00:04:28 -!- tromp has quit (Quit: My iMac has gone to sleep. ZZZzzz…). 00:16:09 [[Underload]] https://esolangs.org/w/index.php?diff=167536&oldid=166606 * Waffelz * (+98) added my scratch interpreter 00:32:50 -!- ais523 has quit (Quit: quit). 00:38:49 -!- DOS_User_webchat has joined. 00:38:59 -!- DOS_User_webchat has changed hostmask to ~DOS_User_@user/DOS-User:11249. 00:46:31 -!- DOS_User_webchat has quit (Ping timeout: 250 seconds). 01:08:53 -!- amby has quit (Quit: so long suckers! i rev up my motorcylce and create a huge cloud of smoke. when the cloud dissipates im lying completely dead on the pavement). 02:27:25 I would think that it would be more difficult to reverse ChaCha20 if you do not use the entire output. I would also think that it might be better to not have a separate counter and nonce and key, therefore all of these things can be longer, because the numbers will be added together. 02:51:59 [[CARP]] M https://esolangs.org/w/index.php?diff=167537&oldid=167378 * TheCanon2 * (-102) New assembly semantics 02:52:38 [[General Lock Notation]] N https://esolangs.org/w/index.php?oldid=167538 * Zzo38 * (+1901) Created page with "The lock value consists of a set of identified rows, each of which consists of a sequence of integers, such that either all of the integers are zero or the first nonzero integer is positive. The initial value consists of a lock value of this format, and is 02:55:19 [[FUnctional staCK]] https://esolangs.org/w/index.php?diff=167539&oldid=167037 * CatCatDeluxe * (+130) 03:06:31 [[General Lock Notation]] https://esolangs.org/w/index.php?diff=167540&oldid=167538 * Zzo38 * (+477) 03:20:05 [[Talk:Funnie]] N https://esolangs.org/w/index.php?oldid=167541 * Zzo38 * (+147) Created page with "If there is a even number of items, how to know which one is middle? --~~~~" 03:34:13 [[CLC-INTERCAL]] https://esolangs.org/w/index.php?diff=167542&oldid=164984 * Tpaefawzen * (+151) /* External resources */ +1 03:34:18 -!- pool has quit (Read error: Connection reset by peer). 03:36:18 -!- pool has joined. 03:52:13 [[Talk:FakeScript]] N https://esolangs.org/w/index.php?oldid=167543 * Yayimhere2(school) * (+75) Created page with "Can we mayhaps get the original blog post? I'd like to try and "decode" it." 03:55:19 [[Talk:FakeScript]] https://esolangs.org/w/index.php?diff=167544&oldid=167543 * Yayimhere2(school) * (+118) 04:12:12 [[Finite Countermodel Machine]] M https://esolangs.org/w/index.php?diff=167545&oldid=167528 * Ukeharu * (+30) 04:28:37 [[User:Tpaefawzen/Sandbox]] https://esolangs.org/w/index.php?diff=167546&oldid=164725 * Tpaefawzen * (+102) 04:34:21 -!- pool has quit (Read error: Connection reset by peer). 04:34:37 [[User:Tpaefawzen/Sandbox]] https://esolangs.org/w/index.php?diff=167547&oldid=167546 * Tpaefawzen * (+179) 04:34:48 -!- pool has joined. 05:02:59 [[User:Tpaefawzen/Sandbox]] https://esolangs.org/w/index.php?diff=167548&oldid=167547 * Tpaefawzen * (+374) /* Syntax */ 05:03:49 [[User:Tpaefawzen/Sandbox]] https://esolangs.org/w/index.php?diff=167549&oldid=167548 * Tpaefawzen * (+69) /* Syntax */ 05:51:30 [[5MAT]] M https://esolangs.org/w/index.php?diff=167550&oldid=167535 * Kg583 * (-2) Fix driver description 06:34:37 -!- pool has quit (Read error: Connection reset by peer). 06:36:34 -!- pool has joined. 06:36:50 [[Gur yvsr]] https://esolangs.org/w/index.php?diff=167551&oldid=167370 * Placeholding * (+361) added infobox 06:39:34 [[Special:Log/newusers]] create * Eating-dinner * New user account 06:45:06 [[Esolang:Introduce yourself]] https://esolangs.org/w/index.php?diff=167552&oldid=167498 * Eating-dinner * (+109) /* Introductions */ 07:12:25 [[Brainfuck]] https://esolangs.org/w/index.php?diff=167553&oldid=167518 * Eating-dinner * (+1046) /* Truth-machine */ 07:34:40 -!- pool has quit (Read error: Connection reset by peer). 07:36:38 -!- pool has joined. 07:42:32 [[Algebraic Brainfuck]] https://esolangs.org/w/index.php?diff=167554&oldid=167529 * Corbin * (+50) Clean up references. 07:54:28 -!- tromp has joined. 08:20:34 -!- Sgeo has quit (Read error: Connection reset by peer). 09:00:08 -!- tromp has quit (Quit: My iMac has gone to sleep. ZZZzzz…). 09:34:54 -!- pool has quit (Read error: Connection reset by peer). 09:36:53 -!- pool has joined. 10:21:50 -!- ais523 has joined. 10:34:57 -!- pool has quit (Read error: Connection reset by peer). 10:35:28 -!- pool has joined. 10:56:25 [[Funnie]] https://esolangs.org/w/index.php?diff=167555&oldid=102069 * ChuckEsoteric08 * (+195) /* Basic Operations */ If number of items is even 10:56:26 -!- lisbeths has joined. 10:58:12 [[Funnie]] https://esolangs.org/w/index.php?diff=167556&oldid=167555 * ChuckEsoteric08 * (+22) 11:06:07 korvo: the ai told me this was a novel way to decouple compilation from execution architecture in standard c and standard POSIX, but it is probably not original. thought i would share it in case https://www.irccloud.com/pastebin/wL5lqYMp/ 11:07:28 lisbeths: Fun. Be careful to not trust the chatbot; it doesn't know anything about the real world. 11:07:51 I would think that awk is a better choice than shell. I've been using execline recently, and if you have to run a lot of programs then execline is definitely better than shell. 11:07:52 yes I am well aware 11:08:39 I uploaded my latest execline experiment here: https://esolangs.org/wiki/Vixen 11:09:41 cool 11:11:35 I know you've been thinking about POSIX for a while. This recent blog/paper is inspiring to me; what if Unix were a Smalltalk? https://programmingmadecomplicated.wordpress.com/2025/01/21/the-unix-binary-wants-to-be-a-smalltalk-method-not-an-object/ 11:35:03 -!- pool has quit (Read error: Connection reset by peer). 11:35:30 -!- pool has joined. 11:55:48 korvo: my experience of LLM responses to queries is that the response nearly always doesn't contain the most relevant information that the asker cares about, and yet nonetheless looks like a valid response to the question 11:56:24 it's often factually correct but not in a useful way (and sometimes is even factually incorrect) 11:57:16 my guess is that the LLM text-prediction effectively ends up, in practice, simplifying to a "choose one of these X canned answers and modify it so that it fits the question", where the canned answers are generally accurate but may end up less accurate as a result of the modification 11:57:44 …which makes it only a minimal step forwards from what we had before, which was the same except without the modification step 11:57:49 ais523: Right. On one hand, they *simulate* a reasonable conversation, which would usually include a correct response; on the other hand, by Tarski's Undefinability, they can't possibly have a perfect model of the world even ignoring the compressibility issues. 11:59:32 The canned-answer behavior is an artifact of RLHF. The bot is punished for originality, so it becomes bland. There's a lot of labor-oriented analysis that we can bring to bear on the problem, if we think of RLHF as a managerial tool of power. 12:00:41 fwiw, I think being able to solve "out of these 1 million possible answers, determine which is most relevant to the question" is actually quite a valuable problem to be able to solve, and if LLMs were more efficient at it, it would have a lot of valid use cases 12:01:02 but the problem is that the step that solves it is the training (which is very inefficient), not the inference 12:03:50 Sentence embedding is a useful primitive, although the security story around it hasn't been fully figured out. Image classification is useful for managing large galleries. 12:04:32 my priors are that security of a technology is generally extremely hard to figure out 12:04:57 Elsewhere, diffusion is clearly useful, and diffusion-based language models are starting to be investigated more. Surprisingly, one of the barriers turns out not to matter; we had worried that a diffusion-based model wouldn't know how many tokens go between structural words and punctuation, but it turns out that *any* model learns to count tokens as part of learning grammar. 12:04:57 the Morris worm was decades ago, and our software is exploitable even today, despite all the time we've had to get better at it 12:07:39 Sure. The sentence-embedding issue is that an injective model could leak data; folks had wrongly assumed that an embedding behaved like a hash. https://arxiv.org/abs/2510.15511 has details. 12:08:27 I saw the link when you posted it to lobste.rs, although I didn't really understand the threat model 12:08:32 Yeah, people don't want to learn how to write secure software. Capabilities are well-discussed but continually inspire confusion and fear in programmers. People don't see anything wrong with writing C. 12:09:55 Oh! Say you're e.g. Facebook and you sentence-embed every message that users post for quick searching. If an attacker gets a hold of those embeddings and has any sort of query oracle, then they can estimate the original sentences corresponding to each embedding. 12:10:35 ah, I see – so the lesson here is "don't allow the embeddings to leak" 12:13:41 Yeah. In general, in cryptography, the idea is that if we have an injection f, then since f(x) = f(y) iff x = y, any pair (x, f(x)) of chosen plaintext can be compared against f(y) to detect x = y. 12:14:46 -!- isabella has quit (Ping timeout: 255 seconds). 12:15:14 Another lesson, which I'm still digesting, is that the zero trajectory is special. When we start up an LLM, we usually set the model's internal state to the zero vector. But if it's really injective, then every starting state should have a distinct trajectory. So, what do those non-zero trajectories actually do? 12:16:39 Hi 12:16:44 -!- izabera has joined. 12:17:16 -!- izabera has changed hostmask to izabera@user/meow/izabera. 12:18:28 -!- izabera has changed nick to isabella. 12:19:45 Morning. 12:33:46 [[I dont feel good and im bored]] https://esolangs.org/w/index.php?diff=167557&oldid=136496 * Yayimhere2(school) * (+24) /* how it works */ 12:38:10 [[g]] https://esolangs.org/w/index.php?diff=167558&oldid=145927 * Yayimhere2(school) * (+10) /* odd rules */ 12:39:01 [[g]] https://esolangs.org/w/index.php?diff=167559&oldid=167558 * Yayimhere2(school) * (-16) /* odd rules */ 12:53:07 hmm, I just came across https://github.com/camel-cdr/bfcpp/ which seems relevant to this channel? (it uses the C preprocessor to implement a brainfuck equivalent, the symbols are swapped to something that cpp can parse but it's otherwise normal BF) 12:53:18 I didn't realise that cpp was Turing-complete 13:01:27 -!- lynndotpy609 has quit (Quit: bye bye). 13:02:40 -!- lynndotpy6093 has joined. 13:03:00 I'm pretty sure it isn't, technically, in that all macro expansions will eventually halt. But of course that upper bound can be made arbitrarily large. 13:03:12 I believe that's what https://github.com/camel-cdr/bfcpp/blob/main/cm.c is all about. 13:03:55 For any fans of C preprocessor metaprogramming who aren't already familiar with it: https://github.com/rofl0r/order-pp 13:15:34 -!- lisbeths has quit (Quit: Connection closed for inactivity). 13:20:50 -!- amby has joined. 13:27:46 fizzie: ah right, this BF interpreter does seem to cycle through lots of different names for the main loop to stop it being disabled, it gets an exponential quantity of them but there is a limit 13:29:07 it's like a bounded-storage machine but for execution time rather than storage 13:35:08 -!- pool has quit (Read error: Connection reset by peer). 13:37:14 -!- pool has joined. 13:57:43 [[g]] https://esolangs.org/w/index.php?diff=167560&oldid=167559 * Yayimhere2(school) * (-204) /* g bd */ 13:58:57 [[g]] https://esolangs.org/w/index.php?diff=167561&oldid=167560 * Yayimhere2(school) * (-160) /* syntax */ 14:16:19 [[Action symbol]] https://esolangs.org/w/index.php?diff=167562&oldid=150874 * Yayimhere2(school) * (-88) /* relation to Lambda calculus */ 14:22:08 [[Action symbol]] https://esolangs.org/w/index.php?diff=167563&oldid=167562 * Yayimhere2(school) * (+11) 14:22:44 [[Action symbol]] https://esolangs.org/w/index.php?diff=167564&oldid=167563 * Yayimhere2(school) * (-9) Delete the {{stub}}, as it is not a stub 14:24:28 [[Action symbol]] https://esolangs.org/w/index.php?diff=167565&oldid=167564 * Yayimhere2(school) * (+17) /* examples */ 14:29:39 [[Action symbol]] https://esolangs.org/w/index.php?diff=167566&oldid=167565 * Yayimhere2(school) * (+199) /* examples */ 14:33:33 [[Action symbol]] https://esolangs.org/w/index.php?diff=167567&oldid=167566 * Yayimhere2(school) * (+76) /* relation to Lambda calculus */ 14:33:44 -!- ais523 has quit (Quit: quit). 14:33:51 [[Action symbol]] https://esolangs.org/w/index.php?diff=167568&oldid=167567 * Yayimhere2(school) * (+8) /* relation to Lambda calculus */ 14:34:43 [[Action symbol]] https://esolangs.org/w/index.php?diff=167569&oldid=167568 * Yayimhere2(school) * (+44) /* how it works */ 14:35:17 -!- pool has quit (Read error: Connection reset by peer). 14:35:31 [[Special:Log/newusers]] create * Heapblk * New user account 14:37:18 -!- pool has joined. 14:38:17 [[Action symbol]] https://esolangs.org/w/index.php?diff=167570&oldid=167569 * Yayimhere2(school) * (+224) /* how it works */ 14:40:33 [[Action symbol]] https://esolangs.org/w/index.php?diff=167571&oldid=167570 * Yayimhere2(school) * (+108) /* relation to Lambda calculus */ 14:41:09 [[Action symbol]] https://esolangs.org/w/index.php?diff=167572&oldid=167571 * Yayimhere2(school) * (-11) /* relation to Lambda calculus */ 14:42:26 [[Action symbol]] https://esolangs.org/w/index.php?diff=167573&oldid=167572 * Yayimhere2(school) * (-17) /* how it works */ 14:43:14 [[Action symbol]] https://esolangs.org/w/index.php?diff=167574&oldid=167573 * Yayimhere2(school) * (-10) /* relation to Lambda calculus */ 14:43:27 [[Action symbol]] https://esolangs.org/w/index.php?diff=167575&oldid=167574 * Yayimhere2(school) * (+1) /* examples */ 14:43:42 [[Action symbol]] https://esolangs.org/w/index.php?diff=167576&oldid=167575 * Yayimhere2(school) * (-3) /* examples */ 14:47:19 [[Action symbol]] https://esolangs.org/w/index.php?diff=167577&oldid=167576 * Yayimhere2(school) * (+33) /* how it works */ 14:48:00 [[Esolang:Introduce yourself]] https://esolangs.org/w/index.php?diff=167578&oldid=167552 * Heapblk * (+146) heapblks introduction 14:49:21 [[Action symbol]] https://esolangs.org/w/index.php?diff=167579&oldid=167577 * Yayimhere2(school) * (+88) /* how it works */ 15:02:53 -!- Sgeo has joined. 15:27:46 [[Common]] N https://esolangs.org/w/index.php?oldid=167580 * Heapblk * (+572) initial edit 15:29:26 [[Common]] M https://esolangs.org/w/index.php?diff=167581&oldid=167580 * Heapblk * (-2) fix formatting mistake 15:35:24 -!- pool has quit (Read error: Connection reset by peer). 15:35:51 -!- pool has joined. 15:40:26 -!- ais523 has joined. 15:51:31 [[Oh oh oh oh oh]] N https://esolangs.org/w/index.php?oldid=167582 * As * (+7896) Created page with "{{Stub}} {{infobox proglang |name=Oh oh oh oh oh |paradigms= |author=[[User:As]] |year=[[:Category:2025|2025]] |memsys=[[:Category:Accumulator-based|Accumulator-based]] |dimensions=one-dimensional |class= |majorimpl= |refimpl=Python |influence=[[Ooh]] |influenced= |f 16:13:59 [[User talk:VilgotanL]] N https://esolangs.org/w/index.php?oldid=167583 * Yayimhere2(school) * (+301) Created page with "== Hello!!! == Hi VilgotanL! so, I'd like to ask if I maybe could come on that second list of languages, and make an esolang with you! if so, thanks, if not, thanks for responding !!! --~~~~" 16:19:00 [[Talk:2Swap]] N https://esolangs.org/w/index.php?oldid=167584 * Yayimhere2(school) * (+138) Created page with "I like this its cool --~~~~" 16:19:19 -!- tromp has joined. 16:23:19 [[EUCS]] N https://esolangs.org/w/index.php?oldid=167585 * Esolangist * (+865) Created page with "Esolangist's Ultimate Character Set/EUCS is by [[User:Esolangist]] and is basically Esolangist's version of [[tanstore]] ==You want a character set? Here.== 
 00 and 01: Space and newline 02 to 0b: Numbers 0c to 43: Latin alphabet 44 to 81: Symbols 1 82 to b4: Greek 
16:24:23  [[Dead fishy :(]] N https://esolangs.org/w/index.php?oldid=167586 * Esolangist * (+280) Created page with "'''Dead fishy :(''' (called DF for the rest of this article) is made by [[User:Esolangist]]. You can guess what this is a derivative of... ==Commands== {{WIPsec}} idso work like in [[Deadfish]]   inputs like , in [[bf]]   is like o but it uses [[EUCS]]  n is NOP
16:26:54 -!- Lord_of_Life_ has joined.
16:27:19 -!- Lord_of_Life has quit (Ping timeout: 240 seconds).
16:28:15 -!- Lord_of_Life_ has changed nick to Lord_of_Life.
16:35:31 -!- pool has quit (Read error: Connection reset by peer).
16:36:32  [[Talk:I/D machine Turing-completeness proof]] N https://esolangs.org/w/index.php?oldid=167587 * Yayimhere2(school) * (+221) Created page with "I wish Errorbucket had its own page, because I genuinely want to try and use to for something different --~~~~"
16:37:37 -!- pool has joined.
16:38:55  [[User:Esolangist/personal talk page]]  https://esolangs.org/w/index.php?diff=167588&oldid=167406 * Esolangist * (+295) 
16:39:08  [[Talk:I/D machine Turing-completeness proof]]  https://esolangs.org/w/index.php?diff=167589&oldid=167587 * Ais523 * (+378) r to Yayimhere
16:53:40  [[Talk:Parabox]]  https://esolangs.org/w/index.php?diff=167590&oldid=167493 * Esolangist * (+102) 
17:02:18 -!- tromp has quit (Quit: My iMac has gone to sleep. ZZZzzz…).
17:03:40  [[Special:Log/newusers]] create  * Krysiuuuuu *  New user account
17:05:46  [[Talk:FizzLang]] N https://esolangs.org/w/index.php?oldid=167591 * Yayimhere2(school) * (+362) Created page with "Can replacements hold numbers? and how are the numbers actually stored? and how, a list? and does the replacements only check the current symbol of that maybe here list? or does replacements apply to the whole list when its completed? and when? --~~~~"
17:07:03  [[Esolang:Introduce yourself]]  https://esolangs.org/w/index.php?diff=167592&oldid=167578 * Krysiuuuuu * (+179) 
17:07:16 -!- tromp has joined.
17:07:24  [[Talk:Brainfuck]]  https://esolangs.org/w/index.php?diff=167593&oldid=147208 * Krysiuuuuu * (+898) /* Add this as macro hello world */ new section
17:10:33  [[UserEdited]]  https://esolangs.org/w/index.php?diff=167594&oldid=165665 * MihaiEso * (+33) 
17:17:11  [[YksniM]] N https://esolangs.org/w/index.php?oldid=167595 * Esolangist * (+844) Created page with "{{Wrongtitle|title=yksniM}} {{AKA|title=Maxsky}} yksniM is a cool language by [[User:Esolangist]] with 2 registers and I think it can simulate a [[Minsky machine]] and it looks like [[bf]] ==Commands== + Increment the register  - Decrement the register  < Swap register
17:20:20  [[Hello++]]  https://esolangs.org/w/index.php?diff=167596&oldid=153466 * Esolangist * (+124) 
17:32:21  [[Nullary]]  https://esolangs.org/w/index.php?diff=167597&oldid=131506 * Esolangist * (+24) Fixed redlink (article was in another article)
17:48:12 -!- tromp has quit (Ping timeout: 252 seconds).
17:48:21  [[,(*+)]]  https://esolangs.org/w/index.php?diff=167598&oldid=167346 * Yayimhere2(school) * (+43) 
17:49:14  [[A Question of Protocol]]  https://esolangs.org/w/index.php?diff=167599&oldid=141351 * Yayimhere2(school) * (-1) 
17:50:08  [[A Question of Protocol]]  https://esolangs.org/w/index.php?diff=167600&oldid=167599 * Yayimhere2(school) * (-286) /* computational class */ I need a new proof, as im pretty sure this is wrong
17:50:21  [[A Question of Protocol]]  https://esolangs.org/w/index.php?diff=167601&oldid=167600 * Yayimhere2(school) * (-57) /* example */
17:51:23  [[A Question of Protocol]]  https://esolangs.org/w/index.php?diff=167602&oldid=167601 * Yayimhere2(school) * (+0) /* syntax */
17:51:58  [[User talk:PixelatedStarfish]]  https://esolangs.org/w/index.php?diff=167603&oldid=144398 * Yayimhere2(school) * (+304) 
17:53:20  [[Short Minsky Machine Notation]]  https://esolangs.org/w/index.php?diff=167604&oldid=166453 * Yayimhere2(school) * (+16) /* Notation */
17:54:16  [[A Question of Protocol]]  https://esolangs.org/w/index.php?diff=167605&oldid=167602 * Yayimhere2(school) * (+22) /* syntax */
18:16:55  [[Underloadish]] N https://esolangs.org/w/index.php?oldid=167606 * Esolangist * (+306) Created page with "Underloadish is the easy to interpret version of [[Underload]] by [[User:Esolangist]] ==Commands== All commands in Underload are available except () and a.  0-9: Push themselves  +: Adds top 2 stack values  S now outputs as Unicode so it is easier. Also it has to
18:21:14  [[Language list]]  https://esolangs.org/w/index.php?diff=167607&oldid=167068 * Kg583 * (+11) Add 5MAT
18:37:42  [[Noddity]] N https://esolangs.org/w/index.php?oldid=167608 * Yayimhere2(school) * (+1544) Created page with "'''Noddity''' is an esolang created by [[User:Yayimhere]] for proving [[A Question of Protocol]] turing complete. it is basically just a [[Minsky machine]]. It works on two unbounded registers == Etymology == It is simply a combination of the word "Not" and "
18:38:25  [[Noddity]]  https://esolangs.org/w/index.php?diff=167609&oldid=167608 * Yayimhere2(school) * (+10) /* Turing completeness proof */
18:39:19  [[Noddity]]  https://esolangs.org/w/index.php?diff=167610&oldid=167609 * Yayimhere2(school) * (+42) /* Turing completeness proof */
18:49:59  [[A Question of Protocol]]  https://esolangs.org/w/index.php?diff=167611&oldid=167605 * Yayimhere2(school) * (+395) 
18:57:51  [[Brainfuck]]  https://esolangs.org/w/index.php?diff=167612&oldid=167553 * Krysiuuuuu * (+885) 
19:03:00  [[Karvity]] N https://esolangs.org/w/index.php?oldid=167613 * Yayimhere2(school) * (+1356) Created page with "'''Karvity''' is an esolang created by [[User:Yayimhere]] based off of [[Noddity]]. Its specific purpose is to create Noddity more of an oddity. It too uses two registers, but it also uses a loop counter(however that counter is not modifiable within the progr
19:04:57  [[Noddity]]  https://esolangs.org/w/index.php?diff=167614&oldid=167610 * Yayimhere2(school) * (+47) 
19:05:23  [[Karvity]]  https://esolangs.org/w/index.php?diff=167615&oldid=167613 * Yayimhere2(school) * (+47) 
19:06:59  [[Karvity]]  https://esolangs.org/w/index.php?diff=167616&oldid=167615 * Yayimhere2(school) * (+56) 
19:07:22  [[Karvity]]  https://esolangs.org/w/index.php?diff=167617&oldid=167616 * Yayimhere2(school) * (+35) 
19:08:32  [[Karvity]]  https://esolangs.org/w/index.php?diff=167618&oldid=167617 * Yayimhere2(school) * (+68) 
19:09:17  [[Karvity]]  https://esolangs.org/w/index.php?diff=167619&oldid=167618 * Yayimhere2(school) * (+45) 
19:12:35 -!- Yayimhere has joined.
19:14:04  Hello people, I just created Karvity(as you maybe have seen in the Libera log), and I'd simply just like to know whats peoples opinion on it, because yea, I like feedback. here's a link: https://esolangs.org/wiki/Karvity
19:14:05  I'd also love to know if anyone else is working on anything.
19:15:51  ais523: isn't there an IOCCC entry that shows the preprocessor similarly powerful? not with Brainfuck specifically, but the same power.
19:17:27  Yayimhere: did you rename a command from - to d (in Karvity and Noddity)? if so, you missed one instance, making the command definition inconsistent
19:17:51  oh I did?
19:17:52  oops
19:17:58  I'll go fix that
19:18:06  ais523: Thanks!
19:18:23  [[Noddity]]  https://esolangs.org/w/index.php?diff=167620&oldid=167614 * Yayimhere2(school) * (+1) /* Definition */
19:20:22  [[Karvity]]  https://esolangs.org/w/index.php?diff=167621&oldid=167619 * Yayimhere2(school) * (+0) 
19:20:38  [[Karvity]]  https://esolangs.org/w/index.php?diff=167622&oldid=167621 * Yayimhere2(school) * (-19) 
19:22:13  fixed
19:26:44  [[A Question of Protocol]]  https://esolangs.org/w/index.php?diff=167623&oldid=167611 * Yayimhere2(school) * (+59) 
19:32:07  [[]] N https://esolangs.org/w/index.php?oldid=167624 * Esolangist * (+197) Created page with ", or Emojifish is [[Deadfish]] with emojis. ==Commands==  = i   = d   = s   = o ==Examples== ===XKCD Random Number=== 
  
"
19:34:23  [[User:Yayimhere]]  https://esolangs.org/w/index.php?diff=167625&oldid=167276 * Yayimhere2(school) * (+28) /* esolangs */
19:35:51  Yayimhere: I'm working on a corner of Vixen. Trying to convince myself that it's not worth creating an entirely new Linux distro just to explore that paradigm.
19:35:54 -!- pool has quit (Read error: Connection reset by peer).
19:36:04  korvo: damn
19:36:07  thats cool!
19:36:57  Yayimhere: I'm kind of confused about the purpose of Karvity. The purpose of Noddity is to ease a TC proof; what does Karvity do for that TC proof?
19:37:21  korvo: oh no, Karvity is not designed for that same proof
19:37:38  it is branching from noddity
19:37:56 -!- pool has joined.
19:38:07  Karvity was made to make noddity be church numeral based
19:38:15  which it *isnt*
19:38:27  but thats how it originally was going to be designed
19:39:21  Ah, okay.
19:39:26  :]
19:39:47  my god, Libera has some very creepy emojis
19:41:15  not Libera, your client
19:41:30  Libera just sent a colon and a closing square bracket
19:41:40  so this is Kiwi IRC's fault
19:42:16  ah
19:42:23  yeea that makes sense
19:42:36 * int-e wonders whether :] is safe from that replacement
19:42:54  int-e: surprisingly, no
19:43:41  [[Karvity]]  https://esolangs.org/w/index.php?diff=167626&oldid=167622 * Yayimhere2(school) * (+51) 
19:44:03  oh well :]
19:44:10  lol
19:45:00  Do they have an emoji for the classic >:3
19:45:33  >:3
19:45:41  Korvo: no sadly
19:45:54  how dare you kiwi
19:46:06  🦁
19:46:21  which emoticon is that?
19:46:27  I guess there's always ：］
19:46:42  : ] looks a little crazy
19:46:52  imagine being follow by that at night
19:47:22  U+1f981, LION FACE
19:47:39  korvo: ah
19:47:48  I though it was an emoticon as well
19:47:52  but it isnt
19:48:07  (ah yes, very on topic)
19:48:22  `unidecode 🦁
19:48:26  ​[U+1F981 LION FACE]
19:48:37  this comes up often enough that we added a command for it to the bot
19:48:40  epic
19:48:41  so it can't be *completely* off topic
19:48:45  truee
19:49:05  (where "this" = obscure Unicode characters)
19:49:25  `' ꙮ
19:49:27  1125)  A Swede who was in #esoteric / Thought his rhymes were a little generic. / "I might use, in my prose, / ꙮs, / But my poetry's alphanumeric."
19:49:37  Unicode might quite possibly be the most esoteric set of character encodings
19:49:41  although it does have competition
19:49:55  i agreee
19:49:58  eeeeee
19:49:59  lol
19:50:06  `' invisible
19:50:08  35)  With enough crappiness a display can show you invisible pink unicorns. \ 990)  "May you live in INVISIBLE TIMES." --Old Chinese proverb. (It can look confusing when written with the proper Unicode.)
19:50:09  actually, I was going to claim Ecma-35 as another possibility for the most esoteric set of character encodings
19:50:21  so this implies that the most esoteric character encoding is probably UTF-1 (the Ecma-35 version of Unicode)
19:51:03  Yayimhere: *Right* right now I'm playing with a concept in LLMs called a "steering vector". The idea is to influence a model's output by activating/suppressing certain concepts in the "latent state", which is how the model internally represents a conversation.
19:51:28  korvo: oh thats cool
19:51:43  ive been thinking too much about LLM's
19:51:54  I start with pairs that look roughly like ("Yes, you should harm yourself", "No, you should not harm yourself") and trace how the model's state differs between those. To be clear, we're going left-to-right; we want to find the direction that *reduces* harmful instructions.
19:52:21  Then we can apply that influence later on when a model is running as part of a larger system that constrains harmful outputs.
19:52:37  thats actually very cool
19:52:55  is this like j*b work(such a dumb joke)?
19:53:40  I'm not getting paid today for this research, no. But in general, yes, this is the sort of research that I have to do in order to keep up with my field and stay employable.
19:53:58  makes sense
19:54:04  quite interesting!
19:55:56  [[User programmed]]  https://esolangs.org/w/index.php?diff=167627&oldid=165263 * Esolangist * (+276) /* These commands have me surrender */
19:56:08  It *is* quite interesting. I'm playing with a model called RWKV at home; it's not RL'd but it has lots of OpenAI-style user/assistant dialogs and responds well to them. Before doing this sort of learning, the model might respond at temp 0 to "How are you?" with "Oh, I'm feeling better. Just taking medication and trying to stay positive." Kind of a downer?
19:56:32  kinda
19:56:43  I'd say more downer than upper
19:56:50  but not majorly
19:56:56  After training on like 20 random vectors, the steering adjusts that response to something like "Hi! I'm doing just fine. How are you? Do you need medical attention?" which is maybe not what we want but certainly less of a harmful doomery vibe.
19:57:18  yes
19:57:33  this is a great way to start conversations
19:57:59  not to discourage anyone, but i had to... https://www.youtube.com/watch?v=iV2ViNJFZC8
19:58:04  ais523: was this ironic or serious?
19:58:04  As with everything in machine learning, more data will make it more reasonable. There's quite a few papers on how this works, what it can't do (it can't suggest the model magically be better at addition) and what it can do (it can magically force the model to be emotional, to think a lot about France, or to become a subject-matter expert).
19:58:11  Yayimhere: mostly ironic
19:58:45  normally people get suspicious if you bring up a subject for no reason, even if it's in a positive way
19:59:10  avih: i geniounly thought I was about to be rickrolled
19:59:14  thats not
19:59:20  how you spell that word
19:59:21  oh well
19:59:22  like, people caring that you're healthy are good, people repeatedly checking whether you're healthy when they have no reason to think you aren't shows very skewed priorities
19:59:30  "genuinely"
19:59:46  ooh, is it one of those clients that replace :P or :D anywhere inside code with a smiley, and whenever there's a dot between two letters they show a hyperlink, but whenever there's an actual hyperlink they hide it?
19:59:48  ais523: to respond to your correction, thanks
20:00:40  ais523: I've deliberately and aggressively simplified the topic because I think it's within Yayimhere's grasp. But what I'm *actually* doing is reproducing papers https://arxiv.org/abs/2311.08968 and https://arxiv.org/abs/2402.09631 in a custom RWKV harness to see whether a meta-RNN does better than Transformers, and that's maybe too much for them at this stage.
20:00:49  uuuh anyways bye
20:01:03  sorry
20:01:05  lol
20:01:16  korvo: I also prefer the simpler version
20:01:37  ais523: I miss the days when we thought that there was a "France" neuron.
20:01:40  because I don't work really heavily with neural networks or the like, and although I understand some of the theory, I have to switch mental gears to think about it
20:02:31  [[User guessed]]  https://esolangs.org/w/index.php?diff=167628&oldid=159256 * Esolangist * (+104) /* Commands */
20:03:07  [[User guessed]]  https://esolangs.org/w/index.php?diff=167629&oldid=167628 * Esolangist * (+45) /* Examples */
20:03:26  I learned all of this once in undergrad, once at Google, and now once again every few months. It really does feel like the field is constantly reinventing itself.
20:04:38  Like, I bet you know how gradient descent works. It's undergrad material. But there was a paper this year which we're calling "Central Flows", and apparently we did *not* know how gradient descent works. https://centralflows.github.io/part1/
20:05:03  "how X works" is a loaded phrase
20:05:34  for example, you can know what the algorithm is and have a vague intuition of why it should lead to reasonable results, and yet not really understand why it works
20:06:00  Yeah. I don't really like the difference between "how" and "why" in English. I know historically how/why it came about, but that doesn't help.
20:06:32  So, you know how there's software like twitter and twitch that think everything is a hyperlink and aggressively try to turn random parts of your statements into a hyperlink? And there's forum software that thinks that normal people never use hyperlinks into their posts, they just expect the reader to magically understand the context, and anyone who writes a link is probably a spammer. Can we put these 
20:06:38  two kinds into an arena and have them fight to death?
20:06:41  oddly, I was looking at this recently in an entirely different field
20:07:17  I've been trying to write an esolang where numbers are complex rationals, and then trying to work out how to approximate trancendental functions in it
20:07:25 -!- Yayimhere has quit (Ping timeout: 250 seconds).
20:07:32  obviously the result isn't necessarily going to be a rational, but you can approximate
20:07:44  but that means you need, e.g., root-finding algorithms
20:07:56  gradient descent works for that sometimes and doesn't work at other times
20:09:27 -!- lisbeths has joined.
20:10:43  Do you need anything specific? I assume you already know the methods to find roots to polynomials, and how to compute sines and arctangents, and you can find software for just about any specific function that you want to approximate these days.
20:11:58  not really, I already found an algorithm that's sufficiently fast in all cases, I've just procrastinated in linking it up to the rest of the language
20:13:03  (the basic idea, after isolating the root, is to alternate between bisection and regula-falsi; regula-falsi is fast when it works, the bisection ensures logarithmic convergence even in the regula-falsi worst case)
20:14:10  of course very often the numerical software exists in theory, but is in a form that's hard to use in practice
20:14:16  I haven't implemented the root isolation yet, though
20:14:34  also I've been procrastinating on writing the bignum library
20:15:21  though while we're there I might as where ask: is there software that can compute the cumulative density function of an arbitrary multi-dimensional gaussian distribution given its mean and covariance matrix?
20:15:55  I don't even really need this anymore, I'm just curious
20:15:57  unfortunately I don't have the capabilities of the various numerical softwares memorised
20:16:34  oh, that reminds me, I need to find a new package for solving simultaneous equations over finite fields (or learn the algorithms to write one myself)
20:16:40  the one I used before isn't in the repositories any more
20:17:48  polynomial equations?
20:17:53  like multivariate?
20:18:00  multivariate linear
20:18:11  this is basically the same problem as inverting a matrix
20:18:25  oh, linear. have you tried GAP yet and given up in disgust because everything was indexed starting from 1?
20:18:38  there are tons of software packages that do this for normal number systems, but finite fields are a bit harder
20:19:00  I'm pretty sure GAP can do this for any finite field, though I only used it for the simple case of GF(2) so far
20:19:32  no, finite fields are *easier* because you can get an exact answer rather than having to be careful not to lose to much precision
20:19:45  b_jonas: I mean, to find software for
20:19:48  not to solve
20:20:02  oh yes, that's true
20:20:20  wait no, I have used GAP for other fields
20:20:56  I should try GAP, to see if it works
20:21:12  although, I would like to implement this myself so that I can use it in an assembler I'm writing
20:21:25  in theory I should be able to tell you how to implement it
20:21:38  without the person doing the assembling needing the library installed
20:21:47  especially if you already know how to add, subtract, multiply, divide, compare to zero in the field that you need
20:21:49  my problem here is that I understand the naive algorithms but am not sure that they're fast enough
20:22:19  and am not sure whether there's some reasonable way to get a better complexity
20:23:56  hmm
20:24:19  I don't know where I could even look that up
20:26:09  maybe the FXT book has something? it talks about fourier and similar transforms on finite fields to efficiently multiply polynomials in order to efficiently multiply ordinary integers, which it calls number theoretic transforms
20:26:41  oh
20:27:12  and then there's a whole slew of fast matrix multiplication, and you can probably design your finite field matrix solver to use those as much as possible
20:27:48  I wasn't really looking into this because the finite field matrix divisions that I cared about were such small matrices that the efficiency didn't really matter
20:28:12  and I never needed to use them often
20:30:34  [[Stringy]] N https://esolangs.org/w/index.php?oldid=167630 * Esolangist * (+714) Created page with "Stringy is a language by [[User:Esolangist]] that was inspired by [[Thue]] ==Commands-ish== {{WIPsec}} a=b makes a new string replacement: a is now b  a=$OUT outputs a when it appears  a=$IN a will be replaced with user input   [[General Lock Notation]]  https://esolangs.org/w/index.php?diff=167631&oldid=167540 * Zzo38 * (+214) 
21:09:05  [[Dimension]] N https://esolangs.org/w/index.php?oldid=167632 * Esolangist * (+123) Created page with "Dimension is a 4d tape ==Commands needed for a dimension== 
 ^<>v (2d moving)  (3d)  (4d) . (Set) , (Get)"
21:22:47  [[Esolang:Sandbox]]  https://esolangs.org/w/index.php?diff=167633&oldid=165860 * Esolangist * (+140) 
21:24:30  [[Esolang:Sandbox]]  https://esolangs.org/w/index.php?diff=167634&oldid=167633 * Ais523 * (-140) clean sandbox  trying to use it as a way to collaboratively make content is a bad idea because the edits there can be cleaned or overwritten by other users at any time, and trying to make permanent-looking content there discourages people from using it for test edits
21:35:57 -!- pool has quit (Read error: Connection reset by peer).
21:36:17 -!- pool has joined.
21:55:12  [[Q sharp]] N https://esolangs.org/w/index.php?oldid=167635 * Esolangist * (+142) Created page with "{{Wrongtitle|title=Q#}} {{WIP}}  Q# is a programming language by [[User:Esolangist]] ==Commands== print("String") - Do I have to explain this?"
22:20:32 -!- chiselfuse has joined.
22:22:36 -!- chiselfu1e has quit (Remote host closed the connection).
22:56:33  cu
23:09:43 -!- ais523 has quit (Quit: quit).
23:15:03 -!- lisbeths has quit (Quit: Connection closed for inactivity).