IDK who needs to hear this, but you shouldn't buy promises of utopian futures with abundance and UBI from dudes who spend a considerable fraction of their time figuring out how not to pay their taxes.
Remote
Alex Russell
@slightlyoff@toot.cafe
mastodon 4.7.3Trying to make the web work for *everyone*.
https://infrequently.org/about-me/
Still not speaking for my employer, lo these many years.
8240 Followers
1619 Following
50 Posts
Joined April 07, 2017
Blog:
Currently:
Microsoft Edge
Previously:
Chrome, Project Fugu 🐡, PWAs, W3C TAG, Web Components, TC39 & ES6, Chrome Frame, Dojo
Also:
Making @phae@status.fberriman.com tea ☕
RE: https://infosec.exchange/@josephcox/117315761403151624
Normally it takes less time for FB to copy Google, but speed-running Duplex 8 years later is still fun to watch.
Quoting
New from 404 Media: Meta announced Muse, its AI agent that can book restaurants for you. We've seen internal Meta messages showing Meta is actually testing having these calls being made by human beings in call centers. In those tests its not AI, it's just a person
https://www.404media.co/meta-tests-muse-ai-agent-calls-that-are-actually-made-by-humans-in-a-call-center/
Open quoted postIf you make websites for a living, stop what you're doing and read this research by @tammy@webperf.social; it shows what many of us have been saying for a long time: even if there is such a thing as "fast enough",[1] it's generally *much* faster than you are targeting:
https://embrace.io/blog/research-core-web-vitals/
[1]: there isn't
Replying to
As a corollary, a useful guide for regulators to know if they're on the right track is the extent to which Apple and Google and Samsung and Qualcomm squeal, falsely claiming the consequence would be mass-scale insecurity.
The more they do, the more correct the policy is.
Our moral and political imagination must concieve and demand Elon Musk behind bars for the rest of his life.
RE: https://mastodon.social/@owa/117013437073165599
If you listen to @pluralistic@mamot.fr interviews, one of the things he repeats that resonates with me is that he's not an optimist; things don't default to getting better or worse, we *make them* better or worse.
We can engineer a fairer power arrangement, and folks like @owa@mastodon.social and @brucelawson@social.vivaldi.net are some of my heros, doing that work every day. Give their interview a listen:
A point excerpted from a talk I gave earlier this week at Microsoft's internal performance symposium (alt talk title: "CSS-in-JS: Frontend's Worst Idea"):
Browsers are virtual machines tuned to efficiently turn *markup* into *pixels*.
If your system tries to generate pixels from not-markup, you'll be working against the grain of the system, and should expect to suffer massive performance penalties -- both in CPU and memory use -- as a result.
For deep reasons, this is not going to change.
Replying to
The idea that users have a "choice" of brickware-from-day-one Androids, or brickware-from-date-of-revenue-collapse iPhones is a scandal, and the solution is legislation that legalises market forces to reassert themselves. That's not possible when hardware is held hostage to extractive OEM and carrier sales incentives.
Replying to
A competent board would have dropped Sundar a thousand times before letting Jeff and Sanjay walk.
Oh look, more React »»» Web Components == $$$ data from Shopify:
https://shopify.engineering/upgrading-checkout-blocks-app-to-polaris-web-components
Boosted by @joe@f.duriansoftware.com
So this keeps happening, and I think we need to talk about it:
> them: React manages state in our app.
>
> me: It does? Then what are for?
>
> them: Oh, those manage state.
https://infrequently.org/2026/07/state-management/
No One is on an absolute tear:
https://nooneshappy.com/article/kernel-anti-cheat-is-an-overreach/
It's very, very hard for me not give in to cynicism about the motives and priors of wealthy tech (and tech-adjacent) folks when the news that seems to move the commentariat to notice stagnant computing resources is that *Apple* is increasing prices.
In no calendar year across the past decade has Apple sold more than 20% of phones or more than 15% of desktop/laptop computers.
And yet.
1.) you should always take Heather seriously
2.) taking Heather seriously, this is a shocking indictment of Andoid and Google, and is 100% in-line with my critiques of DMA enforcement to date [1] and should be read by everyone as "Android, i. practice, is shockingly insecure and out of date":
https://www.wired.com/story/top-google-security-staff-warn-search-data-could-be-hacked-thanks-to-eu-plans/
[1]: https://infrequently.org/2026/04/the-web-is-an-antitrust-wedge/
Tired: this meeting should have been an email
Wired: this app should have been a web page
https://danq.me/2026/07/09/your-app-could-have-been-a-webpage/
TC39ers: the committee is working well and doesn't need fundamental reform.
me:
https://github.com/tc39/proposal-signals
https://github.com/tc39/proposal-decorators
mhmm...I see...yep. "Working" "well".
I'm not saying Safari is the new IE6; that would be unfair...to IE 6. When it came out, it was world-beating in nearly every regard, but got long in the tooth.
No, Safari and iOS are more like IE 7, 8, 9, 10, and 11: predictably disappointing showings thanks to structural under-funding by an organisation hellbent on holding the web back to defend profits:
https://infrequently.org/2026/07/abject-praise/
The framing and tone of this is rancid. The NYT is engaging in excuse-making exercises for Republicans who had, and continue to have, extraordinary agency. Instead of holding them to account or even enumerating their options, it launders inaction as fate.
Replying to
@willrs@social.vivaldi.net Thanks for the reminder about this recording! It's one of the sessions that I couldn't get to in person (was scheduled against something else; can't remember what). Great questions by Ben Kelly about ecosystem incentives; I've thought for a long time that we should expose to developers via reporting API more about when/why they get clobbered (e.g., for OOM/reclamation) or get de-res'd (smaller overdraw skirts, reduced color channels, image eviction, etc.)
Just as I'm planning to move residual devices away from SmartThings, I find out that they:
a.) have a web interface (!!!)[1]
b.) have an *advanced* version of that UI that does significantly more of what I want/need?
They could have blunted our full move to HA had they led with this: https://my.smartthings.com/advanced
[1]: https://my.smartthings.com/
Replying to
@willrs@social.vivaldi.net It won't surprise you, but what we're finding with partner teams at MSFT (often folks building on WebView2) is that an extremely large proportion of their app's memory use is down to legacy JavaScript-based UI patterns that the web platform has made obsolete. Loading less JS, moving back to "use the platform" is both faster, but also less memory intensive along many axes.
Replying to
@parismarx@mastodon.online The throughline is even worse: by failing to act against fraud and user abuse in the ads ecosystem, Google made a low-quality JPEG of the web more attractive than clicking on links. Search "answers" are adblocking with extra steps, and everyone loses because Google was so extractive for so long that what was left wasn't sustainable for anyone.
Replying to
@Girgias@phpc.social The recipe that I've seen work is some combination of:
- small rooms w/ customers/users to iterate on design early
- processes that let real code get feedback quickly w/o risk of burn-in
- wide review
- focused, objective ship gates
Last stage as formality is compatible with that, but we see challenges in both directions re: pace. Some groups too eager, some too timid.
Replying to
@Girgias@phpc.social We introduced the Stages process after the lurching, exhausting experience of ES6, which only got over the line because some of us pushed *very* hard to finally bloody ship.
At that point there was appetite in the room to entertain better processes, and Stages were a solution we (the Web Platform side of the Chrome team, not the V8 side) introduced. Things have gone roughly as I predicted it would, owing to explosion of veto players: some small stuff, very little big stuff, all overdue
Replying to
Contrast that with WICG at W3C where we have many proposals that fail (which is good!) and some which ship at the speed of interest.
TC39 has many proposals that sit in limbo ~forever. Always takes implementer interest to move something on the web, but when that spark catches, process shouldn't extinguish it with overwhelming reliability. Goal should be to nurture, find community, and get things going through coalitions of the willing.
Replying to
@john.kreitlow.rocks "whoa, dude... wouldn't it, like, be strange if this was also more efficient in every other way? We couldn't be that lucky, tho. But what if..."
Replying to
@pointlessone@status.pointless.one Apple primarily "absorbs" component price fluctuations with concentrated buying power and farsighted capital allocations, putting suppliers on notice about raising prices on them (lest they lose it on the other side). Long-duration contracts, etc. etc.
Yes, this is different, but Apple's still gonna be making 30% net margin on desktops and laptops. There was never a chance they'd be eating the cost because modern Apple is managed to the margin.
Replying to
@jaffathecake@mastodon.social @ash@icosahedron.website So the API itself doesn't need any ToS, at least not in other browsers. That's a Chrome product attribute that lots of folks will object to, but it's not part of the API any more than Web MIDI requiring an extension in FF is.
The deeper problem is model interoperability. No idea how we'll get close to that with an API that has no way to specify model or version.
Replying to
@ash@icosahedron.website @jaffathecake@mastodon.social I hear you. Trust is hard to earn and easy to burn, which is why it's so important we API OWNERS don't LGTM proposals that don't have good evidence behind them. I'm still trying to understand the pro/con arguments here. Civil negative feedback does matter.
Replying to
@lauren@mastodon.laurenweinstein.org ...and that's before discussing the layoffs and capital misallocation. Real "hat-trick of value destruction" stuff.
Replying to
@jyasskin@hachyderm.io @jaffathecake@mastodon.social In fairness to folks spotting the I2S late, I did too. I'm not sure how that cuts one way or the other on the substantive critiques, but it seems to me that not having +1s from API OWNERS outside Google when this was known to be controversial is at least a misread of the room.
Replying to
@jaffathecake@mastodon.social Firefox has ToS for features that interact with web content (e.g., autofill sync). The API surface here is pretty rough because it's a superposition: too general to be interoperable using different local models, and any solution to that involves unlicensable blobs/services. A real problem with this design for, but not bigger than the general interop issue.
Boosted by @joe@f.duriansoftware.com
No One's entire guide for classes of native app to avoid are is incredibly thorough, but nowhere is the logic of "app makers" more distilled than this section:
https://nooneshappy.com/guide/a-very-basic-framework-for-apps-to-avoid/#10-apps-that-exist-to-wrap-a-website
The goal of app stores is to rob you of the agency that interoperability and portability through open runtimes (browsers) give you. The gains from that trade are then largely deposited with data brokers and their clients, and no native app store is good about this. Indeed, if the were, they would be ghost towns.
Replying to
@carnage4life@mas.to Better question: how many more people will you need to fix what you didn't write, haven't tested, and don't understand?
Replying to
@AmeliasBrain@mstdn.ca I tried the sublingual drop therapy for a few years without much change, sadly.
Spot Apple's astroturf front group once and you can't help but find their enshittified fingerprints every. Here's Cupertino's favourite sock puppet -- "ACT | The App Association" -- arguing against right-to-repair for smartphones:
https://www.federalregister.gov/documents/2024/10/28/2024-24563/exemption-to-prohibition-on-circumvention-of-copyright-protection-systems-for-access-control#:~:text=Opponents%20ACT%20%7C%20The%20App%20Association%3B
/cc @owa@mastodon.social @pluralistic@mamot.fr
Replying to
@lauren@laurenweinstein.org The truly incredible part is that if ML progress had continued to be pushed forward at the pace of better medicines and energy efficiency breakthroughs, it would have been nothing but celebration.
And yet.
Replying to
@ash@icosahedron.website @jaffathecake@mastodon.social Perhaps you're confused. I'm the person (with standing) asking for the presses to stop on blink-dev. Civil, technically grounded concerns must be how we sort out Intents, and accusations like this don't help your case or mine.
Replying to
@Girgias@phpc.social Where standards intersection helps/hurts is "wide review". In theory TC39 is full of experts. In practice, it's full of people with reckons who go to meetings and can read/write ecmarkdown but who all exercise a veto stochastically on any topic.
Replying to
@ash@icosahedron.website @jaffathecake@mastodon.social "web standards" are misunderstood, in part, because trailing vendors have had a long history of trying to obscure the central role of voluntary adoption in Internet Standards. That's a problem that I've written about:
https://infrequently.org/2025/08/how-do-committees-fail-to-invent/
The firmer ground to argue from is that there doesn't appear to be a path to plausible interoperability.
Replying to
@jaffathecake@mastodon.social @jyasskin@hachyderm.io "abuse of position" doesn't assume good faith, and I think folks working inside Blink need to extend that assumption towards each other (until proven otherwise). That isn't Mozilla's job, of course, but heating the rhetoric after I've called the question on blink-dev may not achieve your goal.