jsheard

12 ms·

241.
by jsheard 9mo ago
> But on the other hand, I'm pragmatic (some might say cynical?), and I'm just left here thinking "what is Signal trying to sell us?" A messaging app? I'm struggling to come up with a potential conflict of intere
242.
by jsheard 9mo ago
In video games this tends to be a function of framerate, with higher FPS being more likely to produce coil whine as the GPU power draw oscillates at a higher frequency. I assume there's something analogous in LLM runtimes when the oute
243.
by jsheard 9mo ago
I'm not sure if a standalone trackpoint really makes sense, the main benefit of those is being accessible from the home row, which you lose if it's not integrated into the keyboard. If you're moving your hand anyway then you
244.
by jsheard 9mo ago
It's been a while since laptops came with track balls, but it's a cool idea nonetheless.
245.
by jsheard 9mo ago
I should have worded that better, I meant no UPX-style end-to-end compression like the OPs project used. NES games had to be much more judicious than that since there wasn't nearly enough RAM to decompress everything at once.
246.
by jsheard 9mo ago
Don't most (all?) browsers consider file:// and localhost to be secure for the sake of enabling those features? https://developer.mozilla.org/en-US/docs/Web/Security/Defens...
247.
by jsheard 9mo ago
Zelda 1 was 128 kB, for those wondering, and that's without any compression. Double that for the sequel.
248.
by jsheard 9mo ago
> Sakurai, director of the game, is bullish on AI for game development [2] Your source is based on machine translation, and professional translators pushed back on the interpretation that he is enthusiastic about genAI. Apparently it cam
249.
by jsheard 9mo ago
JPEG has morphed into a synonym for photo, I guess. See also: Ace Combat 7s infamous JPEG dog. https://youtu.be/IXB1W8va3eg
250.
by jsheard 9mo ago
Netflix did try to turn it into a genre, they made a few dozen CYOA specials before giving up. https://www.polygon.com/22286070/netflix-interactive-shows-m...
251.
by jsheard 9mo ago
> The content is entirely gone right now. It's officially gone, but you can acquire the raw video through other means and plug it into this open source reimplementation of the frontend: https://mehotkhan.github.io/Ba
252.
by jsheard 9mo ago
My point is that you could write the most theoretically bulletproof license in the world and it would count for nothing under the precedent that AI training is fair use, and can legally ignore your license terms. That's just not a prob
253.
by jsheard 9mo ago
Writing the license is the easy part, the challenge is in making it legally actionable. If AI companies are allowed to get away with "nuh uh we ran it through the copyright-b-gone machine so your license doesn't count" then l
254.
by jsheard 9mo ago
On the other hand, sometimes you end up like this guy. Are you feeling lucky? https://arstechnica.com/health/2025/08/after-using-chatgpt-m...
255.
by jsheard 9mo ago
It's not common but few libraries use that to support JSX-like syntax at runtime, without a build step. e.g. https://github.com/developit/htm and https://lit.dev/docs/components/rendering
256.
by jsheard 9mo ago
The compression tricks used in standalone Javascript demos are significantly more cursed. The meta is to concatenate a compressed binary payload and a HTML/JS decompression stub in the same file, abusing the fact that HTML5 parsers a
257.
by jsheard 9mo ago
You have to use the system Webkit though, right? So even though Orion is based on Webkit they are still constrained compared to other platforms, where they can patch the engine if necessary.
258.
by jsheard 9mo ago
Well that and the fact that LLMs love using Tailwind, which puts the vendors in an awkward spot if the Tailwind project implodes. Makes you wonder how much ossification is going to happen because AI models are entrenched in 2023's tool
259.
by jsheard 10mo ago
Wacoms key patents did expire which is why there's such strong competition in the tablet space now, and why prices have fallen through the floor. I wonder why Apple still uses active pens on the iPad when passive EMR has become such a
260.
by jsheard 10mo ago
Browsers have clamped down on that somewhat by enforcing stricter referrer policies by default if the originating server doesn't specify one. It used to be a total free for all where everyone could always see the full referring URL, th
261.
by jsheard 10mo ago
If synthetic data is so much better then what are AI crawlers still DDOSing everyone for? Are they stupid?
262.
by jsheard 10mo ago
> The paid products Adam mentions are the pre-made components and templates, right? It seems like the bigger issue isn't reduced traffic but just that AI largely eliminates the need for such things. Or more cynically that it elimina
263.
How to Vulkan in 2026(howtovulkan.com)
3 points by jsheard 10mo ago|0 comments
264.
by jsheard 10mo ago
Each SM cluster contains 4 independent 32-wide compute units, and GB202 has 192 SMs, although only 188 of them are enabled on the largest shipping SKU. IMO that makes for 752 "cores", but depending on where you draw the line it co
265.
by jsheard 10mo ago
Tell that to Nvidia, Blackwell is already up to 752 cores (each with 32-lane SIMD).
266.
by jsheard 10mo ago
Someone needs to try running Crysis on that bad boy using the D3D WARP software rasterizer. No GPU, just an army of CPU cores trying their best. For science.
267.
by jsheard 10mo ago
I figured that was obvious - users consuming vastly more tokens than they're paying for. When people talked about how much cheaper the subscriptions were compared to doing the same thing via the PAYG API that should have been a massive
268.
by jsheard 10mo ago
They were selling you $1000 worth of tokens for $200, and now they're cracking down.
269.
by jsheard 10mo ago
Every browser requires H2 connections to be encrypted so I don't think a MITM downgrading to it would reveal anything. Downgrading to H1 might do since encryption is optional there, but the proper way to prevent that is to submit your
270.
by jsheard 10mo ago
Yeah, it's not ideal that the standard library currently has blind spots around allocation and I hope those get resolved eventually, but if you're rolling everything from scratch in C anyway then it's not much of a stretch to

More ›