Robin Marx

Network protocols (HTTP/2, HTTP/3, QUIC) and Web performance @Akamai. PhD. Dad. Longsword fencer. He/Him.

@programmingart.bsky.social

[Robin Marx @programmingart]

In case you want to know more about why #HTTP3 is needed and how it works, I have an in-depth article series @smashingmag : https://t.co/Xgf7sqCu50

[Robin Marx @programmingart]

@KeenGamesStudio @Hadbatter bzzzzz bz bzzzz

[Øredev @oredev]

Meet the stars ⭐ lighting up the Web track!
🔷 @andersnoras
🔷 @Horusiath
🔷 @drcircuit
🔷 @helvira_g
🔷 @InesAkrap
🔷 Kristoffer Fredriksson
🔷 Lucky Nkosi
🔷 @ythecombinator
🔷 @patrickbrosset
🔷 @philnash
🔷 @programmingart
🔷 @sergesdev
👉 https://t.co/vzzO8Rwjpz

[christina 死神 @chhopsky]

It's pretty simple.

AI has destroyed the internet and all trust in the people and contents of it. You can no longer trust anything that has been submitted to you on it nor that you find on your own. There is a pre-nuclear steel cutoff point at which the age of a post has become an element in its trustworthiness. That destruction has taken out the work of creating those things and removed humans from the process because greed told business owners they could do it cheaper this way but it just made everything way worse.

And because it kind of sucks but does a passable job if you squint, it's replaced a tonne of juniors while still requiring seniors to oversee it, thus interrupting the flow of juniors who could become seniors and choking off the natural lifecycle of human progression required to have continued service. When the seniors retire we will not have new ones to replace them and their capabilities will be lost forever.

The ability to make kinda shitty versions of stuff and carry on conversations has automated theft, fraud, harassment, and political manipulation on a scale previously unimaginable. Computers that were once deterministic will now lie to you with the confidence of a con-man. The greed and manipulative accounting of shifting revenue around has destroyed consumer computer markets and people can no longer afford what they need to work, or play, once again driving people back into the arms of business owners who can't wait to rent it to you for a subscription. And because investor eyes light up at the new Flavor Of The Month (like they did with blockchain before it) it's being jammed in everywhere it doesn't belong and now every product you liked is worse.

Search doesn't even work anymore and attempting to use any of the AI chatbots for this removes the element of context from your ability to determine trustworthiness of a source, because there isn't one. It's a remix of a remix of a remix, reconstituted from parts, and the average it makes from the slop could be true or it could be full of shit, which is way worse than having to evaluate multiple real human points of view to construct your own. We are taking the averages of averages away from the contexts within which the data points were made in the first place then wondering why everything is slightly wrong now.

And to achieve all this, unaccountable tech billionaires have stolen generations of work from creatives, technicals, and anyone else with a "who will stop us" attitude, making it clear to everyone that the rule of law is over because there's so much profit available it's in their best interests to just ignore it. They know this now, and they will keep doing it forever. We have rewarded the very worst behavior from the very worst people and what little power the individuals had has been transferred back to more corporations.

AI is a race to make everything worse for everyone while a handful get rich, and if you have to stick your fingers in your ears to pretend you just 'dont get why people dont like it' I have more respect for the people doing that evil on purpose than I do for you.

[Patrick Meenan @patmeenan]

It has been a LONG time coming but @sergeyche @programmingart and @yoavweiss weren't falling for my ongoing procrastination.

A test tool (and generator) for compression dictionary deployments: https://t.co/lzJCgq7Saz

[Common Logic Games @CLGBluntBSE]

The Matter of Being is now live on Kickstarter!

Set in the world of #cultistsimulator, it plays like Suzerain if, instead of being president, you were a rebellious spirit on the run from ancient gods.

Check out our campaign @ https://t.co/fuUHM19tnh

[Henri Helvetica 🇭🇹 👨🏾‍🚀 🚀 @HenriHelvetica]

📢 Tomorrow will be H2 protocol's anniversary. With multiplexing, updated header compression and more, H/2 was a timely arrival. To discuss the arrival of H2, we are joined by none other than @programmingart
expert in web protocols!
📆 May 14TH
⏰ 2PM EDT
https://t.co/0VxpiDN1wE

[Shaw @shawmakesmagic]

💀💀💀

[DiscussingFilm @DiscussingFilm]

Apple has landed the rights to turn ‘MISTBORN’ into a film franchise & ‘THE STORMLIGHT ARCHIVE’ into a TV series.

Brandon Sanderson will write, produce and consult on all projects.

(Source: https://t.co/Ka6RvxmT3S)

[Robin Marx @programmingart]

TTFB is a notoriously difficult performance metric, because it has so many different definitions and nuances.

New features like Early Hints and Speculation Rules only make that worse.

My new blog highlights these issues so you know what to keep in mind:
https://t.co/PqLvio1GcG

[Stoyan Stefanov 🪩 @stoyanstefanov]

Web Performance Calendar day 24 with @programmingart on TTFB. Yes, it's one of those X-mas Eves where we cuddle by the fire, drop everything and immerse into re-learning and re-evaluating what we thought we knew about a concept we use daily
https://t.co/H8kX9oppmN

[Stoyan Stefanov 🪩 @stoyanstefanov]

Web Performance Calendae day 13 with Yoav Weiss (@yoavweiss) showing us a clever and creative way to ship *only* critical/necessary CSS for each page type (home, search, etc) *and* also avoid duplicating code in CSS bundles.

https://t.co/N5eI5BALBR

[Mirru 💎 @mirrutatep]

The reason genAI is cheap and unlimited right now is because it is heavily subsidized by blind and reckless investing…

As a business venture it is vastly unprofitable compared to the energy it costs and once the bubble pops the landscape will change

[Robert Graham @robertgraham]

You don't need to understand the OSI Model to work in cyber...

...because nobody in cyber really understands the OSI Model anyway. They've all heard of it. They can all describe parts of it they think they understand. But everyone knows they are hazy on other parts.

The infographic below is wrong, actually. Any graphic like this that you see is wrong. You can verify this yourself by searching for "OSI Model" on Twitter and see whether everyone's pictures of the 7 layers actually match each other. They don't.

In other words, you know they are wrong not because you can find the correct one somewhere (you can't), but simply because they all disagree with each other.

The reason is that OSI was created to for how mainframe networking worked in the late 1970s, to create an official set of standards. The Ethernet and TCP/IP Internet didn't conform to that model, but became the unofficial, de facto standards anyway.

Ever since they've been trying to "retcon" the OSI Model, to reconcile it with Ethernet and the Internet. "Retcon" means "retroactive continuity", like when the second Star Wars movie says that Vader is Luke's father, when in the first movie, it said Vader killed Luke's father.

The same with OSI Model. Instead of paying attention to what the original model actually was, they attempt to redefine it to fit the Ethernet and Internet that we have now.

But there's no standard for this, so everyone's retcon version is slightly different, so no two sources describe the model the same.

Meaning, nobody really knows that the model is.

It's kinda freaky. Everyone is hazy on the details, especially the upper layers. But they are certain somebody, somewhere, has a handle on this, a expert that truly knows the details. There isn't. At least, there's me, telling you it's all wrong. But there's no expert that can challenge me on this, such as somebody who defends the above infographic as being correct.

[Seth Harp @sethharpesq]

Calling it now: There will be no consequences for the creators of the AI bubble that is about to take down the US economy. The main culprits will be made whole by the taxpayer, and the gullible fools in media who promoted the bogus tech will remain in cushy positions of influence

[🔛🔝 @killingherass]

retweet if you actually just fucking hate AI

[karachi memon @reembeem97]

That’s actually the perfect use of generational wealth

[Tom Taylor @TomTaylorMade]

I hate AI with a passion.
It’s an unregulated, unconsidered, human-ingenuity-wrecking, resource-sucking, climate-destroying theft-machine being pushed by short-sighted, greedy individuals putting product and profit over people and planet.
But it turns out… I can hate it more.

[Andrew Healey @healeycodes]

wrote the same word-counting program 5 times, each faster than the last

best result: 494× faster than my first Python version — by using SIMD in C

all are O(n), but some squeeze far more from the CPU and memory!

[DebugBear @DebugBear]

Sometimes servers don't send HTTP response data in the right order.

Our new article explains why request priority matters for HTTP/2 and HTTP/3 connections, and how poor server-side support can impact page speed.

https://t.co/I4yXdq2VJw

[Robin Marx @programmingart]

We're live right now, come join us!

[Henri Helvetica 🇭🇹 👨🏾‍🚀 🚀 @HenriHelvetica]

📢 Excited to have @TimVereecke join me Wednesday to talk shop about 2025 and details from a recent blog post from he and @programmingart re: performance in 2025. Join us!
✨ SPDY STREAM 003 ✨
📆 Wed July 9th
⏰ 1p EST
🔗 https://t.co/Sb4JXIEA6q
🔗 blog post: https://t.co/ixw0iX97ik