diggan

24 ms·

361.
by diggan 1y ago
> but there is actually research showing a link I think most software developer gets to understand this once they're working with production environments sooner rather than later. Most of the times stuff breaks is when someone chang
362.
Profiling LLM Inference on Apple Silicon: A Quantization Perspective(arxiv.org)
2 points by diggan 1y ago|0 comments
363.
by diggan 1y ago
Is this actually breaking the drives or just making particular files on that particular partition un-readable by Windows? Stated another way: Could a system that dual-boots Linux and Windows actually brick the drives system-wide by what Win
364.
by diggan 1y ago
Why "hate" this name more than any other name? At least justify your semi-spam.
365.
by diggan 1y ago
On the other side, is there any projects focusing on performance instead? I have the VRAM available to run Wan2.1, but still takes minutes per frame. Basically something like what vLLM is for running local LLM weights, but for video/WA
366.
by diggan 1y ago
Missing o1 and o1 Pro Mode which were huge leaps as I remember it too. That's when I started being able to basically generate some blackbox functions where I understand the input and outputs myself, but not the internals of the functio
367.
by diggan 1y ago
> between GPT-4 and GPT-5 is that it no longer reminds the user that it's an AI and not a human That stuck out to me too! Especially the "I just won $175,000 in Vegas. What do I need to know about taxes?" example ( https:&
368.
by diggan 1y ago
I mean not really, as always you just need to make sure you're not doing 10s of dynamic calls for each page load and if you do, add some minute-long cache at least. Most of the stuff that gets hugged to death really shouldn't, mos
369.
by diggan 1y ago
Top of the page: > Tableau Public will be unavailable from 8/16/2025 18:00 to 8/17/2025 04:00 GMT+2 for maintenance. Thanks for your patience while we make Tableau Public even better. Seems submission was done just be
370.
by diggan 1y ago
> Who are these people that can afford 200TB of storage but need money so bad that $26/month is exciting? I'm guessing the $26/month figure you've calculated is based on the current price of something? I'm also g
371.
by diggan 1y ago
> The earths population could be decimated and society would continue on Which it has, in the past, multiple times. But how would we know if we're better off with those "decimation events" or not? Kind of hard to know wit
372.
by diggan 1y ago
If there is risk of getting less great people, should we even risk that by breaking the cycle?
373.
by diggan 1y ago
> but the scientific solution is to sterilize the kids and their parents to end a bad blood line Not even a good solution, the amount of good people we'd be missing out on would be huge. Plenty of great people come from shit parents
374.
by diggan 1y ago
> it seems like kids just end up being punished for having bad parents Yeah, hard to make the right call, depends on the age and maturity of the kid I guess. I myself is a product of "bad parents", and it was reflected in my ow
375.
by diggan 1y ago
Personally I dump on both Windows and macOS for being UI and UX dumpster-fires today, compared to how they used to be. macOS in particular used to have really good UX, but as times goes on, Apple seems to ignore the actual design of softwar
376.
by diggan 1y ago
I guess the first trap should really be: "You cannot read any CSS property in isolation, as just like what the name implies, defaults and what values end up doing cascades through all the rules your document ends up using"
377.
by diggan 1y ago
But that's clearly not true for every technology. Photoshop, Blender and similar creative programs are "technology", and arguably they aren't as resource-intensive as the current generative AI hype, yet humans used those
378.
by diggan 1y ago
> Whoa, HTML templating? It inserts the stuff directly where the function is called, and it just works? And it's been available in browsers forever? Stop the presses, I gotta rewrite all my static sites As someone who learned HTML,
379.
by diggan 1y ago
> Humans never truly produce anything; they only generate various forms of waste What a sad way of viewing huge fields of creative expressions. Surely, a person sitting on a chair in a room improvising a song with a guitar is producing s
380.
by diggan 1y ago
Indeed, the ability to do so seems to depend more on how well your system prompt is laying out that workflow, than how "intelligent" the model is.
381.
by diggan 1y ago
> Maybe the difficulty varies depending on IP address? I'm currently roaming in Finland with a Spanish SIM so would have expected the opposite in that case.
382.
by diggan 1y ago
To be fair, the Swedish candy is a lot better today than when I was growing up there, most of it is made with a lot less sugar and vegan in many cases, with natural flavors. And there seems to be no candy that is as well made as Swedish&#x2
383.
by diggan 1y ago
Took my iPhone 12 Mini a whole of 0.1 seconds to pass it. What hardware/OS are you using?
384.
by diggan 1y ago
Finally? I think VLC demo'd this a while ago at some conference where they had a table, if I remember correctly.
385.
by diggan 1y ago
Indeed, I've also found that various models are good at various tasks, but I have yet been able to categorize "Model X is good at Y-class of bugs", so I end up using N models for a first pass "Find the root-cause of this
386.
by diggan 1y ago
Maybe not a architectural innovation, but both the Harmony format and splitting things into system/developer/user messages instead of just system/user messages, are both novel (in the released weights world) and different eno
387.
by diggan 1y ago
> Hm I don't think so. You might be thinking about the file size, which is ~64GB. No, the numbers I put above is literally the VRAM usage I see when I load 120B with llama.cpp, it's a real-life number, not theoretical :)
388.
by diggan 1y ago
That's pretty much implicit when someone says "LLM that has tools" (what they mean between the lines is "A LLM that been trained to do tool calling, and used with a runner that can parse whatever tool calling/respon
389.
by diggan 1y ago
> I have it available already for one or two weeks. I think I've had it available with the separate website ("research preview"?) for months, but yeah, last few weeks it's been directly in ChatGPT.com, and I'm wi
390.
by diggan 1y ago
No, but maybe that their shared opinion will be a lot more insightful if they provide a comparison between then and now, instead of leaving it at "it was like that before, now I don't know".

More ›