leobg

5 ms·

1.
by leobg 3d ago
Sounds like an interesting way to benchmark models. Put two instances of a model in a loop with itself. Run for n cycles. Then look at the output.
2.
by leobg 4d ago
And by the way, I would really love to somewhere find a log file of your personal life and see how often you have promised something, and it didn’t turn out quite to be as perfect as you thought, or took longer than you thought. Or were you
3.
by leobg 4d ago
You are removing agency from everyone who is not visible, even though they might buy trash and make it profitable for thousands of sellers on Amazon to lie, and to fraudsters to lie, etc.. Pointing fingers at those who are the most visible
4.
by leobg 5d ago
> You can make more of these electrical slow waves through an auditory stimulus, if it comes at just the right time. I wonder: 50ms of pink noise at the right time to make your sleep more restorative. Why or how should such a “backdoor”
5.
by leobg 5d ago
He who fights with monsters might take care lest he thereby become a monster. (Nietzsche)
6.
by leobg 5d ago
This is the best explanation. They’ve been doing the same in Google News. Each entry comes not with a URL to the source, but with a hash. To resolve it, you must send requests to Google’s servers. Anyone who wants to create a list of URLs o
7.
by leobg 5d ago
The pain being made is: If you use a search engine as your eyes to see what exists “on the internet”, then, absolutely, whatever Google hides from its results or fails to index is “erased” from “your” experience of the internet.
8.
by leobg 8d ago
Giving something a sinister name doesn’t change what it is. Didnt Altman say a few years ago that the next frontier for training would be synthetic data? It’s what everyone turned to (Tiny Stories, etc.). Using existing LLMs to generate mor
9.
by leobg 11d ago
Do your homework. This has been debunked. Also, it was about emeralds.
10.
by leobg 11d ago
Or about what the word “is” means. If you’re Bill Clinton. :-D
11.
by leobg 15d ago
Thanks. Will try!
12.
by leobg 17d ago
I tried running Kokoro on my iPhone XS (several years old), and it was slower than real-time. So, i wouldn’t expect it to be usable on a Kindle. But if you find a solution, please let me know.
13.
by leobg 21d ago
How can that be legal? You’re assuming anyone famous enough to have a Wikipedia page is granting you a license to republish their content? I think the only thing between your project and a cease and desist is lack of popularity. If proxifyi
14.
by leobg 24d ago
I’d like a book search engine that allows me to search for books covering a certain place and/or time. Like: “Is there a WW2 memoir of someone who passed the place where my grandpa was POW?”
15.
by leobg 26d ago
Calling another person a monster because you disagree with them (or what you heard about them from third parties) is not the pinnacle of civility. Just think about what you’re saying here. Monster: “Malformed animal or human, creature affli
16.
by leobg 26d ago
What’s next? Butter healthier than margerine? ;)
17.
by leobg 27d ago
All this talk of animal rights, and then this! /s
18.
by leobg 27d ago
I bought a cheap laptop to make a single purpose device for my mother in law. Basically a scan terminal. Connected to an old ScanSnap. Drives the scanner, handles cloud backup. Custom TUI. Also, the XTEINK. Made it a single purpose flashcar
19.
by leobg 27d ago
Dumb question: Why not just put an instruction into your favorite harness’ system prompt: “If I give you pseudo code, spell out my intent, and then write and test it in real code.”
20.
by leobg 27d ago
I don’t have MIDI. How about whistling or playing the piano via microphone? Sounds easy. Another 6 month rabbit hole? :)
21.
by leobg 29d ago
IIf you’re over-regulated, you’re bound to get something right every once in a while. To over-regulate and get 100 % of regulations wrong would require someone in the system to be able to tell good from bad, aka competence. :)
22.
by leobg 29d ago
Literally your source: > her stepfather used the Grok chatbot to transform a childhood photo into child sexual abuse material He could have used Stable Diffusion. Or, if this was the 90s, Photoshop. By your argument, we’d have to abolish
23.
by leobg 1mo ago
I was looking for it after reading the post, but there seem to be so many different versions and edtions. Harper, Penguin. College. The New. The Concise. The International. Some say they’re organized alphabetically. How do I know which one
24.
by leobg 1mo ago
I don’t think the “Firehose” level was. I think it was 50k/mo or something. That was before AI training was a thing.
25.
by leobg 1mo ago
https://github.com/mattpocock/skills/blob/main/skills/produc... > Write a handoff document summarising the current conversation so a fresh agent can continue the work. > […]
26.
by leobg 1mo ago
It’s actually your #1and your #3 plus: “Wo kämen wir denn da hin?“ (Just think where this would lead to.)
27.
by leobg 2mo ago
Source reads like Tony Robbins for LLMs: Excuse "Too simple to test" Reality Simple code breaks. Test takes 30 seconds.
28.
by leobg 2mo ago
I don't even watch most videos anymore. YouTube app -> Share -> Gemini -> Summary. Only if that shows interesting content do I actually watch.
29.
by leobg 2mo ago
Fiction from the 1800s? Now that sounds interesting. Can you name three titles that you would recommend? And: how did you get into this? Doesn’t sound like a TikTok trend, haha.
30.
by leobg 2mo ago
Why not generate an artificial dataset using commercial APIs and then finetune a small model on this data? I’ve had success adapting even a 7B model for single-domain tasks that way, including reasoning and tool calling. You can use an open

More ›