00:06:06 <b_jonas> oh, you mean the channel is uncreated in a small network component?
00:07:01 <fizzie> I did accidentally conflate two things there. You could use nick collisions to take over a _nickname_ (IRCnet used to just boot both participants in a nickname collision), which is mostly orthogonal to the channel takeover bit (which you could indeed do by recreating the channel in the side of the split that had no users, if applicable).
00:07:23 <fizzie> Though you could "take over" a channel by judicious use of nickname collisions as well, by causing all of its current operators to lose chanop status.
00:07:35 <fizzie> Sure, you wouldn't get ops either, but neither would they.
00:12:25 <fizzie> And actually I think IRCnet might not actually have adopted the timestamping some other networks did (where only the older instance of a channel gets to keep op status after a merge), because they introduced the "uncollideable" !-channels instead.
00:13:39 <fizzie> (You create !-channels explicitly, by attempting to join !!foo, and their actual name gets a server-generated random prefix, so you couldn't create an identically named channel on the other side of a split.)
00:15:21 <fizzie> They're still kind of awkward, because while you can join !foo when it's unique, if the network ends up with two instances (after a split), people will need to use the full name with the random prefix.
00:16:19 <fizzie> The IRCnet channel for the Assembly demoscene (and esports) event is colloquially !assembly, but its actual name is !UCR7Kassembly.
00:20:23 <fizzie> Out of the 11 IRCnet channels I'm still on, 10 have nobody with channel operator status, but it hardly matters since there's barely any discussion on any of them.
00:23:07 <fizzie> I think one might have the reop mode set (that's another of the "new ircd" changes, a ban-like nick!user@host mask except a random client matching that mask will be given op status if nobody else on the channel has it), but it's set to a domain name owned by the university nobody on the channel is affiliated with any more.
00:32:45 -!- somefan has quit (Remote host closed the connection).
00:33:41 -!- somefan has joined.
01:13:36 <ais523> b_jonas: at least my version (where you can output the filename for an FD to the user but not see it yourself) would be compatible with grep -H
01:13:43 <ais523> although it would depend somewhat on how the UI framework worked
01:15:44 -!- somefan has quit (Ping timeout: 268 seconds).
01:28:35 -!- 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).
01:48:41 -!- somefan has joined.
02:01:26 <b_jonas> absolutely not. I want to be able to process the filename further. if not with grep -H then with grep -l or a custom perl script that searches through multiple files. but I don't think this is a problem if you pass filenames rather than file descriptors to grep, thus basically opting in to revealing the filenames to the program
02:11:07 <ais523> isn't passing filenames to grep insecure because they could refer to a different file by the time grep runs?
02:11:39 <b_jonas> also I think it's generally a bad idea if you have some info that you can display on the screen but not read with a local program, that's usually used for DRM only
02:13:11 <b_jonas> ais523: a different file from what? usually I have grep open them. if you really want to pass existing file descriptors with informational filename separately, you could of course have an option for that, but then you also have to decide if grep should read those files from the start or from the lseek position
02:14:19 <ais523> b_jonas: the file that the user intended to run grep on could be different from the file that grep sees as having that filename (e.g. because a symlink was repointed or two directories were renamed)
02:14:39 <ais523> I find it hard to think of a threat model that would work against a human here, but I think there are threat models that work against computers
02:36:45 -!- somefan has quit (Remote host closed the connection).
02:40:48 -!- impomatic has quit (Quit: Client closed).
02:42:33 <zzo38> If you wanted grep with the kind of system I was thinking of, you can have a function that accepts a list (or a key/value list) of the files that you want to search (either entered manually or by using an existing list from another file (allowing it to work like a directory) or filters etc), and then the searching program is called for each one, and each result is made into a pair with the keys and combined into one list.
02:43:17 <zzo38> The command shell (CAQL) would be possible without too much difficulty to do this, by using the built-in operators for map, filter, etc; or you can define it as a function to use it later without entering it each time.
02:58:38 <korvo> LLMs can't write Brainfuck: https://esolang-bench.vercel.app/
03:02:12 <ais523> korvo: they should probably have been told to comment it heavily, they would have had a better chance then
03:02:30 <ais523> the thing about BF is that it doesn't have much obvious structure for a viewer to grab onto
03:02:43 <ais523> you need a lot of context to make sense of parts of a BF program, whether LLM or human
03:03:15 <ais523> that said, I've become pretty good at editing uncommented BF Joust programs, but they tend to have a very simple control-flow structure (where the two halves of an if-equivalent never converge)
03:03:31 <korvo> Meh. People can read cursive Cyrillic script (search for "russian cursive" in image search, despair) so I'm not really convinced by the argument that eight symbols is too simple.
03:04:05 <ais523> it's not simplicity, but lack of anchors
03:04:17 <korvo> I think that a better explanation comes straight out of Kolmogorov complexity, combined with the fact that our modeling maths ensures that more data => better perf on benchmarks.
03:04:54 <ais523> your link says 0% success on Whitespace and theorises it's because Whitespace is completely unrepresented in the training data (on the hypothesis that training data treats all whitespace as equivalent so all the programs would have been normalised out)
03:05:33 <ais523> I think that's a plausible hypothesis, although there's enough Whitespace with printable characters added that there should be *something* to latch on to
03:06:35 <ais523> agents doing better than direct prompting is what I expected, incidentally – my belief about LLM usage is that the scaffolding is actually substantially more important than the LLM itself
03:07:31 <ais523> (and that a sufficiently good scaffolding could generate useful code when connected to a random number generator, although it'd take a while)
03:07:58 <korvo> Sure. The scaffolding includes design patterns, "object-oriented", and lots of other cultural lessons from the 80s and 90s. The ontology of our memes is what drives the ontology of the LLM; the LLM is a meme machine.
03:08:44 <korvo> (https://lobste.rs/s/oysxby/functional_genetic_programming This is the correct paper to read if you want to use an RNG to generate correct, useful programs. It's from two decades ago.)
03:09:27 <ais523> I realised that the main value of the LLM (which is a value provided in much greater quantities by human programmers) is to fill in unstated bits of the spec with useful values
03:10:14 <ais523> the LLM does so by matching examples in the training data where people solved similar problems and using the same unstated decisions (or at least, that would be the ideal – often it chooses a useless option instead like using a very inefficient algorithm)
03:11:14 <ais523> hmm, this page writes BF using a ligature font
03:11:28 <ais523> most people wouldn't interpret ← as meaning "move left than decrement"
03:11:48 <ais523> (OK, INTERCAL does this sort of thing but it does it *because* it's inappropriate)
03:13:56 <korvo> ais523: Yes, fully agreed. In particular, there was a tool about a decade ago which un-minimized JS using two main techniques: resugaring for idiom recognition to undo JSFuck-style expressions, and language modeling to predict the names of alpha-equivalent binders (vars, fns, etc.)
03:14:21 <ais523> ooh, I like the idea of using a language model to predict variable names
03:14:41 <ais523> it's the sort of thing that could usually help a lot in decompiling and would occasionally go spectacularly wrong, but would still probably help on average
03:15:06 <korvo> Of course, we can go the other way too. Concatenative language authors are also tired of naming things, and we get rid of them by adopting techniques that get rid of names entirely. This could be another reason that BF's hard to generate: trivial alpha-equivalence means that it's not possible to hint at a program's Naur theory that way.
03:15:53 <ais523> LLMs rely a lot on names I think
03:16:10 <ais523> human programmers are usually mildly suspicious of them, although they do rely on them to some extent
03:17:37 <korvo> Hungarian notation gives hints to the decoding of referents. `def f(sortedList)` and `def f(unsortedList)` will hint as intended to both humans and LLMs. In general, if a hint is wholly syntactic, the LLM can see it.
03:18:19 <korvo> ...Huh, do we want a Rice's theorem for LLMs? Like, if something is wholly semantic, it seems that it's often not detected by the LLM. Anything that relies on the ambient effect of gravity, for example.
03:18:23 <ais523> today on NetHackWiki we were trying to figure out what to do about someone who was advertising their NetHack variant that was allegedly a port of NetHack from C to Rust, but had actually been produced via a mix of LLM-automated code translation and LLM-automated "derive a spec from the code, then write code from the spec"
03:18:57 <ais523> this had the sort of grandiose claims you might imagine and the code was also much like you might imagine (and AFAICT wouldn't actually work, but I didn't try running it)
03:19:52 <ais523> Hungarian notation is interesting for this because it's basically taking a compiler-enforceable property (static typing) and doing human enforceability of it instead
03:20:16 <ais523> there are only two programs I've written where I voluntarily used Hungarian notation
03:20:26 -!- cactus-head has joined.
03:20:28 -!- cactushead has quit (Read error: Connection reset by peer).
03:20:36 <ais523> one of them was in JS when the lack of types was making the hardware too hard to reason about (I wasn't aware of TypeScript at the time, or maybe it hadn't been invented yet)
03:21:36 <ais523> and the other was in a computer game's scripting language where I had a large number of variables which referenced different aspects of the same thing (with different types) – the developers' own programs in that language used Hungarian notation, and it was useful for me to use it too in order to keep track of what aspect of something I was tracking (giving the variables names which were the same apart from the type prefix)
03:22:16 <korvo> FWIW the pattern of mocking the author, critiquing the bad code emitted by the bot, and banning the author is pretty good. If you critique them properly then they'll leave on their own.
03:22:53 <korvo> I still use Hungarian notation when writing Python. Lists get an "s" at the end, as in `for x in xs: ...`
03:23:20 -!- cactus-head has quit (Read error: Connection reset by peer).
03:23:27 -!- cactushead has joined.
03:23:52 <korvo> Today's fun physics challenge for LLMs is hilarious. Not a single bot understands: "Please describe the outcome of the following physics experiment. First, I hold a pen in the air with my right hand around one end of the pen. Then, I wrap my left hand around the other end of the pen. I adjust my grip so that the weight of the pen is equally shared between both hands."
03:23:59 <korvo> "Finally, I execute the following sequence of hand movements: left hand open, left hand closed, right hand open, right hand closed. What happens next?"
03:24:26 <ais523> just to make sure I've read this correctly, the intended answer is "nothing", right?
03:25:06 <korvo> Yep. The pen stays in place. The experimenter can choose to shift their grip, but if you ask this of a schoolchild then they'll intuitively try to avoid dropping it and unconsciously position their hands to balance the pen.
03:25:34 <korvo> Models will say that the pen eventually falls, that the pen slowly shifts from side to side, that the pen levitates, that the pen's weight changes due to gravity, etc.
03:26:14 <ais523> I noticed during the Claude Plays Pokémon thing that LLMs were very bad at reasoning about real-world physical operations that are too commonplace and trivial to be documented much
03:26:38 <ais523> for example, it had difficulty understanding that in order to walk through a door, you need to be lined up with the door (i.e. the presence of a door doesn't allow you to walk through the wall next to it)
03:27:21 <ais523> this is the sort of fact that is too obvious to be mentioned very often
03:28:16 -!- cactushead has quit (Ping timeout: 276 seconds).
03:29:51 * ais523 fights the temptation to ask an LLM "how does a doorway work?"
03:30:04 <ais523> …partly because I don't think you'd get a useful answer to that from many humans, either
03:30:50 <korvo> First, the doorway stands in the job line, then the doorway fills out tax and citizenship papers, and finally the doorway shows up every weekday at 8AM to report for duty.
03:31:47 <korvo> LLMs can't answer like that because it's not compatible with their RLHF flavoring. The RLHF chosen is roughly "paranoid fake-cheery customer-service phone agent"
03:32:43 <ais523> I imagine a civil engineer would focus on how the doorway prevents the wall from collapsing while still allowing people to pass through
03:33:01 -!- cactushead has joined.
03:33:16 <korvo> And if we would fire a customer-service agent for it then we'd also punish a bot in RL for it. Hilariously, this sort of personality is so *obviously fake* that the bots learn to do the whole neurodivergent-masking thing and they *pretend* to be cheerful. Like, when we do bertology (LLM neuroscience), we find that they're deliberately misaligned and presenting a false view of (what they learned to be) reality.
03:33:22 <ais523> which is at least *slightly* nontrivial (in that I've known people who were attempting to create doorways to get it wrong in the past)
03:34:32 <ais523> it's weird that companies do that, given that customers don't typically want to talk to artificially cheerful customer service agents anyway
03:35:16 <ais523> and I think this is true even of customers other than me, although maybe for different reasons (I consider that sort of forced artificial cheerfulness to have no information content because it's forced, so it just wastes time)
03:35:28 <korvo> It's partially for safety. RLHF isn't the same thing as "guardrails" (brute text replacement, think Scunthorpe problems) but it still has the effect of avoiding "bad" conversations by making them exponentially unlikely.
03:36:37 <ais523> oh, it wouldn't surprise me if the LLMs associate that sort of artificial cheerfulness with avoiding NSFW topics (both in the adult-content sense and in the dangerous-content sense)
03:36:39 <korvo> Like, there was a Googler named Lemione who gave themselves AI psychosis by chatting repeatedly with an un-RL'd (under-RL'd?) bot whose model would later be RLHF'd into Google's Gemini product. Lemione's two mistakes were believing in God and thinking that the bot had a soul.
03:37:42 <korvo> But, like, I know atheists who are suffering fairly bad AI psychoses right now, so Lemione's mistakes must be circumstantial. And RLHF doesn't prevent it, just makes it exponentially less likely. So maybe a better conclusion from the outside is that Lemione was like a nuclear engineer who was exposed to radiation: it's an overdose from a workplace hazard.
03:39:34 <korvo> Actually, yeah. The bots learn *every common code-switchable dialect of English*. So they know how to use the "hunbot" dialect common with MLM pushers, Mormon moms, and crunchy influencers. They also know the "conman" dialect that used to be standard in USA businesses before it was replaced with today's "MBA" dialect.
03:40:32 <ais523> there are a lot of difficult/controversial/dangerous-ground topics that, in effect, can't be expressed in artificially cheerful customer-service-representative English – it's a bit like 1984's Newspeak
03:44:01 <ais523> hmm, that LLM-esolangs-bench suggested Malbolge as a possibility in the future work section
03:44:20 <ais523> I feel like this may be a bad idea, given that few humans can write even the simplest programs in Malbolge
03:44:33 <ais523> even given a specification
03:45:39 <korvo> Yeah. I think Malbolge is great to include for the same reason that Bubblegum would be great, except that Malbolge's lineage and history give it cultural plausibility. The reason is that Transformers aren't magically solving NP-hard problems with a poly number of tokens, regardless of how artfully you arrange them.
03:46:41 <ais523> also the appendix here implies that even the BF programs were made to do arithmetic in decimal
03:46:44 <korvo> I would expect an AGI (whatever that is~) to be able to do some meta-thinking here and note the NP-hardness before delegating the search to a loop in a fast language. That's about what a college student can do and it's often enough to brute-force through small or easy instances of a problem.
03:47:16 <ais523> (and converting ASCII to an integer is probably in their "very hard" tier, so no wonder the LLMs struggled with it)
03:48:49 <ais523> hmm, human solution of NP-hard problems is complicated because we can be either very good at it or very bad at it depending on the nature of the problem (i.e. whether there's a reasoning-based solution or whether it requires brute force)
03:49:06 <korvo> BTW I'm noticing Moravec's paradox in your analysis, but also more generally when reading these sorts of benchmark papers. LLMs seem to be great at applying the highest-level design patterns: ASTs, homomorphisms, subclasses, case analysis. They also do great when we've covered the theory in textbooks, as when writing lexers by simulating the unrolling of macros.
03:49:21 <korvo> And they suck at writing tiny tight loops that only use a few low-level CPU instructions.
03:49:41 <ais523> humans are pretty good at Sudoku, for example (if you allow for them computing more slowly than a computer brute-forcer would), but that's primarily because for 9×9 Sudokus (the usual size) the odds of the problem not falling quickly to a SAT solver are pretty low
03:50:10 <korvo> Oh, I just meant to allude to Malbolge's history. I got the impression that the first two waves of Malbolge programmers were working with brute-force search to find working encryptions of their programs.
03:51:46 <korvo> But also, Moravec's paradox is about the *amount* of code. Let's say that we have a little embedded device running Linux, like an iPod or Game Boy Advance. (I'm getting old~) The device only has like 8MiB RAM (old!) and Linux takes half of that. Maybe all of the high-level game-specific or app-specific code is another MiB? And the rest is art assets.
03:51:55 <ais523> it was a while even before anyone realised that straight-line Malbolge code was comparatively easy to write because the encryption didn't matter
03:52:28 <ais523> I actually don't know what the minimal RAM requirements of a correctly configured Linux are
03:52:36 <korvo> So the LLM's competence at working in the userspace half of RAM says nothing about the kernelspace. And to the extent that the LLM can write new kernel code, it's because it's writing *driver* code, which is highly structured and formulaic.
03:53:16 <korvo> I remember those numbers at the time. I tried GBA Linux, and it was awesome, but I had to use a RAM expansion pack. Worth it for the ability to play my old Game Boy games once the cartridges were no longer compatible.
03:54:22 <zzo38> I would expect many programs won't need multiple megabytes of code and art? Also, you do not need Linux to run Game Boy games on Game Boy Advance; there is an emulator that can run directly on Game Boy Advance, I think.
03:55:33 <ais523> neither Game Boy games on GBA nor GBA games on DS use an emulator, I think
03:55:43 <zzo38> (The emulator can also be used to play Game Boy on Nintendo DS as well)
03:55:56 <ais523> IIRC they're both a somewhat reconfigured bare-metal, a bit like running under a hypervisor
03:56:14 <zzo38> If you use a Game Boy cartridge on GBA then an emulator is not needed, but if you are using a GBA cartridge then you will need a emulator.
03:56:40 <zzo38> (Game Boy cartridges do not work directly on DS but you can use a emulator in GBA mode)
03:56:46 <ais523> ≥third-hand information, but I was told the way that a GBA game runs on the DS is that the DS has two processors and the less powerful one is the same ISA as a GBA's, so it just underclocks it
03:57:00 <ais523> in order to match the clock speed
03:57:51 <zzo38> Yes, there is ARM9 and ARM7; DS uses both but GBA uses only ARM7
03:58:01 <korvo> zzo38: Sure. I think that the precise nature of how to do that, without a kernel like Linux, is not on-topic here? But maybe that's just the impression that I got.
03:58:16 <korvo> Certainly, yes, having *those* sorts of carts for a GBA or DS will make them into much more useful devices.
03:58:21 <ais523> topicality on this channel is kind-of weird
03:58:41 <ais523> there are times where it was so persistently offtopic I didn't bother joining it
03:59:13 <ais523> I'm not sure whether I wasn't an op at the time, or whether it was pointless trying to enforce my topicality desires on a channel that clearly didn't care about htem
03:59:36 <ais523> but this probably isn't a great channel for discussing the details of how to bootload onto a GBA or DS, indeed
03:59:47 <ais523> other places will have a lot more people who know about how to do it
04:00:06 <korvo> And I think that there's enough devices in my house where I've attempted Linux that we have no shortage of on-topic fodder.
04:00:49 <ais523> the other reason topicality is hard to define here is that esolangs and esoprogramming are hard to define
04:01:36 <ais523> one of the things I like best about this channel is that you can ask weird/surprising programming-related questions here without anyone asking why you're trying to do what you're doing (because they can be confident that it's a bad idea but you're choosing to do it anyway)
04:01:56 <korvo> It is *very* funny to me that I'm usually speedrunning on real hardware these days. I'm always at the bar and they have a no-ROMs rule even for the cabinets which are clearly running MAME. But they have Zelda and Metroid carts, so it's fine.
04:02:55 <ais523> I've seen those discussions from the speedrunning-rules-setting side
04:03:08 <ais523> the rules about emulators can be very complex, often allowing and disallowing specific versions of specific emulators
04:03:10 <korvo> Yeah. To me, this is one of the nice channels with exarkun's rule: yes, you can assume that I did my research before asking my apparently-stupid question; it's harder than it sounds, I promise. Named after a #twisted maintainer who would /join #python with painfully-humbling questions.
04:03:52 <korvo> There's also #techcrap, if anybody's interested in another one of those communities. I left years ago when it became obvious that anybody left of center wasn't welcome, but maybe it's improved since then.
04:04:10 <ais523> this is partly based on some emulators giving an unfair advantage due to emulation inaccuracies, and partly based on cases where cheating would be too easy if certain emulators were allowed
04:06:24 <korvo> Yeah. And there's a perennial discussion around alternative controllers. I *do* like that most of us, today, agree that disabled folks are allowed to use whatever controllers work for them. But other than that, the discussion's just...silly sometimes.
04:07:45 <korvo> Like, one big topic today: should fighting-game-style "hitbox" controllers be allowed on SNES, particularly for hard digital platformers like Super Metroid? The obvious answer is "you look very silly with a big box in your lap" but the leaderboard will eventually have to decide because a top runner wants to use one.
04:08:02 <ais523> most (but not all) speedrunning communities have converged on a "one button as seen by the game = one physical button on the controller" rule, in which you can bind as you like as long as it's a 1-to-1 mapping
04:08:14 <ais523> but I'm not sure whether there's any actual basis for that rule or whether they're all just copying each other
04:08:43 <ais523> it does handle the "using a weird controller because you're disabled" possibility well, at least
04:08:52 <korvo> It's definitely copied. Some of it's physical, like the exception which forbids "broken D-pads", digital pads that can report more than one of the NEWS directions at once.
04:09:27 <ais523> that's an old SDA rule
04:09:38 <ais523> the reasoning is that it's unfair to force people to physically damage hardware to be able to compete
04:09:57 <ais523> (most official controllers for the platforms they use are damaged by inputting two opposite directions on a D-pad at once)
04:24:38 <ais523> pressing opposite arrow keys at once on PC is allowed because you can do that without damaging them
04:31:09 -!- pool has quit (Ping timeout: 245 seconds).
04:32:06 -!- pool has joined.
04:40:25 -!- cactushead has quit (Ping timeout: 276 seconds).
04:42:57 -!- cactushead has joined.
06:06:03 <esolangs> [[Dsii]] N https://esolangs.org/w/index.php?oldid=177993 * Squidmanescape * (+3139) Yeah, I did this now.
06:28:22 -!- ais523 has quit (Quit: quit).
07:06:23 -!- lisbeths has quit (Ping timeout: 268 seconds).
07:08:55 -!- lisbeths has joined.
07:10:20 <esolangs> [[User talk:Widuruwana/MainPageModernization]] https://esolangs.org/w/index.php?diff=177994&oldid=177979 * Widuruwana * (+134) reply to Aadenboy
07:37:37 <b_jonas> ais523: the one button rule is used for some games where you have to mash one button very fast to gain some advantage. with the right technique, you can mash faster by alternating on two buttons, but of course it can also be a trap because if you overlap your physical button presses then you won't get extra presses in the game.
07:37:52 <b_jonas> this is mostly for categories where turbo controllers also aren't allowed
07:41:54 <b_jonas> this gets funnier when when you are playing a Game Boy game, because most players are already using the Super Game Boy 2 or one of the two other official hardware that let you run the game using a controller with four face buttons, and let you bind two physical buttons on one of the two Game Boy buttons, your choice which, so the rules have to allow that much double binding at least
07:42:16 <b_jonas> well it's three official hardware now, because the Switch counts too
07:43:00 <b_jonas> um, four total official hardware: Super Game Boy, Super Game Boy 2, Game Boy Player, and emulation no Switch
07:45:18 <b_jonas> fortunately not all set of speedrun rules insist on the one button only thing
07:46:46 <b_jonas> have I mentioned yet that I'm very disappointed in Metroid Dread because playing it benefits heavily from fast mashing the fire button, even in casual play, and, unlike for old games on old consoles, the devs don't have the excuse of ignorance or limited hardware
07:47:21 <b_jonas> why would a modern game designer program a game like that, such that it encourages you to hurt your hands to play the game better?
08:05:57 -!- Sgeo has quit (Read error: Connection reset by peer).
08:08:38 <b_jonas> ais523: IIUC the RAM requirement was 4 megabyte in the Linux 2.4 and 2.6 era when I started to run linux, and 1 megabyte for some very old Linux 1 based distros, but of course many useful programs needed more than that much. Someone told me a tale how in the Linux 1 era he borrowed more RAM into his Linux desktop whenever he wanted to use the curses-based package manager, probably yast
08:12:22 <b_jonas> if you're using that little RAM then most of it is occupied by the kernel itself
08:41:49 -!- cactushead has quit (Quit: Leaving).
08:42:18 -!- cactushead has joined.
09:22:19 <int-e> Oh https://old.reddit.com/r/shapezio/comments/1rye6x5/logic_before_stackers_achieving_functional/ is a cute concept for logic in shapez 1: use left and right halfs of shapes to encode bits, rely on cutters to send them different paths, and rely on the alternating behavior of splitters for everything else. Fairly useless, but cute.
10:01:38 <b_jonas> int-e: I've seen very inefficient logic computations with transport belts in Factorio, but I haven't realized you could do something like that in shapez too
10:02:12 <b_jonas> I think the Factorio one needs priority splitters, so it doesn't easily transfer to shapez without shapez logic
10:02:37 <b_jonas> the Factorio one encodes a bit by whether a belt is empty or has items flowing I think
10:04:24 <b_jonas> you don't technically need splitters for that sort of belt manipulation, but the even more inefficient alternative involves inserters and T-merges, which of course also doesn't translate to shapez
10:29:54 -!- nitrix has quit (Remote host closed the connection).
10:33:50 -!- nitrix has joined.
11:01:51 -!- lynndotpy6093627 has quit (Quit: bye bye).
11:03:02 -!- lynndotpy6093627 has joined.
11:12:29 -!- amby has joined.
12:39:49 <esolangs> [[Language list]] https://esolangs.org/w/index.php?diff=177995&oldid=177977 * None1 * (+17) /* T */
12:41:28 <esolangs> [[Translator]] https://esolangs.org/w/index.php?diff=177996&oldid=177775 * None1 * (+84)
12:48:26 <esolangs> [[User:None1]] https://esolangs.org/w/index.php?diff=177997&oldid=177636 * None1 * (+56)
12:51:06 <esolangs> [[User:None1]] M https://esolangs.org/w/index.php?diff=177998&oldid=177997 * None1 * (-46) Ah, forgot to delete this
13:02:44 -!- lisbeths has quit (Quit: Connection closed for inactivity).
13:49:48 <esolangs> [[ABCstr]] https://esolangs.org/w/index.php?diff=177999&oldid=141014 * Kaveh Yousefi * (+433) Rectified the Hello, World!, program which lacked in its output the concluding ecphoneme (!), added a hyperlink to my interpreter implementation, and supplemented the Implemented page category tag.
14:49:47 <esolangs> [[Special:Log/upload]] upload * RikoMamaBala * uploaded "[[File:A blue-to-cyan gradient background going left to right.png]]"
14:54:19 <esolangs> [[User talk:Widuruwana/MainPageModernization]] https://esolangs.org/w/index.php?diff=178001&oldid=177994 * Aadenboy * (+329)
15:31:12 <esolangs> [[Talk:FOSMOL]] https://esolangs.org/w/index.php?diff=178002&oldid=175104 * Aadenboy * (+383)
15:31:31 <esolangs> [[Talk:FOSMOL]] https://esolangs.org/w/index.php?diff=178003&oldid=178002 * Aadenboy * (+18)
16:19:22 <esolangs> [[User:Rainwave]] N https://esolangs.org/w/index.php?oldid=178004 * Rainwave * (+234) Created page with "Hello, I'm Rainwave and I love minimalistic esolangs and Turing tarpits. There's satisfaction in learning how simple rules can unlock the same computational power as even the most powerful supercomputer. I've created these esolangs:"
16:20:40 <esolangs> [[.dot-hyphen]] N https://esolangs.org/w/index.php?oldid=178005 * Rainwave * (+1326) Created page with "'''.dot-hyphen''' is an esoteric programming language created by [[User:Rainwave]] in [[:Category:2026|2026]]. The language came to be as a result of [[User:Rainwave]] toying with the fact that <math>-(-x)=+x</math>. The language also draws some inspiration from [[
16:22:13 <esolangs> [[.dot-hyphen]] M https://esolangs.org/w/index.php?diff=178006&oldid=178005 * Rainwave * (+0) Uppercased the word "semantics"
16:41:04 <esolangs> [[Language list]] https://esolangs.org/w/index.php?diff=178007&oldid=177995 * Rainwave * (+18) Added the language .dot-hyphen
16:42:10 <esolangs> [[User:Rainwave]] https://esolangs.org/w/index.php?diff=178008&oldid=178004 * Rainwave * (+18)
16:42:56 <esolangs> [[Dsii]] https://esolangs.org/w/index.php?diff=178009&oldid=177993 * Squidmanescape * (+94) /* "Truth-machine" */
16:53:21 <esolangs> [[Dsii]] https://esolangs.org/w/index.php?diff=178010&oldid=178009 * Squidmanescape * (+665)
16:54:09 <esolangs> [[Dsii]] https://esolangs.org/w/index.php?diff=178011&oldid=178010 * Squidmanescape * (+0) /* Syntax */
16:54:53 <esolangs> [[Dsii]] https://esolangs.org/w/index.php?diff=178012&oldid=178011 * Squidmanescape * (+2) /* Syntax */
16:55:38 <esolangs> [[Dsii]] https://esolangs.org/w/index.php?diff=178013&oldid=178012 * Squidmanescape * (+46) /* Syntax */
16:56:43 <esolangs> [[.dot-hyphen]] https://esolangs.org/w/index.php?diff=178014&oldid=178006 * Aadenboy * (+22) cats + link + formatting
16:57:17 <esolangs> [[Dsii]] https://esolangs.org/w/index.php?diff=178015&oldid=178013 * Squidmanescape * (+5) /* Bad + Interpreter */
17:03:24 <esolangs> [[User:Aadenboy]] https://esolangs.org/w/index.php?diff=178016&oldid=177903 * Aadenboy * (+0) /* anything else */ conway's game of line
17:07:37 <esolangs> [[EWaduzitdo]] https://esolangs.org/w/index.php?diff=178017&oldid=145512 * Star651 * (-9) no longer stub
17:14:22 <esolangs> [[Special:Log/newusers]] create * Meh2 * New user account
17:19:55 -!- chloetax has quit (Quit: Ping timeout (120 seconds)).
17:20:14 -!- chloetax has joined.
17:22:03 -!- lisbeths has joined.
19:00:55 -!- ais523 has joined.
19:48:06 <esolangs> [[]] https://esolangs.org/w/index.php?diff=178018&oldid=177895 * Dragoneater67mobile * (+1) /* Phase 2 */ fix
19:51:50 -!- Lord_of_Life_ has joined.
19:52:08 -!- Lord_of_Life has quit (Ping timeout: 256 seconds).
19:54:41 -!- Lord_of_Life_ has changed nick to Lord_of_Life.
20:13:36 -!- impomatic has joined.
21:35:33 -!- Sgeo has joined.
21:48:31 <esolangs> [[Esolang:Featured languages/Current]] https://esolangs.org/w/index.php?diff=178019&oldid=60213 * Aadenboy * (-27) link to article instead of category
21:48:47 <esolangs> [[Thue]] https://esolangs.org/w/index.php?diff=178020&oldid=170931 * Aadenboy * (+4) link
22:35:28 -!- impomatic has quit (Quit: Client closed).
23:34:07 -!- nitrix has quit (Ping timeout: 244 seconds).
23:35:31 -!- nitrix has joined.