The AI Great Leap Forward In 1958, Mao ordered every village to produce steel. The steel was useless. The crops rotted. Today's top-down AI mandates are producing the same pattern: ba...

This is a fun read leehanchung.github.io/blogs/2026/0...

A package.json with the "prepare" script "compromise-the-supply-chain".

A gentle reminder that while `pnpm install` (or `bun install`) doesn't run the lifecycle scripts of the dependencies by default, it *does* run them from the repo's own package.json.

Let's be mindful when cloning and exploring all those new & exciting projects on our local machines.

All my homies hate `prepare`

We don't need to speed up production. We need to slow down.

We don't need to write code faster. We need to think through what we write.

We've spent decades working to ship faster, and that's landed us in a world of hurt.

Build better products, not faster.

For what it's worth, that's fairly easy to do in any modern, component-driven framework even without AI (and I've accidentally done it myself).

It's a tradeoff of the component model, you gain maintainability but lose context (for both good an bad, reducing context can be helpful at times).

Can someone wake me up when people want to build software for people again? These numbers aren't surprising given the general consensus that software seems to work less and be more error prone across the board.

Shoutout all those dorks who said AI would “fix accessibility”

I mean that seriously -- people were slinging just as much AI slop in December, nothing has changed in the field in the past couple of years. To say it's exploded is incredibly odd given the complete lack of anything noteworthy in the last 3 months.

Nothing’s changed in the last 3 months, I have no idea where this is coming from. Very much the status quo, though there is increasing fatigue.

this is turning out to be one of the onion's all-time classics

Whilst I realize it's probably not everyone's preference, I really enjoy seeing what friends are working on & starring themselves, it's way more fun of a homepage with that.

After remembering the 'Feed' page exists on GitHub, I wrote myself a little user script to swap out the home page for it and I'm quite happy with that.

After months of using it, I find the new home page to be rather bleak; it's a never-ending list of (old) PRs and issues that only really builds.

It does not work on the 'Dynamic Layout' page, entire line highlights only.

Thanks for the report! Vite started seeing this as side-effectful at some point, quick fix:

github.com/preactjs/pre...

Well that’s interesting indeed, will take a look. Dunno what’d cause Vite to fail to shake there.

You can use the `prerenderScript` option to point at any other location, thereby reducing how much (if any) lands in the common entry point.

But indeed, somewhat limited by single build pass.

preset-vite isn't meant to be injecting anything (by default) outside of dev mode, I think it's pretty unlikely to be an issue w/ it but if you can provide a reproduction then we can confirm

Tiny typo in Jovi's name in the "Sources" section, it's "De Croock" rather than "Decroock".

p75 and p90 JavaScript bytes on the web have exploded since May, 2025—easily the largest 10 month period based on HTTP Archive data since 2023.

The long-tail is getting worse, very quickly.

Gemini, what happened in May 2025?

Might be a bit tricky to do with posts that have multiple links within, but I have to imagine there's some way.

It'd be swell if bluesky could actually use anchor tags instead of divs w/ onclick handlers everywhere for navigation, I'd often like to open posts in new tabs but can't easily do that (middle or right click) thanks to current setup

What’s “managing funds really well” in context?

Given the massive divide in maintainer time & compensation, I personally see little issue in paying maintainers as soon as funds are available — it’ll never really be “fair”, and unlike personal finance, there’s little reason to save for a rainy day.

I can’t imagine that’d ever happen, way too big of a lift and wasm is generally good enough.

> EAA in Europe and Section 508 in the US disagree with that stance.

Then they’re objectively wrong, which isn’t the first time. A11y doesn’t work like that, no two apps are equal and user preferences do not hold across them because of this.

I was not referring to the feature at all and a change certainly did not get merged in Chrome 76 as there are people bringing it up still as of October in that thread.

I’m out, and unable to check if it’s been fixed in the last 6 months, but it certainly was an issue recently yet.

System-wide a11y is objectively bad a11y though, global preferences will never work well, is my point. It really has no tie-in to OSes besides (some of them) having a toggle button.

I'm not sure "we're actually not trying to provide a good experience" is a good excuse for not doing *something*.

I've watched this ticket for years now, and it appears to still be open, so I don't think that's the case?

issues.chromium.org/issues/40116...

IMO, it's handled abysmally at the moment, and I really don't understand why you'd say otherwise?

Requiring users switch their preferences OS-wide, and all apps that respect the settings, just to get better a11y on a single site is a horrific failure of accessibility.

No, that Chrome setting only affects browser UI, it does not change the value of `prefers-color-scheme`. I don't understand it, but the Chrome & Safari teams have resisted all efforts to add a toggle.

OS inconsistency is an argument _against_ adding controls? Sounds like giving users less options.

FWIW, Chrome (and I believe Safari?) don't actually have browser-level preferences, and on Linux, there's a dozen different ways to express preferences that the browser isn't guaranteed to respect (I've created gtk configs purely for chromium).

It's really hard to argue it's already handled well.