Just learned that @mu.social lets you edit your Bluesky posts. Since Bluesky hasn’t implemented this, you cannot see if a post has been edited if you are using the Bluesky client.

I did really like the idea of eurosky, but then I went there and their website screamed "GO AWAY".

Evelyn@kandake.africa · 3d ago

SHOTS FIRED, I REPEAT SHOTS FIRED

As of today, the use of AI becomes more transparent in the EU. Here’s how EU law makes sure AI works for you, in full openness ↓

I really hate this too. Podcasts are amazing exactly because they are not attention grabbing clickbait.

David Bushell 🪿@dbushell.com · 3d ago

cold opens are the worst trend in podcasting random talking with no context. I look forward to the vague deja vu I get when it's repeated in 40 minutes 🙄

I think they actually do, but the question has to be asked right. Would you like to have one place where you change your address, and every service just automatically updates? Would you like to be able to easily see every website that knows you address, email, real name?

dame@dame.is · 6d ago

hard pill to swallow for atproto developers/creators: 99% of potential users either do not care about, value, understand, or want to “own their own data” i see so many apps/projects lead with this “value prop”, but it’s not an important or meaningful consideration for most people

I was going to write spist about this being why @nordcraft.com AI harness use retained reasoning and compaction, but I am pretty sure every AI harness does today.

Dare Obasanjo@carnage4life.bsky.social · 6d ago

OpenAI recently showed how tweaking the harness used in a benchmark made GPT Sol 5.6 go from 13.3% to 38.3%. This is a good example of the value of harnesses especially as models commoditize. Getting 3x the performance from using the right harness will likely beat any difference in models.

Film you have watched more than six times with a gif. Hard mode 2025 only

Kpop Demon Hunters Rumi

Alt: Kpop Demon Hunters Rumi

static.klipy.com

Isaac Corbrey@isaaccorbrey.com · 6d ago

film you’ve watched more than six times with a gif. hard mode: no star trek, star wars, lotr, marvel, disney, or ghibli.

I think Sol is much too slow. If we want to make coding faster, this is the most important thing to solve

dax@thdxr.com · 7d ago

idk what size sol is but it's the biggest i want rn

The fact that this happened because of how words are translated to tokens is completely irrelevant. The example is completely valid, It showed people that while LLMs are being sold as super intelligent systems, they can failing the most basic tasks that even a small child could solve.

Liz Fong-Jones (方禮真)@lizthegrey.com · last wk.

it's fascinating how quickly the r's in strawberry discourse dropped off as soon as people just fucking conceded, no, you don't ask LLMs to think in letters, only tokens.

What particularly hit me about “build better gates” is that it seems none of them believed that before AI

Jason Gorman@jasongorman.bsky.social · 2w ago

I hope the "Don't bother reviewing the generated code, just build better quality gates" folks have got good indemnity insurance.

I would go one step further. The promise of “own your data” is meaningless without permissions.

rich harris@rich-harris.dev · 2w ago

the lack of permissioned data is already killing a lot of stuff. there are so many apps that you just can't really build at all right now (without creating your own back end, at which point the value proposition of atproto is massively reduced)

It is really simple: Traditional web is like shopping at the super market. At proto is a farmers market. If you still don’t get it, maybe it is you 🤷‍♂️

𝕮@chrisshank.com · 3w ago

Here's my two minute explainer of atproto: it is in essence a water cycle. - Clouds hold water. - They precipitate water over mountains. - Water funnels down mountains into rivers. - Rivers branch off into lakes. - Water evaporates back into clouds, the cycle repeats.

If you make websites for a living, stop what you're doing and read this research by @tammyeverts.com; it shows what many of us have been saying for a long time: even if there is such a thing as "fast enough", it's generally *much* faster than you are targeting: embrace.io/blog/research-co…

New research: The Core Web Vitals thresholds you trust might be wrong for your site

Core Web Vitals data from ten leading online retailers reveals that "good" for Google may not be good enough for your users or your business.

embrace.io

Stef Walter@stephaniewalter.bsky.social · 3w ago

An now, for my next trick, we will answer the CSS questions everyone had for years: how to center horizontally. Like, really, 100 different ways. Some are quite hacky, but, still, fun to explore just to test the limits of CSS, hehe Tool made by @css-only.dev css-generators.com/center/

What is mu.social? I am positing from it now, but it seems like it is just pink bluesky

I have never built a website than handled different writing modes, but surely it is not this simple? Don’t you always have to adjust every part of you layout in that scenario?

Victor@vpon.me · 3w ago

CSS tip 💡 Instead of writing: padding-left: 24px; padding-right: 24px; padding-top: 16px; padding-bottom: 16px; You can use padding-inline: 24px; padding-block: 16px; or just padding: 16px 24px; (top/bottom, left/right)

Dont stop using js for staggered animations. Sibling-index() is not baseline yet

Lionel Péramo@lionel-peramo.com · 4w ago

Stop using JS for staggered animations. Modern CSS has native functions that handle the indexing logic for you. With sibling-index() and sibling-count(), elements now know their exact position and the total count of siblings natively. #CSS #WebDev #Animation (1/3)

LLMs are not colleagues or tools. They are adversaries in never ending game of sabotage. If they sabotage you, they win, if you stop them you win. They don’t try to sabotage you all the time, then you would just stop using them. They are helpful 90% of the time so you lower your guard.

Andy Bell@bell.bz · 4w ago

I absolutely agree that this is one of the most exciting features in recent years

Josh W. Comeau@joshwcomeau.com · last mo.

Just published a blog post about the new CSS Anchor Positioning API! ⚓ I’ve been learning about it over the past few weeks, and it’s friggin’ great, but it also took me a minute to really get the hang of it. My goal in this post is to share the most useful bits, so you can start using it today. 🔥

One of the strangest things about working with LLMs is that it is so easy to forget that they are not actually capable of reasoning. Luckily they frequently remind you.

"AI-assisted writing helps so many people write better". I have heard this line so many times. I find it deeply troubling. Most of the time people us AI to write, because they are lazy or simply dont value the readers time, but I think this is different.

Alex Suzuki@alexsuzuki.com · last mo.

"You're so arrogant to think AI-assisted writing is beneath you, it helps so many people write better". No. YOU are arrogant for assuming that *my time* isn't worth *your effort* of putting your own thoughts into words.

Bluesky is a beautiful idea with an absolutely terrible execution. The promise of - an open protocol - owning your own data - choosing your algorithm Is great, but I never got the idea that the Bluesky team ever believed in any of it.

Erica Joy Astrella@ericajoy.astrel.la · last mo.

negging is…definitely a strategy. when people start coming over here with fighting energy, just know this is why. (link for veracity: www.threads.com/@bluesky_soc...)

They should keep doing that, it seems to be working. Qwen 3.7 was a big step up.

Bruce Lawson@brucel.bsky.social · last mo.

The company that used the books I co-wrote to train its model without permission is weeping that China "brazenly" and "illicitly" extracted its Claude AI model's capabilities. I'm sending snorts and stares. www.bbc.co.uk/news/article...

I agree with the basic sentiment, but in reality it is very impractical. Coding agents do lots of different tasks: research, planning, implementation and validation. Each has different levels of complexity. Research requires less intelligence than planning etc.

dax@thdxr.com · last mo.

i have low conviction on model routers - very open to changing my mind but this is a snapshot of my current thoughts - i don't think it's good to not be aware of what model you're using. coding with LLMs is a skill you develop and getting a feel for models is part of that 1/4

This is by far the most annoying Bluesky bug. Along with the one where being blocked by one account means you suddenly can’t see the rest of the conversation you were part of.

This is immensely frustrating Anthropic has shown again and again that they are not a well functioning software team. We literally got to read their terrible code, and somehow people are still acting like Anthropic has something intelligent to say about the future of software development.

Jason Gorman@jasongorman.bsky.social · last mo.

Why is anybody taking advice from Anthropic on software engineering? They're very evidently pretty bad at it.

I'm porting Quake to 3D CSS, so it runs in the browser without WebGL. It's open source 👹 Play it! cssquake.com Repo: github.com/LayoutitStud...

a11y people. This is incorrect right? I don't think an <aside> should never contain a <h1>.

I've made a Kubernetes that runs in the browser, and today we're open sourcing it. The hope here is to make super interactive and visual web content about Kubernetes much easier to write, and we have a whole bunch planned this year. Repo: github.com/ngrok/webern... Demo: webernetes-demo.ngrok.app