Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
Terretta
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
17 ms
·
361.
▲
by
Terretta
8mo ago
Simplifying for brevity* -- there are three levels in the GitHub entity: - accounts (personal) - orgs (companies, directories, teams, roles etc.) - enterprises (sets of orgs) Even with enterprise SSO, the initial connect to GH ca
362.
▲
by
Terretta
8mo ago
> built by the unaware awash in bliss
363.
▲
by
Terretta
8mo ago
At least the paper is transparent it's a twist on Edwards cogwheel. I've encountered these in the wild, so interesting to see what's generating them.
364.
▲
by
Terretta
8mo ago
You know this reply is transparently LLM generated, right? It's what happens when you paste the HN feedback into your vibe session as if you wrote it, then copy the response back here.
365.
▲
by
Terretta
8mo ago
Has EU weather sources per credits (DWD, ECMWF, EUMETSAT -- roughly what it's doing is graphing multiple models), but if you are into weather apps you're likely best off with Carrot that (a) lets you design your own UI including m
366.
▲
by
Terretta
8mo ago
Compare 1996's "FITALY": "For the FITALY layout, we have obtained an average travel of 1.8, to be compared to an average travel of 3.2 for the QWERTY layout. (For prose, involving few numbers and symbols, the results ar
367.
▲
by
Terretta
8mo ago
Almost no such thing as acting. It's all casting . It seems fewer than 1 in 20 actors act, the rest are cast.
368.
▲
by
Terretta
8mo ago
> But it's a good reminder that a lot of denialists are waiting for a personal, practical reason to care. Four out of five denialists agree!
369.
▲
by
Terretta
8mo ago
> If you're looking for something beyond corporate safespeak or stylistic pastiche, they drain the blood out of everything. Strong agree, which is why I disagree with this OP point: “Stage 2: Lexical flattening. Domain-specific j
370.
▲
by
Terretta
8mo ago
> most of the time you only get what the most boring person in the most boring cocktail party would say don't be mean, it's median AI à la mode
371.
▲
by
Terretta
8mo ago
Better link shows exchange: https://news.ycombinator.com/item?id=9224 Or this pulls the exchange under the famous HN post itself: https://news.ycombinator.com/item?id=8863
372.
▲
by
Terretta
8mo ago
What happened to this block and the keyboard key arrangement? ESC [ { 11011 FS \ | 11100 GS ] } 11101 Also curious why the keys open and close braces, but ... the single and double curly quotes don't open and clos
373.
▲
by
Terretta
8mo ago
Love Halide, and Pro Raw and Nitro, and pay for these even the iPhone's looks grid + RAW meant I didn't really have to. This could bring me back, the B/W with shoe-black blacks is lovely. Hopefully some folks using the previe
374.
▲
by
Terretta
8mo ago
> The fact that you're getting 15-30 tok/s for text gen on phone hardware is wild — that's basically usable for real conversations. Wild how bad it is compared to, say, Russet for iOS/ipadOS, which runs these same
375.
▲
by
Terretta
8mo ago
This isn't a ______. This isn't a ______. This is ______. The message is clear: ______ isn't a nice-to-have. It's a ______. But here's what nobody's talking about: ... I can't force myself to spend m
376.
▲
by
Terretta
8mo ago
How do you feel about Safari tab groups? They're perfectly happy to snap open a thousand tabs at once, from however many such groups you like, all synced across all the (Apple) devices. And on the left of the Safari window in a slidin
377.
▲
by
Terretta
8mo ago
This doesn't sound like a tech support question. Sounds like stepping up to the mic at an emotional support group: "Am I the only one?" Well no, this group is your people and you're speaking to the choir.
378.
▲
by
Terretta
8mo ago
Right, “like a stereogram”! https://en.wikipedia.org/wiki/Autostereogram
379.
▲
by
Terretta
8mo ago
> That's smart - but it is still… > That's a clever… The difference in my approach… Are you using LLM to help you write these replies, or are you just picking up their stylistic phrasings the way expressions go viral at
380.
▲
by
Terretta
8mo ago
If you put two proofs side by side, you can view from the right distance then uncross or cross your eyes like a stereogram till they converge, which makes differences shimmer. Instant "spot the difference" solve. // Long
381.
▲
by
Terretta
8mo ago
I'd like to imagine I know which of each example were better designed, but the handle going to the side opposite from the label was making me second guess. Move handle away from the label to reveal is how I took it, so hope that'
382.
▲
by
Terretta
8mo ago
> I wrote it The commits feel more human than usual these days, as does the timeline.
383.
▲
by
Terretta
8mo ago
Love the idea and the approach, both. Intend to try it. Dislike READMEs written by or with the LLM. They're filled with tropes and absolutes, as if written by a PdM to pitch his budget rather than for engineers needing to make it work.
384.
▲
by
Terretta
9mo ago
> what makes you think that the features that are personally important to you are the same features that other potential users need? Good question, what's the pitch: “Vecti is a browser-based UI design tool built from the ground
385.
▲
by
Terretta
9mo ago
This is a delightfully informative blog for 15 years. Amazing these days.
386.
▲
by
Terretta
9mo ago
> you first have to train the model on a BUTT TON of that new language Tokenization joke?
387.
▲
by
Terretta
9mo ago
> > But “the path forward is brutal” > Great insider feedback. Also a common LLM trope.
388.
▲
by
Terretta
9mo ago
> the thing that actually executes the work ^H^H^H^H toil
389.
▲
by
Terretta
9mo ago
My opinions are my own … but rhyme with this, particularly the final paragraph, with a twist: the strong foundation with in-house mastery that can sherpa an atelier of agents.
390.
▲
by
Terretta
9mo ago
> You're no longer developing software, you're doing therapy for robots. Or, really, hacking in "learning", building your knowhow-base. > But it delivers enormous compounding gains, and you can use your agent to
More ›