Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
al2o3cr
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
18 ms
·
211.
▲
by
al2o3cr
2y ago
Only sensible that a "polygraph in your pocket" is as unreliable and useless as an actual polygraph. Heckuva job, AI!
212.
▲
by
al2o3cr
2y ago
It's not at all "new" anymore (published 1999), but if you enjoy source-code commentary then check out ISBN 978-1576104699 - "Linux Core Kernel Commentary: Guide to Insider's Knowledge on the Core Kernel of the Linu
213.
▲
by
al2o3cr
2y ago
The tricky part would be maintaining the equalization + delay matching in a changing environment - imagine a person suddenly walks in between you and the speaker that's playing the music. Unless your correction system can respond in re
214.
▲
by
al2o3cr
2y ago
Nah, they just think they do because they fucked up the counting rimshot
215.
▲
by
al2o3cr
2y ago
My guess is that this will extend to knowing not to open weird attachments from strangers. I've got bad news for ya, 2005... :P
216.
▲
by
al2o3cr
2y ago
The BIG observatories are operated semi-remotely; the onsite people are choosing from a preplanned set of observations based on sky conditions etc. https://www.gemini.edu/observing/start-here There's plenty of IT-
217.
▲
by
al2o3cr
2y ago
The documentation for that specific module even calls it out as "painfully hard-coded biquad filter code", YMMV. I'd guess those files aren't what the author actually edits - there are commits that suggest that they are
218.
▲
by
al2o3cr
2y ago
Step #2 is trickier than it seems, because you can't just make up an identity and you need to include details. Step #4 (not shown) is where the entity that you did this to has their lawyers grind you into a fine paste. That's why
219.
▲
by
al2o3cr
2y ago
"Cracked" in the exact same sense that changing a URL from /accounts/1234 to /accounts/1235 is "cracking" website security: if it works, you didn't have anything but make-believe security before.
220.
▲
by
al2o3cr
2y ago
The best/worst of those videos are the ones where you can actually HEAR a bunch of metal bits hit the ground right after the shift :)
221.
▲
by
al2o3cr
2y ago
I find UML too detailed and not much different from code-level details That's the twist: it was supposed to REPLACE the code-level details!
222.
▲
by
al2o3cr
2y ago
You can get inexpensive "programmable" lasers that are controlled by DMX, but that protocol isn't fast enough (250kb/s) to do anything besides pick preprogrammed patterns. If you're up for some hacking / revers
223.
▲
by
al2o3cr
2y ago
A quick skim of the posts on that account shows they're all pointing to your Medium blog: https://news.ycombinator.com/submitted?id=arijo See also from the FAQ: "Please don't use HN primarily for promotion&qu
224.
▲
by
al2o3cr
2y ago
LOL gtfo https://news.ycombinator.com/item?id=40852038 https://news.ycombinator.com/item?id=40847926 https://news.ycombinator.com/item?id=40834134
225.
▲
by
al2o3cr
2y ago
Since Rule 30 is dense in the space of configurations, at some point it produces a binary sequence that spells out "BRAZZERS" in ASCII - does that count as Rule 30 Rule 34? :P
226.
▲
by
al2o3cr
2y ago
TBH the LLM seems worse than useless for a lot of these tasks - entering a netlist from a datasheet is tedious, but CHECKING a netlist that's mostly correct (except for some hallucinated resistors) seems even more tedious.
227.
▲
by
al2o3cr
2y ago
From what I understand Unless you've got experience in international corporation law, I'd recommend you get a LOT more sure about this - otherwise you could end up back here on HN posting "My external CEO stole my w
228.
▲
by
al2o3cr
2y ago
The only person responsible for somebody who thinks "1*1 = 2" is the dumbfuck who believes that nonsense.
229.
▲
by
al2o3cr
2y ago
This video is a good summary: https://www.youtube.com/watch?v=aitVHsg0rWA
230.
▲
by
al2o3cr
2y ago
Nyanning left banking to found a consultancy focused on agentic AI use cases, Agxency.ai. Holy motivated reasoning, Batman!
231.
▲
by
al2o3cr
2y ago
General comment: the README opens with that replaces tables/joins with a lightweight semantic binding layer But then the included example doesn't do any joins. :( Similarly, the "Concepts" section of the documen
232.
▲
by
al2o3cr
2y ago
They certainly aren’t paying us for access to our websites. That's... not how websites work?
233.
▲
by
al2o3cr
2y ago
There’s no shortage of ways I could be wrong. For instance, when you started by believing every LLM marketing pitch...
234.
▲
by
al2o3cr
2y ago
LOL if you think this is a new thing - check out the "child beauty pageant" scene that's featured parents pimping out their kids for DECADES.
235.
▲
by
al2o3cr
2y ago
All this needs is a big "definitely not the feds THIS TIME" sign :P https://www.inc.com/jennifer-conrad/the-fbi-created-its-own-...
236.
▲
by
al2o3cr
2y ago
the Henry Kissinger Distinguished Professor Checks out.
237.
▲
by
al2o3cr
2y ago
Classic "I have a PhD in economics, so I ALONE understand nuclear physics" not-even-wrong gibberish.
238.
▲
by
al2o3cr
2y ago
"100x improvement" in _memcpy_ performance? On a benchmark ("Matched Parallel Algorithms") where a Xeon is 10x _slower_ than an i7? Their page that should have details is equally vague: https://flow-computing.
239.
▲
by
al2o3cr
2y ago
If you've got $4B to set on fire, you can have $2B of "revenue" selling dollar bills for 50 cents. That doesn't mean that throwing another $20B on the fire is going to suddenly turn profitable. The better question is: di
240.
▲
by
al2o3cr
2y ago
Can every text generated quickly by an algorithm be quickly verified for truthfulness by a human or automated system? The algorithm prints out "all the nontrivial zeros of the Riemann zeta function have a real part of 1/
More ›