kangalioo

What happened to Wuffs? https://github.com/google/wuffs

A programming language specifically geared for file format libraries, including safety and performance

Especially because the last step is less "make it" and more like "npm add it".

The browser has "ready-made components". React (et al) also has "ready-made components". React's selection is a thousand times larger. React also allows you to make your own. The browser doesn't [1].

Who can blame developers for skipping the inconvenient half-solution and going straight to the all-encompassing ecosystem

[1]: It does, but web components are too unwieldy to compete with React

This article misses the most crucial point.

Developers don't "use the platform" because all the building blocks for any component you can think of are already there. When "the platform" adds yet another set of not particularly pretty and badly configurable components, why would we use them? When there's tons of mature, deeply thought out, battle tested, cleanly abstracted components out there - especially when _those_ components are made of divs and layouting and CSS and are therefore malleable to the pixel instead of being made of magic browser pixie dust?

How does it compare against the Pi harness, which I thought was the unofficial harness champion so far, in your workloads?

Where can you watch it happen? Is there a video visualizing the diffusion process on text?

I assume because promising trustworthiness by sourcing information from the web is specifically Perplexity's shtick. The fact that this study undermines the quality of random web sources hits Perplexity's value proposition the most.

Is this sarcasm?

Thinking mode is exactly this. The LLM is allowed to dump thoughts, correcting itself along the way, and only surfaces the cleaned-up final answer to the user

From what I've heard, Ollama has a bad reputation because it's a thin wrapper around llama.cpp without attributing it properly, thereby stealing recognition from the maintainers doing most of the work

$500 would amount to just 10k requests. One million requests is $50k

0.05 cents even. The beautiful thing is, minuscule amounts of micropayments are already enough to fix the incentives

Every git repository hosting platform - GitHub, GitLab, Bitbucket, Codeberg - necessarily gives end users control over filenames.

"User-controlled strings as parameters" is not "access to your git" but it is still an attack vector.

How does Pangram detect this? How do we know this is any more reliable than just asking any HN user to judge a text for AI signs, i.e. how is this more authoritative than the comments you're responding to?

I haven't heard that number before and I'm actually impressed.

For a file-by-file close to literal translation of a memory-unsafe language to Rust, isn't that pretty good?

I was expecting the post-rewrite Bun's unsafe usages to be in the order of magnitude of C2Rust (https://c2rust.com/). I was _not_ expecting it to be in the same ballpark as hand-written Rust.

In what sense was Bun's rewrite performative? If I recall correctly, Jarred deliberately tried to fly a bit under the radar to prevent the inevitable outcry

What would be a better way to incorporate AI as a spell checker?

In comparison to non-AI traditional tools, AI has the advantage of "understanding" the text, reducing the number of "stupid" mis-corrections. And its spelling correctness is usually already impeccable, so what is there to gain by interfacing it with traditional solutions, and how can it be achieved?

Is there research about this?

That sounds like it would make productive AI usage much easier, but it also sounds very brittle

The "frontend skills" whose growing irrelevance are bemoaned in this article consist largely of navigating a minefield of unintuitive edge cases, browser incompatibilities, historic baggage, exceptions to exceptions to exceptions.

Modern frontend, or the "tower of leaky abstractions", is finally a common-sense mental model for web development. Supplanted by force on top of an exploding bag of eccentricities that is web standards and conventions. The fact that it works at all and is merely a little leaky is an accomplishment in itself.

If you're running Claude Code in a container anyways, why does `--dangerously-skip-permissions` not work for you?

To be fair, the culprit in the article is _less complex_ than branch prediction: "with random data, bits are flipped often, and bit flips in transistors inherently draw power" is less mental gymnastics than "with random data, the cpu fails to predict the future, causing redundant speculative execution"

I agree. The article mourns the death of pentesting as an art form due to automation. But you could also celebrate the death of pentesting as an arduous necessary evil due to automation

So glad I found someone mention this. Amazon.de, SPIEGEL.de is down. Highly prominent sites unreachable. I wonder how long this will last and how big of a thing this ends up being once people talk about it :o Feels big to me

A much cheaper way might have been to buy a couple of Novation Launchpads. 8x8 full RGB tactile buttons for 90€, MIDI-controllable. Four of those next to another for 16x16 at 360€ plus a little bit for cables and controller comes out at 1/3 the price

Someone using JPEGXL in a real world product

If someone made that happen I'd be ecstatic

Which is, the UK? China?

But what's wrong with sending the number as a string? `"65537"` instead of `"AQAB"`

Why does the specification specifically mention them, then?

I'm afraid this introduces too much magic

Personally, "0%" doesn't mean "hasn't started" to me, it means "not enough progress has happened to reach 1%". Assuming I'm not alone with that, the rounding becomes a simple truncation `roundedPercent = int(percent)`

An implementation as simple as the concept, which is a good sign in my experience

Can be pretty fast if you know your tools well :)

Just tried in VSCode; 2:38min using its fabulous multiline editing. 1:34min on second try due to practice