best part is this was mostly written by claude - it loves this approach to declarative reactivity because it just sits there and writes basic JS without a thought or care in the world

I've now got a 15 page, 200+ component app that is heavy on data munging, editing, adding, updating, importing all centered around gpx route mapping - all the data is stored locally in cache and the entire app uses only 9 explicit signals (tracked properties), 8 of which are for cursor hover state.

and because I buried the lead, essentially the only thing you change in your app is `@tracked` to `@field`, and ime you pretty much can delete 50% or more of associated signals you were probably using before to juggle syncing state from multiple sources.

Whether a resource is tied to session-storage, local-storage, cache-storage or your API is just config.

And anything can be turned into a resource now, including services, components, generic state objects...

a key thing this unlocks: easily building new features with locally stored/modeled data when you don't have an API yet (or will never have an API) with the same simple patterns you use for everything else.

TL;DR signals as sync engines. Turns out to work pretty darn well. Signals might be the fundamental unit of sync.

I've been exploring a bunch of new ideas for WarpDrive. Stuff that is in the space of where I thought the project should go but haven't talked about publicly much. And well... not only has it been a delight to use... it has me thinking there is a key bridge missing in WD between sync engines and it.

So I've been kinda quiet on WarpDrive lately. Some of it is the way AI has shifted where my work is needed internally, some of it is I'm helping to put on a 100 mile race in June - and there's a lot of work for that, but a lot of it is that in my limited time for OSS I've been deep on something 👀

Dunno - Claude seems to output reactive code naturally and easily. Have had to wade through very little slop with it - had expected to be fighting imperative biases and haven’t been

There's a lot of ceremonial stuff that bigger companies do that's just an absurd time-waste, but yearly goals is a contender for worst. "predict what will matter to the company in the next 12 months, be super aligned and specific, btw the only time anyone will ever see this is you when you write it"

Bam saw Kuminga drop 27 on the wizards and realized what that really meant.

Gui Santos is busy proving all the reasons it was time to move on from Kuminga

Only if by drama you mean people that haven’t watched a game of basketball. Good riddance.

Has he Played ok against a couple of bad teams lately? Sure, but he’s getting extra minutes with starters sidelined and still only putting up 3rd option numbers as the top scorer on the floor, and is now predictably “injured” again. The guy misses games if he passes gas.

Honestly getting annoyed with how diplomatically polite Kerr and Dunleavy are about Kuminga.

He was shit, and a liability. He’s still shit and a liability. Yeah, he is good when he does everything he’s supposed to: the problem is he never ever does everything he’s supposed to.

Bay Area is either on the third false spring or it’s actually summer already. Won’t know for a month 😂

the reason I find this interesting is I don't feel like there would have been a ton of examples of this in its training set - as its decidedly not how humans approach problems on average.

one of the things I find the most interesting about LLM code output for complex things like codemod work is the ease with which it uses recursion to simplify/achieve complex tasks in a small number of lines. Stuff we know we should do but have a hard time reasoning through and thus don't.

More sites need to offer a manual uninstall or refresh button. Also it’s crazy how hard browsers make it to find and remove them

The answer is easily melon

Games like today are why I continue to believe the warriors are closer to a championship than not. Before the Butler injury I even felt like it was maybe this year. They have the pieces.

Next year with the cap space they have is gonna get interesting.

Some folks worry what this does to their career. Personally I find it’s letting me get back to being more creative and more effective and do a lot of the things that originally turned me into an engineer in the first place but for which I never had enough time to do.

And importantly - validation of ideas matters a lot less when it’s the person with the need producing the feature. A whole industry exists because folks couldn’t solve their own problems and needed others to figure out how to understand their needs - and that just matters less now.

Except it was.

The cost of producing bespoke software was too high even when the needs were well understood and organizational overhead was low. Even higher when needs needed to be translated. Multiple layers of human and time costs have vanished here.

This is really similar to another point I’d make about LLM assisted coding’s effects on velocity: I’m watching tons of non-technical folks build software solutions for their small business needs.

I think small businesses ultimate benefit the most from AI coding. “The code was not the bottleneck”

Companies still spending tons of time deciding what to try instead of trying lots of things quickly and iterating on what’s working will fail

Historically, spending a ton of time on research and planning was beneficial because the cost of building the wrong thing was so high. That cost is no longer high - the calculus has changed.

The enormous, depressing, overwhelming, unconquerable human costs of convincing folks what to try vanish when it becomes easy to just do or show. Code velocity changes organizational velocity significantly.

I disagree with this take.

I watch product teams spend tons of time trying to validate ideas that turn out wrong. Or products fail because the cost of prototyping ideas is so high that most ideas are left untested.

Product velocity absolutely can be the hardest part of it all… and that’s changing

Got a catalogue? I’m curious mostly because I’ve never hit one of these and so now I want to know what features I’m not using but should be 🫣