kacey's recent activity

  1. Comment on Zig creator weighs in on the Bun Rust rewrite in ~comp

    kacey
    Link Parent
    Ach, good find. It looks like his original letter is from ~2023, so this isn't a recent thing either. Sigh.

    Ach, good find. It looks like his original letter is from ~2023, so this isn't a recent thing either. Sigh.

  2. Comment on Zig creator weighs in on the Bun Rust rewrite in ~comp

    kacey
    Link Parent
    Aaaaugh HN. I hoped never to see that orange banner again. I think you're linking to this comment? You can get a direct link by clicking the "X days ago" text. Agreed re. JS. People've been...

    Aaaaugh HN. I hoped never to see that orange banner again. I think you're linking to this comment? You can get a direct link by clicking the "X days ago" text.

    No i'm definitely summarizing and essentially agree, but I've met the kind of people who thought everything should be made with JS (before TS even existed), and I think seeing a language in the wild/reality is important. Things that don't matter for one scope matter a LOT for another, and given "will i need to manually handle memory" is one of the biggest decisions at the product creation, Zig's performance there on a major product strikes me as very very important.

    Agreed re. JS. People've been zealots for their favourite tools forever. But it's disappointing to me that a debate akin to Milwaukee vs. Dewalt seems to be elevated to the level of pseudoscience (pseudoengineering?) in software dev circles, whereas -- afaict -- tradesfolk are generally not actually frothing at the mouth crazy when they defend their preferred power tool colour. Dunno, maybe people with more experience can chime in; I only have Reddit comments to rely on.

    It cannot, and neither can rust fully catch them.

    Yep. Not sure where the framing of my claim that Rust fully solves "them" came from. Your full statement was:

    When everyone swears their product is the best (again nothing really new, just inflated thanks to the money) getting valid data points is important, and if you just hear "oh wow Zig just can't handle catching some memory errors while rust can" that's a BIG statement given that is probably one of THE largest problems in coding and any tool in this day and age not solving it leads you back to "might as well use C, at least I can hire for it" or similar problems.

    My response was that it is provably the case that Zig can't handle some memory errors while Rust can. I was not making a universal statement, so I'm not sure why I'm being framed as having done so.

    "solving this with GC" is.....kinda wild take to me. [...] However I have friends who absolutely cannot use a GC language for their line of work due to the performance requirements they need to meet.

    Ah, that's on me; I added a statement in out-of-sequence in a previous paragraph to couch my claim:

    [...] delta some realtime-critical applications.

    High frequency trading, aerospace, or nuclear control systems, all of which have liveness and realtime constraints which a GC generally cannot meet. Please note, though, that the statistically average professional software developer is writing with a language that would be a PIA to extract the GC from (Stack Overflow 2024 Survey; to my eye, maybe 15% -- optimistically -- used Rust/C++/C that year) so evidence strongly leans towards the observation that GCs must have solved a huge section of these problems for the overwhelming majority of situations. Either that, or 85% of programs don't have this problem to begin with, so we're back to saying it's only a minority concern.

    Please understand that I am not stating that every memory-related software defect can be addressed with a GC. I am stating that a GC fixes the problems that a GC fixes, that Rust's borrow checker fixes what Rust's borrow checker fixes, and Zig's arena allocators (iirc that's how it handles memory allocation) fix what Zig's arena allocators fix. These all correspond to addressing a not-necessarily-overlapping set of defects in applications. The claim I made was that there are a large set of memory use-related errors that Rust prohibits at compile time (when using the borrow checker properly) which Zig will not (without excessive hoop jumping).

    Re. Bun and the necessity of garbage collection, all I can point to are Maven, Gradle, Bazel, and all the existing Javascript tooling to say that they didn't need to abandon the GC to hit whatever their performance goals are unless they're extremely memory constrained for some reason. Bazel especially rips through enormous builds by dint of running a persistent build daemon. To be precise about my point: "performance" is squirrelly and ill-defined. We have evidence that, for some definition of "performant", we can build performant build systems that use GCs.

    A random blog writeup that is inherently biased (in both cases) strikes me as a less than useful datapoint. Git activity and actual code gives plenty of real examples to dive into to at least asses "oh god Zig will be a minefield of gotcha's" or "jesus no wonder Bun is whining, this code is crap".

    I think we're in agreement? I'm once again being bitten by my terrible communication skills. I was trying to imply that a third party's coherent, low bias analysis of the problem is what's necessary for coming to any reasonable conclusion. When each side slings isolated fragments of evidence at one another, we lose the context, distribution, and frequency information which is critical for understanding if they're e.g. cherry picking in order to "win" the debate.

    It's just more front and center now.

    Pretty much agreed.

    3 votes

  3. Comment on Zig creator weighs in on the Bun Rust rewrite in ~comp

    kacey
    Link Parent
    Fair enough! As noted in my previous comment, I suppose I prefer a world with more professional courtesy, but I can understand that the lines between that and "cover-up" can become blurred -- and...

    If they’re wrong to have shared this then - again - they’re sharing even more with the world than what people normally do. And I appreciate that. Now we have dirt on both parties.

    Fair enough! As noted in my previous comment, I suppose I prefer a world with more professional courtesy, but I can understand that the lines between that and "cover-up" can become blurred -- and given the choice, I'd lean towards preferring openness. I still hope for a world where we can trust one another enough not to speak ill of us behind closed doors, without the threat of NDAs/non-disparagement contracts to keep everyone in line.

    4 votes

  4. Comment on Zig creator weighs in on the Bun Rust rewrite in ~comp

    kacey
    Link Parent
    Ah, apologies, I should've been more precise. The context of this discussion is comparing Linus' jackassery to Andrew's, with respect to defending the honour of their respective OSS projects. I'm...

    Ah, apologies, I should've been more precise. The context of this discussion is comparing Linus' jackassery to Andrew's, with respect to defending the honour of their respective OSS projects. I'm claiming that a broad drop in confidence in Zig's ability to perform as a programming language is orders of magnitude less important to the software development community as a whole than Linux's. I don't feel that either maintainer's response is appropriate, but since there are actual stakes at play if the industry as a whole decides to move on from Linux because someone powerful dictates that "it can't scale" or something, it doesn't seem like an apt comparison.

    Taken to its logical extremes, if we accept that toddlers can cry because they can't have juice after bedtime, does that justify Andrew throwing a tantrum because a donor pulled out?

    4 votes

  5. Comment on Zig creator weighs in on the Bun Rust rewrite in ~comp

    kacey
    Link Parent
    IMO, understanding where tools have fallen short doesn't make them good or bad on a flat, 1d spectrum, it just indicates where the risks could be in adoption. Even if it can't "scale" (for however...

    On the other side though this is a big issue. If multiple people come out and just say "yeah zig doesn't scale" that's a problem, and a major project is a very loud datapoint on that.

    IMO, understanding where tools have fallen short doesn't make them good or bad on a flat, 1d spectrum, it just indicates where the risks could be in adoption. Even if it can't "scale" (for however the term is being used in this case), that still doesn't remove its other proven benefits. Bun's post on this is extremely verbose, but IMO since every tool has a set of tradeoffs associated with it, it's still very informative: not everyone is the literal creator of Zig and capable of writing pristine code.

    If we were actually doing software engineering, the takeaway from viewing a failure would not be "oh no this whole thing is bad! We'll never use any of these tools!", it would be understanding the root causes of those issues in order to better develop plans to avoid them, since all tools have flaws and nothing is perfect. Analogously: just because a house catches fire once doesn't mean we should stop building houses and return to the caves -- which are clearly the superior form of housing, since they don't catch fire!! -- because everything is a tradeoff. People like having walls, and windows, and etc. So the takeaway should be to install fire sprinklers, or AFCIs, not to discard an entire toolchain because someone rich and super smart wrote some blogspam about how they've abandoned the surface world. The weird, uncritical, cargo cultness of software development really irks me.

    But that's a subtle point. I think people love the language benchmark game or the TIOBE index since they provide a clear and unambiguous ranking of an otherwise complicated problem, which -- ultimately -- generally has little to no impact on your team's ability to execute on a given project. Given the number of projects barely limping on with extraordinarily subpar stacks in the world (lord knows I've seen enough of them), we have existence proofs that one can accomplish basically everything with anything these days, delta some realtime-critical applications.

    oh wow Zig just can't handle catching some memory errors while rust can

    Quick sidebar: it can't though, can it? Zig doesn't appear to have a borrow checker, and from a skim, it doesn't have a GC either (or ARC). Assuming you were thinking of memory leaks when you mentioned errors? (and not, like, use-after-free, array bounds checking, or etc.) Also also we kinda solved this with garbage collection a couple decades ago; for whatever reason, marketing teams have successfully unsolved it for many folks, but I've very rarely seen GC behaviour cause an issue under load where there wasn't some underlying problem (e.g. memory leaks in unmanaged pools, or some algorithmic error which would've thrashed a manual allocator anyhow).

    Both sides could be providing code snippets they thought were unreasonable.

    I fundamentally don't understand what quoting snippets at each other would do in this case, though. Like, is the goal to learn what went wrong, so we can do better as an industry? It really feels like petty squabbling.

    Linus is not known for his private beatdowns, it just wasn't as visible because you had to be a linux kernal dev or interested in that to even see it. It had similar ramifications though in that the outcomes of those shit slinging sessions often would determine major standards and tech adoption.

    Just addressing this separately, but imo Linux isn't a great example: languages are a dime a dozen (Zig is an uncomfortable middle child between golang, Rust, and C++), and they can be swapped out pretty easily and without major consequence. If one project catches fire and explodes, sure, fine, do a rewrite and move on.

    If the Linux project went down -- especially for a critical decade or two in there -- I'm not sure another project could've taken up the reigns? Sure we have some weird nix's in there (GUIX, Solaris, ... Darwin (_shudders)) ...), but they weren't anywhere near as mature or multipurpose. Linux can work on anything from your microwave to a car to a supercomputer, which would've been extremely challenging to reproduce in another OS.

    6 votes

  6. Comment on Zig creator weighs in on the Bun Rust rewrite in ~comp

    kacey
    Link Parent
    Strong disagree. They're acting like consultants in this scenario: a fee is paid to their organization so that they'll listen to your concerns, and then provide advice on how to fix them. We truly...

    I think it's entirely reasonable to share this with the world.

    Strong disagree. They're acting like consultants in this scenario: a fee is paid to their organization so that they'll listen to your concerns, and then provide advice on how to fix them. We truly should not create the norm that consultants leak your dirty laundry: the entire point is that you can be fully open with them, but if you need to hold information back for fear of their badmouthing you later, then the entire relationship is soured.

    To be clear: I'm not talking about actually important things, like exposing SA or whatnot. This is just business. When two groups of people speak with the expectation of privacy, violating that trust is a bridge you can't uncross.

    For example, I've met a founder in SF who I'm pretty sure is a sociopath. I heard him say some wild stuff about his employees.

    Mmhm, and same. But that's very different than trashing someone's reputation given your insider knowledge because they've broken ties.

    I dunno; I get that people want others to be more open in general, but this feels like if your doctor started a blog describing how much of a piece of crap their patients in particular are, and called them out by name. It seems terribly immature imo.

    6 votes

  7. Comment on Zig creator weighs in on the Bun Rust rewrite in ~comp

    kacey
    (edited )
    Link Parent
    I ... guess? I suppose I'm used to e.g. Bjarne Stroustrup not coming out swinging whenever someone bails on C++ for, like, D or something. Or I guess Oracle dropping a diss track for every press...

    I ... guess? I suppose I'm used to e.g. Bjarne Stroustrup not coming out swinging whenever someone bails on C++ for, like, D or something. Or I guess Oracle dropping a diss track for every press release praising Kotlin. And there doesn't seem to be animosity between the Kotlin and Dart crews.

    If someone stops using your project then great, more power to them. If they misunderstood the assignment, and had been misusing your tool the whole time, then I guess that's on them. The need to "well ackshually" every critique leveled against your project via competing open letters is a trend that I've grown to despise in the modern software development community.

    I'm probably just being nostalgic for the bygone days of yore, where the people engaging in the language wars were just us sweaty code monkeys, swinging away in the pits. It's very odd to see this sort of poop slinging being done by the ostensible responsible adults in the room.

    I am curious what all the fallout from this will be. It strikes me that both sides have objectively provable claims(either zig doesn’t scale well in buns case or bun had slop code) so it shouldn’t be that hard to point to “like this line here” but so far there’s a lack of that

    I don't think those are objectively provable? I can jump into nearly any codebase and find regrettable choices, especially once they crack the mid-five digits. Similarly, you can find C++ projects pushing several hundred thousand lines which compile faster than mid-sized Rust projects, because the people architecting them understand how to leverage their (admittedly, distributed and caching 😅) compiler toolchains extremely well. So ... everything can scale. And everything can be written poorly. IMO at least.

    Comparing programming languages is super difficult, which is why I've historically selected them based primarily on what I know about the teams that're going to support them. I'd choose a vanilla TS/JS project ten out of ten times if I knew the crew that was running it was allergic to Kotlin Jetpack, but if the app devs needed to stretch their wings out into a small web service, I'd lean towards the tool they already know.

    (edit) Sorry; I let this go unstated in my original message, but to note it: I'm aware that this is a marketing stunt done by some egotistical jerks manipulating their respective communities into donating their time, money, or mindshare. That's just the core of what's disappointing here.

    14 votes

  8. Comment on Zig creator weighs in on the Bun Rust rewrite in ~comp

    kacey
    Link
    Holy crap, I'm never touching Zig or Bun with a ten foot pole. What a bunch of unprofessional jerks. So what they're saying is, if I use their obscure language and ask for help (or even donate!)...

    Holy crap, I'm never touching Zig or Bun with a ten foot pole. What a bunch of unprofessional jerks.

    The post claims they were fuzzing their Zig code, while during our calls the whole Bun team told us that they were not fuzzing anything. This appears to be an outright fabrication.

    So what they're saying is, if I use their obscure language and ask for help (or even donate!) -- then dare criticise them -- they'll whip out the call logs to start railing me in the socials? What the heck?

    Bun always gave me the ick, and Zig felt like a solution in search of a problem. I've had the privilege of working with genuine professionals: they could wipe the floor with me engineering-wise, and were level headed when the seas became rough. But these guys ain't it. They really deserved each other.

    19 votes

  9. Comment on Modern, abstract art makes me angry in ~arts

    kacey
    Link
    I don't have anything intelligent to say on the matter, but I wanted to share a piece that I'll probably never see in person, but that has interested me for years! 20:50 is sculpture (per the...

    I don't have anything intelligent to say on the matter, but I wanted to share a piece that I'll probably never see in person, but that has interested me for years! 20:50 is sculpture (per the artist) which is a room that appears half-filled with recycled motor oil, creating a darkened reflection of the ceiling and walls enclosing it.

    It really does feel like it's saying something about the modern world's dependency on oil, but after having watched that interview, it sounds like the artist just wanted to make a big mirror 😅 still seemed important to me.

    3 votes

  10. Comment on Fines doubled as teens outsmart Australia's world-first social media ban in ~tech

    kacey
    Link Parent
    Yup, thank you for mentioning that! Canada Post has a similar system where, for a fee, they can verify someone without their every needing to hand over an ID to Discord/Facebook/etc. -- this sort...

    Yup, thank you for mentioning that! Canada Post has a similar system where, for a fee, they can verify someone without their every needing to hand over an ID to Discord/Facebook/etc. -- this sort of stuff should have been put into law as a requirement, instead of allowing an unlimited number of third parties store sensitive personally identifiable information.

    5 votes

  11. Comment on Coffee co-fermentation, navel gazing or scandal? in ~food

    kacey
    Link
    I hadn't heard of this before, but it feels like it shouldn't pass the sniff test? Typically you ferment in order to get new, different flavours (often sour tones, but also richness/sweetness from...

    I hadn't heard of this before, but it feels like it shouldn't pass the sniff test? Typically you ferment in order to get new, different flavours (often sour tones, but also richness/sweetness from chemical breakdown), not the same flavours. Eg. marmite doesn't taste like sourdough, (sour) pickles don't taste like cucumbers, aged hard cheeses don't taste like milk, etc.

    You'd normally do an infusion with some sort of solvent in order to directly transfer flavours from one foodstuff to another, like when you dump herbs into vodka, or marinade a steak.

    But yeah I don't think anyone is a rube for thinking this could work; I don't know a tonne about how coffee is fermented, and it seems reasonable to trust manufacturers not to lie through their teeth.

    Edit: also fermenting one foodstuff with others is a practice as old as time, since often you want to inoculate the ferment with cultures that're often present on another. So that probably has a lot of history. But unless you incorporate that other item in the final product, or bathe it in solvent, you're relying on air transport and chemical volatility to get those flavour compounds into your target material.

    Also I should note that I'm just interested in this stuff and not an expert XD this is not culinary advice

    11 votes

  12. Comment on The Amazing Digital Circus - Ep 9: Remember in ~tv

    kacey
    Link Parent
    np! I distantly recall that there was something, but it's been so long now that I've forgotten. I think a few of the comments noted that the full cam copy that was floating around has it, if...
    np! I distantly recall that there was something, but it's been so long now that I've forgotten. I think a few of the comments noted that the full cam copy that was floating around has it, if you're interested?

  13. Comment on No, artificial intelligence is not conscious in ~tech

    kacey
    Link Parent
    Sorry to interject and nit pick this, but language is kinda my thing 😅 I can't speak to much, but I can say this with certainty: the 21 GB of 4 bit floats sitting on my SSD does not have a...

    [...] it's processing language in a way that is proximate in similarity to our mind's processing of language. That's the entirety of what an LLM is doing.

    Sorry to interject and nit pick this, but language is kinda my thing 😅

    I can't speak to much, but I can say this with certainty: the 21 GB of 4 bit floats sitting on my SSD does not have a Wernicke's area, and my brain does not have the latest FlashAttention optimizations installed. Equally, I can't keep ~2000 pages worth of content in my working memory (context, analogously), but an LLM trivially can and must in order to work properly.

    They seem pretty dissimilar IMO.

    4 votes

  14. Comment on The Amazing Digital Circus - Ep 9: Remember in ~tv

    kacey
    Link Parent
    Hmm. I think I'd disagree, but it's because I have a different interpretation of Caine's role in potential reversibility. An AU comic I read at some point put it well: Caine doesn't understand...

    I think abstractions being actually reversible would create some problems. It would imply that Caine could revert them all along but chose not to.

    Which, not only would make his redemption a bit harder to swallow, but also would be weird because he likes humans. You would think then that he would jump at the chance to fix the abstractions and bring them back.

    Hmm. I think I'd disagree, but it's because I have a different interpretation of Caine's role in potential reversibility. An AU comic I read at some point put it well: Caine doesn't understand humans well enough to be able to do the in-depth emotional surgery to re-alive them, so he'd need the help of the more capable circus crew to do the heavy lifting. We already saw via Pomni's perspective that Jax is reachable, and from some art released post-ep 9, not continuously on a rampage or anything.

    I dunno, I think it'd be cleaner from an overall message perspective to say that either (A) friendship and acceptance cures all (i.e. abstracted people can "come back" with enough love and time), or (B) some wounds can never be healed (i.e. Caine doesn't make a third act reappearance). I'd prefer A XD

    2 votes

  15. Comment on The Amazing Digital Circus - Ep 9: Remember in ~tv

    kacey
    Link
    Saw this one in theatres a few weeks ago! I liked it, but it's been a hot minute, and I didn't take notes on my thoughts at the time. Couple quick points in the spoiler block below. 1. I'm not...

    Saw this one in theatres a few weeks ago! I liked it, but it's been a hot minute, and I didn't take notes on my thoughts at the time.

    Couple quick points in the spoiler block below.

    1. I'm not fond of the implication that abstraction is irreversible. It really feels like Caine's arc is supposed to demonstrate redemption for one's acts, but where is that for Ribbit? And we never even _learn_ what happened to Queenie. Eternity is a long time to stay dead (or the next best thing). 2. OK but I still love Caine so I forgive Gooseworx for everything 3. The Jax character arc is ... interesting overall. Continuing from (1), it feels like the author is trying to say something about the magnitude of their crimes, but the point feels muddled. Cool to see more trans representation in media though! 4. Aww, y'all don't get to see the fish come back to tell you not to post spoilers. Now that the episode's out, it's probably kosher [to share it](https://www.youtube.com/watch?v=_NXC7SIZrAc). 5. I think it was about a ~6.5/10 overall -- quite watchable, and enjoyable overall!
    3 votes

  16. Comment on No, artificial intelligence is not conscious in ~tech

    kacey
    Link Parent
    Perhaps! I'm not super deep into the weeds on this, as I'm merely an amateur playing with local models, but my understanding was that MTP layers materially assist the optimizer in training the...

    I'd argue that multi-word prediction is still effectively word-at-a-time, just accelerated with a cheap model front-running the main model to break dependency chains and allow for parallelizing the main model.

    Perhaps! I'm not super deep into the weeds on this, as I'm merely an amateur playing with local models, but my understanding was that MTP layers materially assist the optimizer in training the model, completely distinct from their performance benefit. I suppose, the null hypothesis for "MTP layers are just an inference-time optimization" w.r.t. impacts on training is that they don't help, since they're inference-time only, but since we do see an improvement ... I dunno, maybe it suggests there's something more?

    I did a cursory Google search and turned up this ArXiv paper which talks about something similar, but tbh I don't know if I trust any non-peer reviewed ML papers after ~2024.

    It's interesting stuff! If this field weren't actively obsessed with destroying the world economy, and I were a decade or two younger, I'd probably be working on them professionally XD

    1 vote

  17. Comment on Should the plural of "milf" be "milfs" or "milves"? in ~humanities.languages

    kacey
    Link Parent
    Oh no I edited my post to remove what I felt was a subpar joke clause D: apologies

    Oh no I edited my post to remove what I felt was a subpar joke clause D: apologies

    3 votes

  18. Comment on No, artificial intelligence is not conscious in ~tech

    kacey
    Link Parent
    Mmhm, well put! I'm having trouble thinking of a sci-fi setting where a post-scarcity economy is portrayed poorly. Generally it's considered a time for celebration, but in reality, it seems likely...

    When you live in a world where your value as a person is based on your economic output, and something of equal output can be created on a whim, what worth do you even have anymore?

    Mmhm, well put! I'm having trouble thinking of a sci-fi setting where a post-scarcity economy is portrayed poorly. Generally it's considered a time for celebration, but in reality, it seems likely that the "haves" would wind up hoarding their infinite goods from the "have nots", justifying their actions the same way that charity has been withheld by generations prior in service of preventing moral decay (see famine walls for an example of this).

    8 votes

  19. Comment on Should the plural of "milf" be "milfs" or "milves"? in ~humanities.languages

    kacey
    (edited )
    Link Parent
    Hah, yeah, I littorally went overboard with that one :3

    Hah, yeah, I littorally went overboard with that one :3

    5 votes

  20. Comment on No, artificial intelligence is not conscious in ~tech

    kacey
    Link Parent
    That was a joy to read; thank you :) I agreed with 99% of what he said, and therefore I think his arguments are sound XD I'm not sure if you were linking in order to have a discussion, or just to...

    That was a joy to read; thank you :) I agreed with 99% of what he said, and therefore I think his arguments are sound XD

    I'm not sure if you were linking in order to have a discussion, or just to share the post, but if you were -- I'm not sure I agree with this statement:

    What I am saying is that if code like this—code that was not explicitly designed to mimic the architecture of an organic brain—ever does wake up, it will not be like us. Its natural state will not include pleasant fireside chats about loneliness and the Three Laws of Robotics. It will be alien.

    I think it's because I really want it to be true. To be witness to an inhuman intelligence would be a true privilege; we cannot know what we don't know, and interacting with something new could give us new perspectives with which to understand ourselves better. I dunno, it'd be a real philosophical moment for me.

    But LLM's ain't it. I would disagree with the premise that a lack of explicit intention of organic brain mimicry is necessary; on a fundamental level, Artificial Neural Networks are intended to mimic any function that they can observe outputs to. All of the machinery we've bolted on top of a basic, gigantic, fully connected network has been to make it faster to train and execute, but it shouldn't change the fact that they're still theoretically capable of approximating any function.

    I think the question I'd like answered someday is, is our speech all there is to be human? Or is there some necessary component in our minds which grows and acts independently across the course of our lives that makes us special, in some way? It's fascinating, if it weren't also connected to a 2.5 trillion dollar investment bubble that will either pop and destroy the world economy or directly replace every job with robots, and apparently those are the only two outcomes

    (edit)

    I do love how Blindsight seemed to foresee a lot of the qualities of modern LLMs a good 15+ years before their invention. The first-contact conversation is remarkably like talking to one -- a glib, slippery, humanlike persona trained entirely on intercepted human communications, something with no real inner life that makes subtle errors and seems impossible to pin down. (There's a similar scene in his Rifters books when talking to a "headcheese" neutral network.)

    Ach, and yes, agreed! I have only read Blindsight, however, but it was a nail on the head for the experience of chatting with an LLM for the first time.

    4 votes