abound

11 ms·

121.
by abound 1y ago
While StarLabs has shipped products before, I'd be careful with them. That StarFighter laptop was supposed to ship years ago, and I don't think any have been delivered yet. Here's a Reddit thread [1] from three years ago sayi
122.
by abound 1y ago
This intuitively makes sense (like deep-frying a JPEG), but it doesn't seem to happen in practice, as modern models are frequently trained on text both output from other models, and curated from other models. Realistically, going forwa
123.
by abound 1y ago
Interesting, thanks for the links! I had read this part of the article: > All competitors, including OpenAI, were limited to identical hardware provided by AtCoder, ensuring a level playing field between human and AI contestants. And ass
124.
by abound 1y ago
I'm guessing he didn't have access to any LLMs while competing, but I think a "centaur" approach probably would have outperformed both "only human" and "only LLM" competitors. Reading through the chal
125.
by abound 1y ago
Not sure if anything major has changed recently, but I think you bring your own CMS. I've used Astro successfully with Strapi, for example.
126.
Show HN: m(ctf)p – A semi-automated environment for solving CTF challenges(git.sr.ht)
2 points by abound 1y ago|0 comments
127.
by abound 1y ago
Similarly: https://flic.io/ Little programmable buttons, they're great. I've used them for similar habit formation stuff, in my case not ignoring my alarm, by making the button the only way to shut it off, and put
128.
by abound 1y ago
If you're just here looking at the HN comments: check out the article. It's really well-written and has a bunch of nice visualizations if you like astronomy things.
129.
by abound 1y ago
Current LLMs are multimodal -- you could record yourself singing + beatboxing, upload it to an LLM and ask it to translate it to Strudel code. My best guess is that it won't work well, but we'll never know until someone tries it.
130.
by abound 1y ago
Is the argument that entire categories of software should not be used because they could be backdoored? That'd seemingly rule out messenging apps, as well as the examples you mentioned (VPNs, remote viewing tools) I think a more reason
131.
by abound 1y ago
One surprising thing to me is that using model outputs to train other/smaller models is standard fare and seems to work quite well. So it seems to be less about not training AI on its own outputs and more about curating some overall qu
132.
by abound 1y ago
As far as "declarative abstraction layers" go, I think the reason that distros don't have them is because they exist one level of abstraction higher in tools like Ansible or Chef.
133.
by abound 1y ago
I feel the same way, but to defend GP, if you're dealing with unknown JSON, the experience of deserializing `any` (or `map[string]any` if you know the top-level entity is an object) can be extremely tedious. For things like structured
134.
by abound 1y ago
Same, I think I originally got it from the Arch Linux wiki[1], which explains it well. It in turn references HN, to come full circle. [1] https://wiki.archlinux.org/title/Dotfiles#Tracking_dotfiles_...
135.
by abound 1y ago
I hit this yesterday even though my account is on Tier 2 or 3. In my case, the issue was that I was using an old model name (exp-03-25 or something) in requests. Update to the latest pro-preview-whatever and the rate limit issues should go
136.
by abound 1y ago
I was curious about this, so I looked for data from prior to 1970 [1] and it's all clearly trending in the same direction. So 1970 isn't special or misleading in this case, there does not appear to be any bamboozlement going on. [
137.
by abound 1y ago
I wanted to turn left off of Stow St (and so did everyone else it seems), but at the intersection, there were no left-facing arrows, so we're still on Stow St.
138.
by abound 1y ago
I think it's pretty clearly hypocritical, but also if his actions are (far) more than offsetting his own emissions and impact, it's still a net positive. Of course, he could choose to not live a super-high consumption lifestyle in
139.
by abound 1y ago
I didn't see anyone mention the McMaster-Carr website [1]. It may not be the "densest" out there, but it's clean, functional, and nicely presents a lot of information at once. [1] https://www.mcmaster.com&#x2F
140.
by abound 1y ago
Like the GP said, the whole topic of this article is getting an accurate figure, and how the 2%/9th largest figure is incorrect.
141.
by abound 1y ago
Your other comments indicate you work there, you might consider mentioning that.
142.
by abound 1y ago
Depending on how long you're planning on sabbatical-ing, you might consider applying to the Recurse Center [1]. It's basically a (free) 6- or 12-week self-directed learning program that you do with others. I had done a half batch
143.
by abound 1y ago
> Unconventional postcards: A thin sheet of plywood Can confirm, I laser cut wedding invitations out of 1/4" plywood and mailed them out like that. I think it required some "non-machineable" stamp or similar, but they
144.
by abound 2y ago
For what it's worth, I don't think there's any hard evidence that big companies are doing this. Aside from the technical difficulties of constantly listening to 1+ B mobile devices and running speech-to-text, it'd also b
145.
by abound 2y ago
The title is "What the Press Got Wrong About Hitler", and the majority of the piece is about German and foreign press consistently mocking and underestimating him even as he rose to power. It seems like the title and article are p
146.
by abound 2y ago
I think the tricky piece is that you'd need to find that inode number without using any dynamically linked libraries , including ls and cat in the author's case. And debugfs is likely dynamically linked too (I just checked and it
147.
by abound 2y ago
Reminds me of the Jeff Dean "Numbers you should know" schtick about latency. Other people are asking about how to use this "tool", I think it's just a rough reference. I almost see it as a kind of art/poetry, t
148.
by abound 2y ago
They could very well just be temp files, I know the Go standard library will write large multi-part file uploads to temp disk, for example. It'd be better to totally stream it of course, but that's not always an option for one rea
149.
by abound 2y ago
In a similar vein: I had a new coworker named "Jason" join a team once, and for the life of me I couldn't help but type "JSON" first when writing emails/messages with his name. It took weeks before I could reli
150.
by abound 2y ago
As a remote tech worker who moved from the Bay to a (much) cheaper place - I agree that the expensive places are expensive for a reason - but if one can articulate what specifically they're looking for, it's almost always possible

More ›