Did you forget to mention the work on AsyncContext, or just too early to mention?
The Bloomberg contributions in the performance WG are much appreciated!
10.03.2026 00:23
👍 3
🔁 0
💬 1
📌 0
Did you forget to mention the work on AsyncContext, or just too early to mention?
The Bloomberg contributions in the performance WG are much appreciated!
Now you have the opportunity to sort the keys alphabetically :D
Honestly that's an impressively thorough job of destruction! And you know now they are addicted to it...
Slow clap, from one dad to another.
I don't get it... Is this uncommon?
Yeah I think 2026 will be an exciting opportunity.
I think individual sites or nimble RUM tools will be able to move fastest here, but we're slowly trying to gather data for CrUX. Not sure how much interest there is for that?
One caveat-- those are averages across a large range of sites, some of which are popular but only minimally use the soft navs pattern, not full spa.
Many sites-- esp sticky sites-- will have higher ratios. But these are a minority of the overall.
My recording from @perfnow.nl is now up on YouTube: www.youtube.com/watch?v=sDHS...
I was a bit nervous, and very tired, but I think it went okay.
Japanese tourism sign says "please refrain from any acts of nuisance" and my whole life philosophy is shattered...
I couldn't find a metric for "turn off js" but I'm sure its trivial.
More useful would be "desired JS didn't have a chance to run" but that's much harder to define. We won't have single perfect number for that.
But I'll see if I can get some insights on Monday.
Ah neat. This answers the question I had last night.
Interesting stuff this signal graph...
Another wonderful @perfnow.nl conference has flown by.
Love this community, and I return energised with ideas and motivation.
Well, mostly energised...
As we already have framework app templates and create-blank-app in code editors, we will have whole framework-version-specific context hooks as well.
My 2cents: It's not really what LLMs could do, its: why might developers choose to specifically constrain LLMs to a specific tech stack?
As we evolve to develop new APIs or new idioms, there is a bootstrapping & training problem, no doubt, but this is where agents/context engineering plays a role.
Sure, if I just ask a chatbot to hallucinate a demo it might pick a random popular option-- but I could also frame my requirements.
I think I disagree with the premise here.
I think that LLMS only need *sufficient* context to become good at writing in particular style, not necessarily *the most& context.
Like humans, LLMS can become expert developers at many frameworks regardless of which is most popular in the training set.
<activity/> is sweet.
Seems there are two use cases in particular (right from the docs)
- prerendering hidden new content
- Keep rendering hidden old content
Pre-re-rendering you might say :)
React 19.2 update and the react conf content from yesterday has been great.
I enjoy seeing the references to INP on guiding a lot of the performance and scheduling/coordination work, and the nice integration with modern web features.
Exciting stuff every year.
Omg genius
(Ah. Astro+VT used to be coupled with ClientRouter, but afaik has been de-coupled. I assumed you mentioned cross-document-VT, which are also now supported.)
HTML swap I think is default, but you can opt to persist nodes (component islands, or dom nodes with state). The default might change...
@danabra.mov I've been enjoying your recent posts on RSC.
Re the astro post: overreacted.io/rsc-for-astr... I don't see mention of <ClientRouter> which both changes some of comparison, and, I think, validates the RSC model.
docs.astro.build/en/guides/vi...
Have you looked at that? Very htmx-ish.
m.xkcd.com does some of these things already. Unfortunate that you don't get redirected, automatically.
Naming aside, that presentation was exceptional.
It put into clear terms patterns we were struggling to articulate.
A wild Michal appears!
Your last question in this episode, with an explicit call to "be specific", was about the nicest "I call bullshit" I've ever heard ;)
Darned. Sorry to hear that. You are the second person this week who I've heard this happen to.
Where are these slopes?
That's not dodgy. You should see me try overhangs!
That's actually genius! Until you forget to turn it off while on a roaming hotspot :)
But now you have to undo on error or if user changes their mind...
Or maybe optimistic + route changes are a bad idea and you should only use it for simpler things?
What happens when you mix:
- transition to a new state
- route change at new state
- useOptimistic (eager state update)
In most of the guidance for transitions they recommended integrating route changes after successful render, not at start.
I think useOptimistic means eager URL update?
Ricky, I've been left in <Suspense>.
I could really use() an answer! Could you kindly route() me, so I can start to Transition?
I eagerly await your Promise to resolve my query. Please don't reject.
(Why yes, I am dad joke levels of vibe today :)
I, too, often feel this way when measuring event timing ;)