LambdaDuck (@Anka213) header
LambdaDuck (@Anka213) profile picture

Functional programming and type theory enthusiast #Haskell #Agda. They/Them mastodon: https://teapawt.net/t.co/yi842YCJhb bluesky: @anka213.bsky.social

Joined May 2011
Photo by LambdaDuck (@Anka213)Photo by LambdaDuck (@Anka213)Photo by LambdaDuck (@Anka213)Photo by LambdaDuck (@Anka213)Photo by LambdaDuck (@Anka213)Photo by LambdaDuck (@Anka213)
Parser in a tweet: import Control.Monad.State import Control.Applicative import Data.List data T = T [T] deriving Show ch c = mfilter (== c) $ StateT uncons parens = T <$ ch '(' <*> many parens <* ch ')' parse = runStateT -- $> parse parens "(()())" -- Just (T [T [],T []],"")
16
5
1
137
this sketch is evergreen: (english subtitles available) m.youtube.com/watch?v=zvgZtd…
Frankly, they need better dirty tricks. The “everyone is Hitler” attack is sooo tired 😴
2
9,526
As of this post, we're now active on Bluesky as well! bsky.app/profile/roborosewat…
1
2
17
6,445
Replying to @TheFigen_
Real race
13
45
7
475
47,872
the best summary i’ve seen on how and why the modern tech market became shit youtube.com/watch?v=4EmstuO0…
2
3
1,733
8
81
7
910
30,857
this makes me wonder how well they managed to protect it against indirect prompt injection? they probably gave it zero privileges, but that doesn’t necessarily make it safe. at the very least you can make it give the wrong summary
Replying to @rowancheung
3. Apple Intelligence automatically summarizes your emails and notifications for easier reading. It also brings your most important notifications to the top of your stack, helping you focus on what is important in the moment.
1
1,230
For the purposes of enabling local reasoning and simplifying formal verification of software, you don't actually need immutability, what you need is no _mutable aliasing_. Here's a master's thesis which uses this equivalence for verification of Rust code: github.com/Kha/electrolysis
1
592
Here is an example of how such translation from linear mutation to immutability can happen: teapawt.net/Anka213/status/1824735…
any program with no shared mutability can trivially be translated into an equivalent program with only immutable data and therefore be reasoned about similarly. let mut x = ...; f(&mut x); g(&x); is equivalent to let x = ...; let x1 = f(x1); g(&x1);
501
No one’s doing it so I guess I gotta promote myself (it hurts)
what's the funniest tiktok video you've ever seen?
305
6,260
337
82,592
6,192,206
LambdaDuck retweeted
haskell is the cutest programming language (^+^)
35
214
27
2,480
82,191
It's done — the start of something great. The first part of Pragmatic Category Theory for Beginners has been published! dev.to/chshersh/pragmatic-ca… Thanks to everyone who reviewed the draft and shared the feedback. I'll work on the video version next.
22
71
8
452
131,209
Can you really call yourself a Software Engineer if you don't know at least basic Category theory?
“You don’t need to know math to be an SWE” “You don’t need to know CS to be an SWE” “You don’t need to know Category Theory to be an SWE” “You don’t need to know many PLs to be an SWE” “You don’t need to know the latest framework to be an SWE” Do SWEs know anything 😭
2
393
LambdaDuck retweeted
the title of this email always cracks me up
18
440
7
9,916
133,173
I guess it’s open season on any Microsoft product then?
235
2,862
182
18,539
386,821
If you've ever wondered how the OOM-Killer works, wonder no more!
13
64
6
602
49,823
Sci-Fi Author: In my book I invented the Torment Nexus as a cautionary tale Tech Company: At long last, we have created the Torment Nexus from classic sci-fi novel Don't Create The Torment Nexus
303
30,557
2,825
115,252
Watching in real time as "slop" becomes a term of art. the way that "spam" became the term for unwanted emails, "slop" is going in the dictionary as the term for unwanted AI generated content
it’s cool how every google search now starts with a wall of LLM slop that is completely useless and takes up half the screen
29
294
58
2,690
320,898
Replying to @svpino
Llama 3 had an interesting response...wait until the end!
58
16
12
690
203,151
9
67
497
19,013
I think I got my answer - "LLMs are trained with 200,000 years worth of reading material and are still pretty dumb." teapawt.net/ylecun/status/17842103…
As long as AI systems are trained to reproduce human-generated data (e.g. text) and have no search/planning/reasoning capability, performance will saturate below or around human level. Furthermore, the amount of trials needed to reach that level will be far larger than the amount of trials needed to train humans. LLMs are trained with 200,000 years worth of reading material and are still pretty dumb. Their usefulness resides in their vast accumulated knowledge and language fluency. But they are still pretty dumb.
1
2
377