the wrong button, you might be delighted by BA's _ancien_ loading spinner, indicating that you have passed out of your body and that the river Styx will be visible just beyond the next rise.
July 30, 2025 at 10:00 AM
infrequently.org
Absolutely no chance this will age like milk left in the sun. None whatsoever:

webpagetest.org/video/compare.ph…
An up-to-the-moment project structure that brings hundreds of MB of deps in tow. 70% of this _static web page_ is JavaScript, or 155K compressed/505K unzipped
July 26, 2025 at 10:35 PM