jedbrooke

6 ms·

1.
by jedbrooke 4d ago
ok, what is taste? how does one acquire it?
2.
by jedbrooke 9d ago
I wonder if another path to “success” for ladybird could be rather than end users downloading it for the main browser, if it gains enough dev interest, we could get some kind of competitor to electron and it would end up getting bundled in
3.
by jedbrooke 12d ago
looks like it only has the physical connections for USB 2. Fun fact about usb 3, it’s really a completely separate physical standard from usb 2, just designed in such a way that a single port can have BOTH usb 2.0 and 3.0 connectors in it a
4.
by jedbrooke 13d ago
according to https://downdetector.com/ Gemini is down too (and copilot, but that just uses ChatGPT right?)
5.
by jedbrooke 14d ago
I tried this in my mac mini m2 16GB, unfortunately I have to use an usb disk for the model weights, and I’m getting 0.5 tok/s. Still, being able to run (heh maybe crawl is more accurate) a 100B model on this computer AT ALL is pretty c
6.
by jedbrooke 15d ago
surprised to see nothing at null island[0]. Good job cleaning your data [0]: https://en.wikipedia.org/wiki/Null_Island
7.
by jedbrooke 19d ago
I have to say though as a xfinity customer, being able to get wifi basically anywhere urban is pretty nice
8.
by jedbrooke 20d ago
https://xkcd.com/927/ To actually add something to the discussion though: I guess getting LLMs to control robots is cool, but would potentially have more severe consequences than pure software since it interacts more w
9.
by jedbrooke 22d ago
that and always putting the “current year” at the end of the search term (so the results are more recent, I guess?), except that “current year” consistently ends up being 2-3 years ago since I guess that’s what’s in the training data (even
10.
by jedbrooke 22d ago
see also https://en.wikipedia.org/wiki/Telecine
11.
by jedbrooke 26d ago
I gotta laugh at some of the models it suggests, for example: > AnkitAI/Parable-Qwen3-4B-Claude-Fable-5-GGUF you’re telling me you managed to fit Fable 5 into just 4B?
12.
by jedbrooke 27d ago
maybe a stupid question, but, don’t you have access to all those things now? What’s stopping you from becoming Da Vinci now?
13.
by jedbrooke 28d ago
huh, sounds like they’re talking about over fitting and datasets etc, it seems like this is almost more like a fine tune/distill than just a pure quantization
14.
by jedbrooke 1mo ago
I feel like the current “reasoning” that LLMs are doing has got to be a dead end eventually. Every time I have to read another answer with “but wait” and “Actually,” as they “reason” their way to a (sometimes) better answer, I feel like the
15.
by jedbrooke 1mo ago
nope just my normal m2 mac mini. I bought it just as a normal computer to do my taxes and whatever, so it’s mind blowing that I can run this kind of AI workload on it. Well, “run” might be generous, it gets like 3-5tok/s I’m working on
16.
by jedbrooke 1mo ago
100k is all the context I have ram for, this is with any auto-compact turned off. This is using Cline in vs code. I’m sure I could tune the system prompt and mode switching more to work better with this specific model, but I haven’t gone do
17.
by jedbrooke 1mo ago
I hope the bonsai team makes another 1bit quant of this model (or releases code/instructions on how to do it), using the Qwen3.6 27B on my 16GB mac mini has been wild . The 1bit quant feels like opus level… for the first couple turns.
18.
by jedbrooke 1mo ago
it’s missing spinners for 5-10 seconds while it loads endless megabytes of javascript doing who knows what. Seriously how did we get to this point? I remember growing up in the 2000’s and all the marketing about computers was “it’s fast!” a
19.
by jedbrooke 1mo ago
Sorry to be the guy on HN complaining about scrolljacking on the article, but this has to be one of the worst I’ve seen. Text is cut off near the bottom of the description, but if I scroll down to see it, it jumps to the next “slide” and I
20.
by jedbrooke 1mo ago
the “Actually… But wait!” style responses are so annoying, even Claude opus struggles with this so I’d be interested if meta has done something to cut down on that while still giving good responses
21.
by jedbrooke 1mo ago
This is a nice little archive of all these different installers. I’d almost forgotten about all this madness since I’ve been using homebrew to install stuff for over a decade
22.
by jedbrooke 1mo ago
I’m surprised I haven’t seen anyone mention video models yet. I don’t know how many fps 17000 tok/s translates to exactly but it’s gotta be a lot. Might make real time AI video possible. Now that I think about it, real time AI video mi
23.
by jedbrooke 1mo ago
I wonder, if indeed there is astroturfing going on, if instead of a foreign actor it’s competing AI companies. Eg OpenAI astroturfing opposition to when Anthropic wants to build a datacenter, and vice versa
24.
by jedbrooke 1mo ago
I mean, if you had fast enough internet connection, you could just stream it over https
25.
by jedbrooke 1mo ago
nonexistent functions being referenced is pretty bad, I haven’t had that issue since like opus 4. Shows either complete lack of oversight and/or using a bargain basement model (so no even having financial skin in the game). Most likely
26.
by jedbrooke 1mo ago
This is one of my pet peeves too. Like when someone says “ChatGPT says: …” I just want to stop them and say “Look, you are a human with a functioning mind and thoughts and I’m sure YOU have something valuable to contribute here, I want to h
27.
by jedbrooke 1mo ago
the bonsai 27B 1bit quant version of Qwen3.6 27B is even more nuts, model fits in 4GB, and with 100k of content model+kv cache fits in 8GB. I’ve been running it locally on my mac mini 16GB. it gets around 4-6 tok/s, so not quite real-t
28.
by jedbrooke 2mo ago
from the readme: > the filtering is optimised for framework 12's hinge stiffness. never in a million years would I expect to have to optimize software for a given hinge stiffness, this made me chuckle just reading it also: > why
29.
by jedbrooke 2mo ago
it would be interesting to see results from works that don't have notable existing film/tv adaptations or illustrations (though I suppose there will always be fan art) to see if it’s really able to generate something novel. I’ve h
30.
by jedbrooke 2mo ago
UCLA is often used as a stand in for many colleges in movie sets (notably Harvard, as one of the buildings in UCLA has the Harvard crest on it specifically for this purpose) due to its proximity to Hollywood. It likely ended up in the train

More ›