iiinova — where your digital self keeps evolving

hfcRed @hfcred.dev · 21/09/2026

Yes ofc, most people who do webdev do, it's the industry standard. If you look at Stack Overflow and State of JS surveys, around 80% of people use VSCode for webdev

020

hfcRed @hfcred.dev · 21/09/2026

VSCode basically holds all the "market share" for web dev

130

hfcRed @hfcred.dev · 08/09/2026

I love CSS too much to not do it by hand. And from my experience LLMs only care about the visual result, and not the path to get there, so they often end up writing very rigid and brittle CSS that breaks under non ideal conditions. I still have to go in and fix most of the styling by hand.

061

hfcRed @hfcred.dev · 07/09/2026

There should be a PR or issues by @scottwu.ca to update the docs to mention that async Svelte supersedes the await block.

130

hfcRed @hfcred.dev · 07/09/2026

{#await} is soft deprecated/replaced by async Svelte. Use the await keyword directly in the markup instead svelte.dev/docs/svelte/...

121

hfcRed @hfcred.dev · 07/09/2026

The Svelte Discord server has a Bsky feed that shows all posts mentioning Svelte so that's why you sometimes got ppl dropping in to help. I've been keeping up with this thread and it seems like a lot of these are XY problems. Are you using async Svelte yet? Do you have any reproductions?

110

hfcRed @hfcred.dev · 05/09/2026

That's why this tech is also just stupid. More realistic just means more bland lighting, because real life is bland. You can see in the video how it instantly sucks out all color and brightness. It's also heavily biased towards faces, that's why every demo focuses on that.

010

hfcRed @hfcred.dev · 05/09/2026

Ironically it does less for S1 because DLSS5 banks on physical based rendering being used. It's only purpose is to make already realistic looking stuff more realistic. S1 is too flat for there to be anything to "enhance".

110

hfcRed @hfcred.dev · 04/09/2026

My example still holds in that case, you can just do $state(await functionToGetTranslation()); and then plug that into the setContext. You can also drop the await, that's how you get a promise into the context, but I personally don't see the point

000

hfcRed @hfcred.dev · 04/09/2026

You can setContext an async value but I would honestly just do this instead svelte.dev/playground/b...

120

hfcRed @hfcred.dev · 03/09/2026

Super cool well done

010

hfcRed @hfcred.dev · 01/09/2026

No problem! One thing to know about reactive classes is that they can handle reactivity for the consumer. For example in this playground: svelte.dev/playground/4... you can see that I never use a $derived() anywhere, but the function call in the markup still re-runs when the count changes.

140

hfcRed @hfcred.dev · 01/09/2026

No measurable impact no. It's just a normal .ts file that you can use reactivity in, reactive classes are a very good pattern to follow. You can check out some of the util classes from Runed for examples of reactive classes github.com/svecosystem/...

github.com

runed/packages/runed/src/lib/utilities at main · svecosystem/runed

Magical utilities for your Svelte applications. Contribute to svecosystem/runed development by creating an account on GitHub.

110

hfcRed @hfcred.dev · 01/09/2026

If you own the class then you can just use reactivity inside of the class in a .svelte.ts file. If you do not own the class just build a wrapper class for it that maps the internals to a reactive class.

130

Reposted by hfcRed

barinky @barinky.net · 22/08/2026

@hfcred.dev does insane work, the #picoCAD2 Web Viewer is rad. -> picocad2-web-viewer.hfcred.workers.dev you can render at higher resolutions + tweak a bunch of effects and settings. recently updated to support beta branch of P2. #lowpoly #cyberpunk

227952

hfcRed @hfcred.dev · 18/08/2026

Could you give some examples?

000

hfcRed @hfcred.dev · 10/08/2026

It's actually a ~51% winrate 😭

000

hfcRed @hfcred.dev · 10/08/2026

Two ways to do Dialogs. Singleton where you have only one instance at the root of your app that you can call from anywhere: svelte.dev/playground/e... Or an instantiable component: svelte.dev/playground/3...

110

hfcRed @hfcred.dev · 09/08/2026

The lesson here is that a lot of people need to work on their mental. Instead of blaming a weapon for your losses, look and learn from your mistakes, that's the best way to win against 52 in the future.

040

hfcRed @hfcred.dev · 09/08/2026

That doesn't mean 52 isn't a problem, it still needs to be nerfed as you can see on the right side of the graph. A nerf wouldn't even impact it below that 0.1%, the winrate would remain the same on the left side, it would just be flattened more on the right side, aka it would be more balanced.

120

hfcRed @hfcred.dev · 09/08/2026

People will always put the blame on something. If 52 wasn't around, a different weapon would take it's place as the scapegoat. These stats look even cleaner for other queues. Outside of X Rank, 52 has an almost 50% flat winrate, and a slightly above average pickrate.

120

hfcRed @hfcred.dev · 09/08/2026

People in that group of 0.1% of players are right to complain about 52, however that is not most people. Most people who blame 52 on their losses are just looking to put the blame on something so they don't have to confront themselves with the possibility that they just made mistakes.

120

hfcRed @hfcred.dev · 09/08/2026

52 Gal winrate and pickrate by XP (100 XP steps). I put this together because I'm getting tired of ppl blaming 52 for their losses, when in reality it has no impact on 99.9% of players. 52 hate is mostly a mix of selection and confirmation bias, which causes people to overestimate it's impact. 🧵

2224

hfcRed @hfcred.dev · 09/08/2026

Basically they dont want to use file based routing and are upset that they can't use code based routing

000

hfcRed @hfcred.dev · 08/08/2026

Additionally you avoid binding and needing an extra variable just to hold the DOM reference.

010

hfcRed @hfcred.dev · 08/08/2026

Attachments offer encapsulation. Code inside an onMount or $effect could be anything, so you have to read all the code just to puzzle together all the relations. With an attachment you have everything in one place, and it becomes apparent what exactly it does. And it's reusable.

110

hfcRed @hfcred.dev · 08/08/2026

That's true tho it depends on how extensive the feature as a whole is

000

hfcRed @hfcred.dev · 08/08/2026

Ideally the parent would handle this yeah. Components should stay "stupid". Their primary concern is just rendering out the data they receive, not transforming or managing data themselves. This might also be a good usecase for a reactive class in a .svelte.ts file.

111

hfcRed @hfcred.dev · 08/08/2026

In general $derived or explicit function calls in the same code block that changes the state cover most usecases. $effect should be reserved as a last ditch effort. If you like you can share some code or create a REPL on svelte.dev/playground/h...

000

hfcRed @hfcred.dev · 08/08/2026

I dont see why you would need effect here even when no re-mount happens

100

hfcRed @hfcred.dev · 14/07/2026

Total X Power share by weapon per season. Now that blasters are weaker, 52 gets to pick up a lot of momentum. Right now 340/800 weapons across the X leaderboards are occupied by 52 gal, with X Power shares still trending upwards steeply.

0102

hfcRed @hfcred.dev · 17/06/2026

I actually implemented it because I saw Ranys post here and thought it was a cool idea

030

hfcRed @hfcred.dev · 18/04/2026

Right bsky.app/profile/hfcr...

041

hfcRed @hfcred.dev · 30/03/2026

It can comfortably render 4000 meshes with 300000 polys on my phone at 60fps so should be fine. Only one badge will be rendered at a time, and there will be model size restrictions in place.

110

hfcRed @hfcred.dev · 30/03/2026

Idea for a new badge system where the actual models will be rendered in real time, fully interactable. Model by @brushmommy.bsky.social

1535

hfcRed @hfcred.dev · 21/03/2026

Its in the global settings now

110

hfcRed @hfcred.dev · 19/03/2026

And typescript lol its really nice for updating stuff like nested generic intersect types and then resolving the resulting type errors

020

hfcRed @hfcred.dev · 19/03/2026

So the skill and knowledge is proven to be there, but LLMs can often save you a lot of time when writing schema #381 or similar.

120

hfcRed @hfcred.dev · 19/03/2026

You're fine, it is still a correct observation. Claude is used during development sometimes for small/tedious tasks, but we both double check eachothers commits regardless. I think it's important to note that the majority of the site was built long before LLMs.

220

hfcRed @hfcred.dev · 19/03/2026

Core features and migrations are all hand made, LLMs are sometimes brought in for tedious or minor tasks with strict supervision. We also did experiment with some agentic workflows, but have since removed them all again as they did not produce the results that measured up to our standards.

061

hfcRed @hfcred.dev · 19/03/2026

This is false, the majority of it was hand written, and at no point was Claude a requirement for making the changes possible. Both Sendou and I possess the skills required to achieve the new feature set. Experimenting with Claude is not the same as relying on it.

2113

Reposted by hfcRed

Sendou @sendou.ink · 19/03/2026

Biggest update to sendou ink ever is now live! New layout, friends system, SQified tournament subs, site-wide chat, new mobile UI and so much more... Huge shout-out to @hfcred.dev for all the work on this with me. Wouldn't be possible without their efforts. #sendouink

713733

hfcRed @hfcred.dev · 19/03/2026

These changes took 4 months, with 462 committed changes, adding 35.761 lines of code, spanning 830 files. It was a huge effort, so I hope it delivers!

0170

hfcRed @hfcred.dev · 18/03/2026

High res hfcRed jumpscare

001

hfcRed @hfcred.dev · 13/03/2026

We are trying to distance ourselves from Discord considering the recent events around it

131

hfcRed @hfcred.dev · 10/03/2026

This is photoshoped

100