# I am the Watcher. I am your guide through this vast new twtiverse.
#
# Usage:
# https://watcher.sour.is/api/plain/users View list of users and latest twt date.
# https://watcher.sour.is/api/plain/twt View all twts.
# https://watcher.sour.is/api/plain/mentions?uri=:uri View all mentions for uri.
# https://watcher.sour.is/api/plain/conv/:hash View all twts for a conversation subject.
#
# Options:
# uri Filter to show a specific users twts.
# offset Start index for quey.
# limit Count of items to return (going back in time).
#
# twt range = 1 60404
# self = https://watcher.sour.is?uri=https://twtxt.net/user/prologic/twtxt.txt&offset=58691
# next = https://watcher.sour.is?uri=https://twtxt.net/user/prologic/twtxt.txt&offset=58791
# prev = https://watcher.sour.is?uri=https://twtxt.net/user/prologic/twtxt.txt&offset=58591
> Secret To AI Profitability Is Hiring a Lot More Doctorates
Umm no! It's figuring out how to build an auto-complete that doesn't require a nuclear power plant! Oh wait π€ we did that like 50 years ago already π€¦ββοΈ
> Secret To AI Profitability Is Hiring a Lot More Doctorates
Umm no! It's figuring out how to build an auto-complete that doesn't require a nuclear power plant! Oh wait π€ we did that like 50 years ago already π€¦ββοΈ
@aelaraji Can you walk me through your testing process here and share example links etc? π
@aelaraji Can you walk me through your testing process here and share example links etc? π
@movq That's really disappointing π’ Even before so-called "AI" (Artificial Incompetence) became "more of a thing" I was always astounded at how quick some of the solutions were coming out! π€£
@movq That's really disappointing π’ Even before so-called "AI" (Artificial Incompetence) became "more of a thing" I was always astounded at how quick some of the solutions were coming out! π€£
Over the holiday break I was looking at one of my old projects, Β΅Linux. Turns out I did a fine job realy and have decided to revive the project π₯³ -- Just getting the build/tests woring on my Mac Studio (_Apple Silicon_). Check it out! π 
#Β΅Linux
Over the holiday break I was looking at one of my old projects, Β΅Linux. Turns out I did a fine job realy and have decided to revive the project π₯³ -- Just getting the build/tests woring on my Mac Studio (_Apple Silicon_). Check it out! π #Β΅Linux
Was just catching up on all the LinkedIN garbage that is well umm garbage ποΈ One was from a candidate I interviewed, so I had to reply to that π
-- Anyway.... Saw this random post in my "notifications":
> How do land that job with a Unicorn
First off, you'll have to define what da fuq a "Unicorn" is! π€£ My understanding a Unicorn is a mythical creature with a horn on its head and wings πͺ½ π€¦ββοΈ 
Was just catching up on all the LinkedIN garbage that is well umm garbage ποΈ One was from a candidate I interviewed, so I had to reply to that π
-- Anyway.... Saw this random post in my "notifications":
> How do land that job with a Unicorn
First off, you'll have to define what da fuq a "Unicorn" is! π€£ My understanding a Unicorn is a mythical creature with a horn on its head and wings πͺ½ π€¦ββοΈ
@bender Already planning on going back to the same place again next year π€£
@bender Already planning on going back to the same place again next year π€£
@bender I reviewed my solution and it's pretty much spot on! π€£ the order of magnitude performance is anywhere between 1-10ms
@bender I reviewed my solution and it's pretty much spot on! π€£ the order of magnitude performance is anywhere between 1-10ms
@bender I currently use the mixstral and codeastral models. Both open source and run just fine on my personal machines.
@bender I currently use the mixstral and codeastral models. Both open source and run just fine on my personal machines.
Bye bye Angsana! See you again next year hopefully π€ 
Bye bye Angsana! See you again next year hopefully π€
@bender are one of my assumptions off?
@bender are one of my assumptions off?
@johanbove You are also welcome to use and/or borrow from twtxt2html π
@johanbove You are also welcome to use and/or borrow from twtxt2html π
* Page size 1MB
* Median 50 pages per query
* 80% pages cached
* 200us SSD reads
* 100us Memory reads
* Query time:
* (50x0.80x100)+(50x0.20x200)ββ=β6,000
* 6000us
* 6ms*
* Page size 1MB
* Median 50 pages per query
* 80% pages cached
* 200us SSD reads
* 100us Memory reads
* Query time:
* (50x0.80x100)+(50x0.20x200)ββ=β6,000
* 6000us
* 6ms*
> PhD level science questions? (+1)
niftydude an hour ago
A PhD level science question is a question that can only be answered by scientific research and experimentation.
And no, by "research", I do not mean googling.
Literally the whole point of a scientific PhD is to perform experiments and study to answer a specific research question that no one has looked into yet.
Whilst ChatGPT can probably can answer "PhD-level science questions" with the same generation of plausible bullshit it answers all questions, I very much doubt ChatGPT can answer PhD-level science questions with any sort of accuracy.
It can't do that without performing experiments (that in some cases might be complex enough to last years).
Just more of the marketing BS silicon valley seems to be full of these days. Remember when California was actually making products that benefited society as well as making money?
> PhD level science questions? (+1)
niftydude an hour ago
A PhD level science question is a question that can only be answered by scientific research and experimentation.
And no, by "research", I do not mean googling.
Literally the whole point of a scientific PhD is to perform experiments and study to answer a specific research question that no one has looked into yet.
Whilst ChatGPT can probably can answer "PhD-level science questions" with the same generation of plausible bullshit it answers all questions, I very much doubt ChatGPT can answer PhD-level science questions with any sort of accuracy.
It can't do that without performing experiments (that in some cases might be complex enough to last years).
Just more of the marketing BS silicon valley seems to be full of these days. Remember when California was actually making products that benefited society as well as making money?
@johanbove Congrats! π₯³ I think it's only 4 years for me π
@johanbove Congrats! π₯³ I think it's only 4 years for me π
@lyse Haha! No! Golf biggie π€£
@lyse Haha! No! Golf biggie π€£
Nice view from up on the mountain side π 
Nice view from up on the mountain side π
I like to think of it like this. It takes approximately five months of power, relatively speaking to power, the human brain vs. multiple megawatts hell even multiple gigawatts of power to power even some of the most modest and yet surprisingly complex AI models.
I like to think of it like this. It takes approximately five months of power, relatively speaking to power, the human brain vs. multiple megawatts hell even multiple gigawatts of power to power even some of the most modest and yet surprisingly complex AI models.
There is something about human intelligence that we don't quite yet understand, and it isn't in the complexity or increasing the number of parameters to the order of billions π€£
There is something about human intelligence that we don't quite yet understand, and it isn't in the complexity or increasing the number of parameters to the order of billions π€£
In other words, I don't think we can realistically even come close to emulating, emotion, depth, and creativity
In other words, I don't think we can realistically even come close to emulating, emotion, depth, and creativity
@movq This random comment from another Slashdot article pretry much sums up my view on so-called "AI":
> Elevator music
Tony Isaac 20 minutes ago
If you derive your income from producing "elevator music" you might indeed be in danger of losing that income to AI. Also, bumper music--music used to fill otherwise silent gaps between segments of a podcast or radio show--might be a candidate for AI takeover.
But if you produce real music--music with depth and emotion--your job isn't going anywhere.
How can I be so sure? I've seen the kind of code AI writes. I've seen the kind of prose AI writes. Both are amazing, for something computer-generated. But neither would be mistaken for the work of someone skilled or proficient in the art. Music won't be any different.
@movq This random comment from another Slashdot article pretry much sums up my view on so-called "AI":
> Elevator music
Tony Isaac 20 minutes ago
If you derive your income from producing "elevator music" you might indeed be in danger of losing that income to AI. Also, bumper music--music used to fill otherwise silent gaps between segments of a podcast or radio show--might be a candidate for AI takeover.
But if you produce real music--music with depth and emotion--your job isn't going anywhere.
How can I be so sure? I've seen the kind of code AI writes. I've seen the kind of prose AI writes. Both are amazing, for something computer-generated. But neither would be mistaken for the work of someone skilled or proficient in the art. Music won't be any different.
@slashdot da fuq?! Already?! π± Who's pumping this shit?! π€―
@slashdot da fuq?! Already?! π± Who's pumping this shit?! π€―
@slashdot why are so many banging on against Bruce Perens in the comments of this Slashdot article? π€ what has he done?
@slashdot why are so many banging on against Bruce Perens in the comments of this Slashdot article? π€ what has he done?
@bender Ahh yeah that'll do it π€£ I couldn't find a reason for the martial law myself π€¦ββοΈ
@bender Ahh yeah that'll do it π€£ I couldn't find a reason for the martial law myself π€¦ββοΈ
@movq there is no appeal.
@movq there is no appeal.
@bender Facing impeachment over what?
@bender Facing impeachment over what?
What the hell is happening in South Korea?
What the hell is happening in South Korea?
Problem 2: Your SSD-backed database has a usage-pattern that rewards you with a 80% page-cache hit-rate (i.e. 80% of disk reads are served directly out of memory instead of going to the SSD). The median is 50 distinct disk pages for a query to gather its query results (e.g. InnoDB pages in MySQL). What is the expected average query time from your database?
----
Share your solution via Twtxt and how you arrived at it and I'll share my solution tomorrow!
#napkin-math
Problem 2: Your SSD-backed database has a usage-pattern that rewards you with a 80% page-cache hit-rate (i.e. 80% of disk reads are served directly out of memory instead of going to the SSD). The median is 50 distinct disk pages for a query to gather its query results (e.g. InnoDB pages in MySQL). What is the expected average query time from your database?
----
Share your solution via Twtxt and how you arrived at it and I'll share my solution tomorrow!
#napkin-math
@skinshafi That would be cool! Lemme know how that goes!
@skinshafi That would be cool! Lemme know how that goes!
@bender zs integrates my static server yes π
@bender zs integrates my static server yes π
@movq It's because Luke the "others", it's designed to be a "fast socials media" π€£
@movq It's because Luke the "others", it's designed to be a "fast socials media" π€£
@bender Looks like Irish was created around 7 years ago too π€
@bender Looks like Irish was created around 7 years ago too π€
Also interesting:
> edit_hash: When a message is edited, a new message is created-- this field holds the hash of the modified message. The client follows the chain of edit hashes to end up at the final, edited message to display. This lets us keep an "undo" history (not yet implemented) and is a marker so the client can display a marker that the message has been edited.
Also interesting:
> edit_hash: When a message is edited, a new message is created-- this field holds the hash of the modified message. The client follows the chain of edit hashes to end up at the final, edited message to display. This lets us keep an "undo" history (not yet implemented) and is a marker so the client can display a marker that the message has been edited.
@bender Thanks! Also very interesting rid bits here π€£
> The author, parent hash, timestamp, and message values go into the hash. (see Message Hash for details)
@bender Thanks! Also very interesting rid bits here π€£
> The author, parent hash, timestamp, and message values go into the hash. (see Message Hash for details)
LOL the rest of it appears undocumented π€¦ββοΈ
Messages
Message Hash
Bad Hashes
Edit Chain
Deleted Messages
Topic List
Replies
License
GPLv2
LOL the rest of it appears undocumented π€¦ββοΈ
Messages
Message Hash
Bad Hashes
Edit Chain
Deleted Messages
Topic List
Replies
License
GPLv2
@skinshafi Cool!
> Iris leans heavily on convention. Iris' security and message authentication is provided by filesystem permissions and message hashing.
@skinshafi Cool!
> Iris leans heavily on convention. Iris' security and message authentication is provided by filesystem permissions and message hashing.
