scratchyone

5 ms·

1.
by scratchyone 13d ago
any way to see the tok/s for all the models listed on your homepage? curious which has the best speed/quality tradeoff for me
2.
by scratchyone 2mo ago
we have this!! https://www.pangram.com/solutions/chrome-extension (disclaimer, i am a pangram research scientist)
3.
by scratchyone 2mo ago
I contribute to Storyteller in my free time and I absolutely love using my Boox with it, honestly my dream reading setup.
4.
by scratchyone 3mo ago
fully self hosted without any external dependencies is incredibly impressive, amazing work
5.
by scratchyone 3mo ago
ooh, this seems super cool. i see that he doesn't print jackets, have you gotten them printed elsewhere for that?
6.
by scratchyone 4mo ago
This is an article from almost a year ago, last edited July 3, 2025...
7.
by scratchyone 4mo ago
thank you lol, this has convinced me to switch over to fastmail
8.
by scratchyone 4mo ago
As someone who also loves Kagi, what do you like so much about Fastmail? Seeing it compared to Kagi makes me want to give it a try as well haha.
9.
by scratchyone 4mo ago
I've had generally good experiences with Exa, I use it as an MCP in all my various coding tools and use the API whenever I build an AI thing that needs search. Also an extremely happy Kagi user but can't speak to their API, haven&
10.
by scratchyone 4mo ago
To be fair, just because you understand the code you see doesn't mean its safe to copy-paste. Many of these compromised sites will show something that appears to be a benign command but will be something entirely different when you cop
11.
by scratchyone 4mo ago
They will never remove tissue if you're still alive. This is the reason organ donation is most common in brain-death cases, because the tissue is still alive but you are entirely dead. As you point out, it would be horrible to dismembe
12.
by scratchyone 4mo ago
Honestly still insane to nuke a high-volume client's business after a single payment issue. There would be no reason for Google to believe that a single hiccup like that is evidence that they won't get paid and have to cut account
13.
by scratchyone 4mo ago
https://github.com/instructure/canvas-lms/pulls?q=is%3Apr+is... haha i went to go check and they haven't merged a PR since 2017
14.
by scratchyone 4mo ago
felony murder is pretty widely regarded as a leading factor in incredibly unjust prosecutions and sentencing decisions. perhaps not the best concept to build your ideas on top of.
15.
by scratchyone 5mo ago
that's quite smart. i was almost stupid enough to paste it into a terminal to check if it worked before deciding to wait and let others analyze it first haha
16.
by scratchyone 5mo ago
funny how he didn't care about ethics the moment it was more profitable to release it than to talk about dangers.
17.
by scratchyone 5mo ago
unfortunately the page can also lie to you haha. it seems people have reviewed the code by now, but running suspicious shellcode you don't fully understand is never a great idea.
18.
by scratchyone 5mo ago
my company intentionally avoids all AI usage and yet our enterprise google workspace is still constantly sending us ai generated summaries of all of our various sensitive documents
19.
by scratchyone 5mo ago
tbf most companies don't have a potentially world-ending product. only real similar field is defense contractors who typically can't brag about unreleased ideas as they're classified.
20.
by scratchyone 5mo ago
Don't worry, these companies will make sure we get to experience both nightmare futures.
21.
by scratchyone 5mo ago
100% agreed. That's part of the issue imo, these companies pretend their new models are "too dangerous" to seem like they care about the world, yet they have no qualms deploying existing models in warfare or bragging about im
22.
by scratchyone 5mo ago
I mean, it's the BBC and the article doesn't have any typical AI tells, where is this idea that it's AI written coming from???
23.
by scratchyone 5mo ago
Honestly we should have learned this claim from AI companies was purely fear-mongering back when GPT-2 was "too dangerous to release".
24.
by scratchyone 5mo ago
I agree, it seems like disabling them is quite easy. I'm more speaking to the new user perspective. From looking at their site I would have no clue that that product even had features aside from AI. I would never have a chance to find
25.
by scratchyone 5mo ago
In all honesty, the people who want to turn off AI won't be downloading Warp in the first place. I know Warp has interesting terminal innovations because I've been familiar with it since before the AI boom, but new users can'
26.
by scratchyone 5mo ago
also > A self-contained security audit prompt is available at docs/security-audit.md. lmfao
27.
by scratchyone 5mo ago
i notice that as well. most of the time when i see those it has a retry counter also and i can see it trying and failing multiple requests haha. almost never succeeds in producing a response when i see those though, eventually just errors o
28.
by scratchyone 5mo ago
is GLM genuinely comparable to claude models? haven't had a chance to test it yet.
29.
by scratchyone 5mo ago
does google actually host anthropic models themselves?? surprised anthropic allows that, given how notoriously crazy they are about distillation or weight leaks or any hints of their models being used in the wrong way.
30.
by scratchyone 5mo ago
maybe, but the response to GPU shortages being increased error rates is the concern imo. they could implement queuing or delayed response times. it's been long enough that they've had plenty of time to implement things like this,

More ›