sharktheone

That is a lot. I thought Anthropic models would just do the opposite because they are greedy for money.

Interesting to see it now. I've used it a bunch before it came out and i pretty much didn't notice it. It might have been slightly better code quality, but still not great in that. I guess it just was slightly less frustrating to work with, but still AI...

The realtime dashboard was really interesting. I think I've never seen any lab being so open about this

That makes sense. I mean just that it is hard doesn't mean it is impossible

Honestly, Flock should be dropped in general.

But someone will be able to get vehicle data out of them was just a matter of time.

Such systems are never great. Even when you ignore the obvious reasons on why not to do that, security is just the last nail in the coffin. Now just imagine what bad actors can do with that data. They now helped the people they are trying to avoid and they always will

Don't really understand what the drama here is. They said it was a temporary uplift and even though they could just take the whole uplift away, they just cut it by 25% which is better than before.

I really think Anthropic deserves much shit, but that's just shitting on the for the wrong reasons. And I hate that I am defending them here.

I hope GPT-3.5 will be released at some point. Would make me really sad if that gets lost at some point

I don't understand why you are saying this. That just seems a bit inappropriate

I hope that the CC0 license doesn't make any problems. It might though since decompilation does not remove the copyright of Capcom. Maybe just hope that they just don't care anymore

Very cool too see. So happy we actually have SIMD in WASM

I feel like archive.today is as important as webarchive is. Not only for uptime redundancy, but also redundancy in what they capture.

> The first thing that caught me off guard — and honestly, who would’ve expected this to be the memorable part — was IDE support, or the near-total lack of it.

Yeah this was also surprising to me the first time I tried Zig.

ZLS exists, but back when I last used it it was unstable and crashed frequently.

I wrote a small toy JVM. In case anyone interested: https://github.com/Sharktheone/yeeeeem

I think that wouldn't change much. They would just make use of more grapheme clusters.

For emojies they already make heavy use of the Zero-Width-Joiner. So a woman firefighter is the woman emoji + ZWJ + fire engine. Sure the UTF-8000 approach is much better encoding size wise.

Even if they wouldn't be vibecoding. They were able to write slop before AI

It would be actually funny if a LLM wants to just put it's weight here during benchmarking.

Probably Mythos / Astra will just be way too large

Spotify talking about Quality and it's product is funny to me. It is some of the most buggy software I (sadly) use day to day.

Sure it has improved, but sometimes Audio playback just doesn't work or is degraded in an unacceptable way. That is the core of the product and appears to still be unstable after all these years.

Having spoken to some of the core people working on Firefox, not all of them seem too happy about all of the the AI integrations.

I think this is mainly Mozilla trying to stay at least partially relevant and going on the hype train :/

> If all that comes out of this is a few people learning how to build Firefox from source I'd consider it a win.

Firefox is surprisingly easy to build! Only around 6 commands with cloning the source.

For anything interested: https://firefox-source-docs.mozilla.org/setup/linux_build.ht...

For me the compile process that took no more than 15-20mins. Partial builds are even faster

it tells something that the word "aperture" is only mentioned 3 times in the test

Interesting to see what they got out of basically the same camera sensor. It seems largely just like way better denoising capabilities. Maybe that is why they needed the extra NPU in the A20 chip.

But even with that they are still behind Android phones, while they dominated some years ago

So interesting to see. Though I think if Cloudflare really wanted to safe more memory, there are even more strategies. Just starting with pointer compression, even if that is not an option, other techniques exist in rust. Assuming some numbers in from the previous post, with 250GB per byte wasted, you could shave off 2bytes of every pointer or 4 if you can fit them into a context related memory cage. And even that is just a beginning here.

Ouch, this seems so bad. Just another punch into AOSP and custom ROMs like GrapheneOS

This is really interesting. I wonder how this might affect Chromes performance if they decide to implement such thing or just optimize the C++ code responsible for emitting sub optimal skia instructions

Very happy for it finally being there!

I attempted to implement the Temporal spec myself and it really isn't fun. After the tc39 meeting in May, I also switched to `temporal_rs` for `yavashark`. My own implementation ways everything else than optimal, I didn't support the RFC9557 timeformat. In theory you could even use temporal_rs to provide an perfect polyfil with wasm.

I highly doubt that. At least not in a native way. However, you could make an adapter from the native interface to wasm components. Would be interesting to look into!

The acronym was basically there from Day 1, I am not the Person who came up with it and I also wasn't there at Day 1.

Servo and Tauri still let the Application use JavaScript, which is just very wasteful. I was thinking about making it possible to write dynamic websites completely without JavaScript, by building up WASM support and making that even faster. But that's a not the goal for now. Indeed, we don't even have a JS API for websites yet...

Well, the application is basically unoptimized. However, we also don't have many components yet that can eat up your memory. But we try to keep the memory usage as low as possible. We also thought about integrating it into something like Electron, and for something like that it is important to have a low memory usage.

Lateron, we might have some compile time flags to reduce memory usage in cost of a bit of speed. Most Electron apps don't require that big oomf performance on the startup, since everything is locally and can be cached in a better way.

Yes, the HTML parser, probably the component that is the most spec compliant. I think the CSS parser also is. For the initial phase, I think it is more important to get something working. For a later implementation, the spec should dictate what we do and what not. But with a small team, it is just not feasible.