Rage-Quit Maneuver

[JM]: I'd like to start off the show today by continuing our long tradition of mentioning interesting things that have recently appeared on neal.fun.

[JM]: Yes, once again, we're talking about n-e-a-l.fun.

[JM]: Because this week it seems there's a new project there called Sandboxels, which I imagine to be a clever mashup of Sandbox and Pixels.

[JM]: Because the idea is you have a graphical sandbox where

[JM]: And then you have a bunch of tools or materials with which to play in said sandbox.

[JM]: And there is a lot in here.

[JM]: You default to the land tab in which there's sand, dirt, mud, rock, snow, gravel, mulch, etc.

[JM]: And when you click and drag on the screen, you drop these various materials into your sandbox.

[JM]: And then there are additional tabs for liquids, plant and animal life, powders, solids, energy, weapons, gases, food, machines, and the special tab.

[DJ]: Oh, I hadn't looked at the special tab yet.

[DJ]: Oh my.

[JM]: Yeah.

[JM]: The special tab has fun things like antimatter, ice nine, Midas touch, gray goo.

[DJ]: Yeah.

[DJ]: Ice nine.

[DJ]: Are these all just things that destroy other things?

[DJ]: Yeah.

[DJ]: I imagine a lot of them are.

[DJ]: I mean, there's one labeled Sun.

[DJ]: This is the best.

[DJ]: We're just going to spend the rest of the episode playing with this.

[DJ]: It's amazing.

[JM]: You can put a virus, you can put malware into your sandbox.

[JM]: And then you can apply heat, subzero temperatures, lots of fun things you can do in the sandbox.

[DJ]: This is what I used to do in like MS Paint in 1994, but it actually has like a physics model imposed on it.

[DJ]: So it's incredible.

[DJ]: This is my new favorite thing.

[JM]: As with all things Neil.fun, definitely an entertaining way to spend a little bit of time for sure.

[DJ]: Oh man, there's the life tab.

[DJ]: You can actually just pour human bodies onto your, and then cover them with algae.

[DJ]: I'm not sure why you would do that, but.

[DJ]: And then dump acid on top of the whole thing.

[DJ]: Well, I don't know about acid.

[DJ]: I mean, but you could definitely use a heat ray.

[DJ]: Let's try that.

[DJ]: Oh, no, no.

[DJ]: There's acid in there.

[DJ]: I checked.

[DJ]: Oh, no.

[DJ]: I know there is.

[DJ]: I'm just saying, why would you do such a thing?

[DJ]: It's much better to incinerate everything with this heat ray.

[DJ]: Yep.

[DJ]: You're now listening to the Sandboxles podcast going forward.

[DJ]: It's just going to be us playing with this.

[DJ]: Oh, man.

[DJ]: Neil.fun, the website that doesn't stop delivering.

[JM]: Yeah, definitely batting 100% for sure.

[JM]: One of the things that I noticed at the bottom of this very fun page is a link to Discord, which brings us to one of the more significant stories of the week, which is that Discord has announced that it is rolling out Age of Verification on its platform globally starting next month.

[JM]: And it will automatically set all users accounts to a teen appropriate experience unless they demonstrate that they are adults.

[JM]: Discord's global head of product policy said, quote, for most adults, age verification won't be required as Discord's age inference model uses account information such as account tenure and

[JM]: device and activity data, and aggregated high level patterns across Discord communities.

[JM]: Additionally, Discord will implement its age inference model, a new system that runs in the background to help determine whether an account belongs to an adult without always requiring users to verify their age.

[JM]: Some users may be asked to use multiple methods if more information is needed to assign an age group.

[JM]: So I want to make sure that I have this straight.

[JM]: It has been a whopping, checks calendar, five months since Discord publicly disclosed a data breach which exposed Discord users' government IDs that were used for age verification.

[JM]: And now they want everyone...

[JM]: to submit to a face scan, or to present government ID to verify their age.

[JM]: That seems like a lot to ask users to do to say, yeah, we know we've demonstrated that we can't be trusted with your personal data, including images of your face and images of your government ID, but we're going to start requiring it anyway.

[JM]: And the part that's even worse, as far as I'm concerned, is that it's not just that Discord is requiring your photo of your face and or government IDs to verify your age.

[JM]: They're also going to be running this so-called age inference model.

[JM]: against everything that you do on Discord in order to infer your age.

[JM]: So when you're typing your chats, whatever you do, make sure you type like an adult.

[JM]: Because if you don't, you might be classified as a non adult.

[JM]: And thus, your access to parts of Discord will presumably be restricted.

[JM]: And we have talked about this whole age verification thing before.

[JM]: This just feels like this inevitable step further along a path we have already described, that we have already foretold.

[JM]: And we are creating, and by we, I mean companies like Discord and the governments that ask for, or I should say mandate, age verification.

[JM]: They are creating a two-tiered internet system.

[JM]: where we are going to see people who are willing to give up their selfies and their government IDs having one experience, while those who can't, won't, or don't have those things, for whatever reason, are going to miss out on those experiences or be outright excluded from entire areas of the internet.

[JM]: I don't know exactly where this is going to go.

[JM]: It seems pretty clear where it's going to go.

[JM]: But today, where we have some agency and we have some choices in terms of what we can do, I don't know why in the case of Discord, anyone would be willing to

[JM]: to give up your government ID to web services like this that have such demonstrated poor security in a time when identity is being weaponized by state actors.

[JM]: I have no idea why

[JM]: people would willingly hand this kind of personal data over.

[JM]: Because you can see how this is leading down a very dark path.

[JM]: Dan, you're a fiction writer, but you don't have to be a science fiction writer to be able to extrapolate where this all goes.

[JM]: So speaking for me personally, I saw this news.

[JM]: I have a Discord account that I created in 2018.

[JM]: And I immediately...

[JM]: logged in for the first time in a very long time, because this is not an account that I use very often.

[JM]: And I deleted my account.

[JM]: And I'm going to use scare quotes when I say deleted, because when you go to delete your account, you'll see a notice that says Discord account deletions take 15 days to process.

[JM]: Our support team is unable to speed up this process.

[JM]: And once you delete your account, if you try to log in, you see account scheduled for deletion.

[JM]: Your account is scheduled to self-destruct soon.

[JM]: Change your mind, restore your account.

[JM]: And a link where you can undo this so-called deletion and restore your account.

[JM]: So this to me is somewhat bad enough that presumably when you go to create your account, it's instantaneous, right?

[JM]: You'd have to wait 15 days when you create your account.

[JM]: But they make you wait 15 days to delete it.

[JM]: Because, of course, that only really benefits them.

[JM]: They're not doing this, I think, for our benefit.

[JM]: Like, sure, yeah, maybe every once in a while someone, I don't know, doesn't see all the confirmation dialogues and accidentally does it.

[JM]: Or does it in a rage quit maneuver and then the next day they...

[JM]: bemoan their action and wish they could undo it.

[JM]: Sure, maybe that happens.

[JM]: But as far as I'm concerned, there absolutely should be a button that says no, no, no, I'm so very sure I am once again, going back to the question of age, I am an adult, I am saying I am so very confident about this, I should be able to delete it immediately.

[JM]: But as if that weren't bad enough, the email confirmation that they send says the following.

[JM]: We have received a, dare we say, terrifying request to permanently delete your account.

[JM]: Your account has been deactivated and will be permanently deleted after 14 days.

[JM]: Just in case you change your mind or were a bit too curious with pushing buttons or did not request to permanently delete your account, please log into Discord to cancel this request.

[JM]: How very cheeky of them.

[JM]: They joke that my request is terrifying.

[JM]: How amusing.

[JM]: What's terrifying is everything else that I've just described.

[DJ]: What's terrifying is when some company who could not care less whether you live or die acts sad when you decline to use their services.

[DJ]: I was thinking, like, I think Justin's overreacting a little to the, like, you know, the waiting period, the grace period on account deletion, or at least I didn't feel as strongly about it.

[DJ]: But now I hate it because I do really hate that pattern.

[DJ]: And Discord is not the only one who does it.

[DJ]: Where, you know, a sad crying cat face emoji appears on the screen.

[DJ]: And it's just like, you want to leave?

[DJ]: And it's like, shut up.

[DJ]: Just delete my account.

[DJ]: Stop acting like you give a crap.

[DJ]: Like, stop trying to emotionally manipulate me.

[DJ]: It's gross.

[DJ]: Your guilt trip is not going to work, bro.

[DJ]: No, it's super gross.

[DJ]: So anyway, I don't like it.

[DJ]: And I will be going through this process soon to probably like I actually really did use discord a lot.

[DJ]: I can actually fathom why people would submit to the age verification stuff, even though it's bad for a bunch of reasons, because there are real relationships with other human beings at stake here.

[DJ]: Like this isn't just some service that, you know, oh, I use this to like back up my some personal like my notes or something like it's data that only I see and care about.

[DJ]: And, oh, I don't really need it or I've moved it somewhere else.

[DJ]: But it's really hard if you have like a bunch of friends on the Internet and the only way you can really keep in touch with them is through a service.

[DJ]: But if that service says, hey, we're doing something, we're going to require you to do something that you might not want to do.

[DJ]: it's genuinely difficult to go, oh, well, wait, if I stop using this, how do I talk to my friends?

[DJ]: That's a big thing.

[DJ]: It's why this age verification stuff is so insidious.

[DJ]: And it is really frustrating, as we've said in the past, that it's coming from our elected representatives.

[DJ]: This is not purely just avarice on the part of tech companies now.

[DJ]: Now it's like,

[DJ]: Yeah, absolutely.

[DJ]: Otherwise, we're going to assume you're a child and you can only look at the Disney branded version of this website or whatever.

[DJ]: And so actual minors will be in that position, which is fine.

[DJ]: But then anyone who doesn't want to go through these verification techniques is also is just going to have to find alternative places to go.

[DJ]: This is one of these attempts that like.

[DJ]: We'll have a bunch of negative second-order consequences and won't be effective in achieving its aim anyway.

[DJ]: And I think more to the – what's really frustrating about that is where's the follow-up?

[DJ]: Where's the accountability?

[DJ]: Who's going to be able to put out a report two years from now when it's give us your passport in order to look at this website or go away?

[DJ]: That goes, good news, everybody.

[DJ]: We can somehow quantify that we have protected 50,000 children from harm.

[DJ]: It doesn't work that way.

[DJ]: So we're essentially ruining a giant part of our shared economy.

[DJ]: global online experience and capability in return for nothing like something that a bunch of people a bunch of frankly ill-advised in my opinion at least people are saying oh this is going to have a bunch of positive consequences it's like is it I

[DJ]: I don't believe you.

[DJ]: And even if you're right, how are you going to prove that exactly?

[DJ]: I guess what I'm saying is call me cynical, but unfortunately, I think that children will continue to be harmed even after you can't do online shopping without submitting your fingerprint.

[DJ]: This isn't actually going to fix anything.

[DJ]: It's just going to ruin a bunch of stuff, which sucks.

[JM]: Precisely.

[JM]: It is security theater.

[JM]: when you say that there won't be positive consequences, oh, there are going to be positive consequences for the governments and the companies that they mandate do this kind of thing.

[JM]: There just won't be the positive consequences that you're talking about, the ones they say they care about, which is for the children.

[JM]: Instead...

[JM]: the positive consequences are going to accrue to the governments that mandate them and to the companies that have to implement these policies.

[JM]: Because as we've talked about before, it's hard to buy any argument that someone might make that

[JM]: These companies, oh, well, they're just mandated.

[JM]: They're told they have to do this by law, so they don't really have a choice.

[JM]: Okay, but let's just say for the sake of argument that some other piece of legislation, it doesn't matter what kind, let's just say that there's some piece of legislation that regardless of its content,

[JM]: is a mandate to companies that they do something.

[JM]: Again, it doesn't matter what.

[JM]: And as the end result of whatever this legislation is, the end result is that companies make hundreds of millions of extra dollars per quarter.

[JM]: Let's just say that that's an effect of some theoretical legislation or regulation.

[JM]: It's hard to feel bad or to say that these companies aren't complicit in some way when they clearly benefit from it.

[JM]: Because that is what is going to happen here.

[JM]: It's what's already happening here.

[JM]: By collecting images of your face and being able to match it against data brokers' data and doing the same thing with your government IDs,

[JM]: What better way is there to target advertising at you?

[JM]: What better way is there to monetize your data than by knowing exactly who you are, where you live, all of the other information that's on your government ID?

[JM]: Look, I didn't use my Discord account very much.

[JM]: The only reason I even created it is because there were certain communities, say if I were going to speak at a conference, oftentimes the way that you collaborate or interact with

[JM]: other conference goers is in a discord that I'm not choosing the conference organizers are choosing that.

[JM]: So that's why I created the account in the first place.

[JM]: But this wasn't an account that I used very much.

[JM]: So discord has not really lost very much by me deleting my account.

[JM]: But deleting it is what I can do.

[JM]: So it was important to me that I do it mainly to send a signal that what discord is doing is not okay.

[JM]: Hopefully enough other people will do the same, making it clear to companies like Discord that their actions against our privacy, security, and liberty come with significant negative repercussions for them.

[DJ]: I agree with that.

[DJ]: And as someone who has used Discord, it is for me like the principal way that I stay in touch with some of the people is

[DJ]: In my life, I'm going to be trying to move some people to alternatives.

[DJ]: In my case and in the specific cases I'm thinking about, it's probably going to be Signal.

[DJ]: Signal is not the perfect alternative for a bunch of things.

[DJ]: And we'll have a link in the show notes to a good article about Discord alternatives.

[DJ]: Because a lot of people use Discord as like group conversations and community conversations.

[DJ]: conversations.

[DJ]: I personally don't really do that, so I'm mostly looking for an alternative messaging app.

[DJ]: Signal is already a better choice for that since it's end-to-end encrypted and Discord is not, but I've just never had a compelling enough reason to try to take certain conversations off of Discord, but now I think I do.

[DJ]: So speaking of links, one other thing that I want to point people towards is a Discord chat exporter.

[DJ]: This is an open source app.

[DJ]: It's available on GitHub, and it does a really great job of pulling all channels and direct messages out of Discord into data files that you can then do whatever you want with.

[DJ]: I've used this before in the past because it's important to me to at least occasionally exfiltrate data.

[DJ]: I'm really fond of the word exfiltrate.

[DJ]: I think it's a cool word.

[DJ]: So that's how I always describe it.

[DJ]: It's important to me to occasionally exfiltrate

[DJ]: pull my data out of various services in case of situations like this, where I do want to stop using them.

[DJ]: When I quit using Twitter, the very first thing I did was, okay, what are the ways to get your data out?

[DJ]: To the website formerly known as Twitter's former credit,

[DJ]: They did at least make it easy to just dump everything out of your Twitter account, including direct messages with all the media, etc.

[DJ]: That's cool.

[DJ]: Discord, the last time I checked, does not make that as easy.

[DJ]: But there is this really good tool out there that you can use to do it.

[DJ]: So I encourage people to do likewise and look for alternatives, knowing that, you know, you might lose contact with some people who are only going to talk to you through Discord.

[DJ]: But at the very least, if you have past data from past conversations that you care about, there is a way to get it out of the service.

[JM]: Indeed.

[JM]: And as a reminder, we will have links in the show notes to how to delete your Discord account, how to export your chat history, as you mentioned, Dan, and also some alternatives that you might use when you are trying to find a chat tool that

[JM]: for your next conference or whatever community that you want to curate.

[JM]: And I was already thinking of what are the discord alternatives even before I saw this article that describes them and I thought before I tapped on it, I wonder if they'll mention

[JM]: any really good options that I haven't thought of.

[JM]: And I have to say, sadly, that I didn't really see any in there, because there really just aren't great options for this.

[JM]: Unfortunately, I think that if signal does what you need it to do in terms of communicating with other people, it is an excellent tool.

[JM]: And if all you're trying to do is send text messages to your friends, I think you can stop looking.

[JM]: It should just be your go to solution for some other use cases where you're trying to have a larger group conversation, like in a chat room signal is less well suited for that use case.

[JM]: But I think if I had to find a solution for that tomorrow.

[JM]: I would probably choose Zulip, which is an open source written in Python and Django, if memory serves, tool that kind of combines chat and elements of forums.

[JM]: So it's a little bit of a odd duck in this space, but it's something worth looking at if you're trying to find an alternative for your next event or community.

[JM]: All right.

[JM]: In other news, Dan's favorite music streaming service and the one I refuse to even try to pronounce.

[JM]: So this is your cue.

[JM]: Are you talking about QuoBuzz, Justin?

[JM]: I think you must be talking about QuoBuzz.

[JM]: Yes, that is the one that we were talking about.

[JM]: I just need to get you saying it like on a soundboard so I can just push it every time I need to say it.

[DJ]: Yes, you do.

[DJ]: You just need a soundboard.

[DJ]: And actually, you should probably play that in between every segment, even when they're not QuoBuzz related.

[DJ]: Just it'll kind of like wake people up, you know, like get them to pay attention.

[JM]: This is a fabulous idea.

[JM]: Let's do that.

[JM]: So that streaming service...

[JM]: is taking a human-first stand on software-generated music.

[JM]: And they have released a post where they talk about music being about human expression, and they are...

[JM]: underlining their commitment to focusing on human music.

[JM]: So they describe their human first approach while acknowledging that it is not always possible to perfectly detect when a piece of music is software generated and not generated by a human.

[JM]: So they're trying to be realistic about the challenges and difficulties in detecting software generated music.

[JM]: But their approach is that their editorial and human curation will remain 100% human.

[JM]: So their album of the week, playlists, and reviews will be from their team of music experts.

[JM]: And their recommendations are going to highlight human artists.

[JM]: They say they're going to filter out software-generated tracks.

[JM]: They're going to identify fraudulent streams, excluding them from reporting and royalty statements so that humans are getting the appropriate visibility and compensation.

[JM]: And they commit to protecting listeners' access to human-crafted music and the artists who create it.

[JM]: And they've gone on to link to their AI charter, which details their principles, ongoing commitments and challenges that they're going to work through.

[JM]: And I am really happy to see this, right?

[JM]: Like a lot of times things that we talk about on this show are, you know, sometimes not super uplifting as you know.

[JM]: And this was cool to see.

[JM]: It's cool to be able to highlight a company that's doing something positive and trying to do the right thing.

[DJ]: Yeah, a lot of what we talk about is technology companies being bad.

[DJ]: This is at least one example of one company being good, I think.

[DJ]: And as a user of this service, I switched to it sometime last year.

[DJ]: I...

[DJ]: was happy to see this proclamation, it makes me feel additionally confident, like, oh yeah, okay, I backed a winner, you know, like I picked, I made the right choice.

[DJ]: And I also like that in this statement, there's some nuance to it.

[DJ]: Like they acknowledge the use of so-called like AI driven, you know, we tend to refer to like large language model driven software because that's what we're really talking about.

[DJ]: They acknowledge the use of those tools in the mixing and mastering and even composition process of music by people.

[DJ]: What they are objecting to is the complete generation of music through just prompting, like people using a large language model and giving it a prompt like generate a hit song in the style of whatever, and it'll do it.

[DJ]: They do an impressive job at this point in time.

[DJ]: But a lot of people inside and outside of the music industry really want to preserve, I would say rightly, and want to enshrine the importance of a person actually being deeply engaged in the process of composing music.

[DJ]: So I like the fact that they're making that distinction.

[DJ]: If this isn't like a reactive AI is bad kind of thing...

[DJ]: They're saying, okay, well, these tools are making their mark on the music industry in a variety of ways.

[DJ]: Some of them are empowering human beings to perform the act of creating music, while others are, for better or worse, empowering people to use software to create music that you can't really say the human composed can.

[DJ]: That's not as straightforward a distinction as it seems on the service.

[DJ]: So we don't need to get too much further into that necessarily.

[DJ]: But I like the fact that Qobuz is at least pointing it out in their statement.

[DJ]: So yeah, I'm pretty pleased to see them do this.

[DJ]: And just in general, I've been pretty happy with them as an alternative to the music providers I was using.

[DJ]: So go Qobuz.

[DJ]: You have a silly name, but we love you.

[DJ]: Also, I think I accidentally slipped into pronouncing it the way they want you to, which is co-buzz, which generally I resist because it's much funnier to pronounce it the way it's spelled, which is quo-buzz.

[DJ]: At least that's how I think it's spelled.

[JM]: I was sitting over here being like, should I point this out?

[JM]: That first it was quo-buzz and then it was co-buzz?

[JM]: I'm like, no, I'm just going to let it go.

[JM]: But I like that you decided to do it anyway.

[DJ]: That's right.

[JM]: Quo.

[DJ]: Never forget, people.

Yeah.

[JM]: As usual, I'm going to put a link to this in the show notes, but by all means, take a look at their charter because they talk about these are our principles.

[JM]: This is what we do.

[JM]: These are our red lines.

[JM]: We will never do this.

[JM]: We will never do that.

[JM]: We will protect this.

[JM]: We'll protect that.

[JM]: What are our ongoing commitments?

[JM]: What are our limits in gray areas?

[JM]: This is just a really cool document.

[JM]: I wish more vendors produced documents like this for all kinds of things.

[JM]: Like this is our position on these topics.

[JM]: And what a great differentiator for QuoBuzz to...

[DJ]: There it is.

[DJ]: I knew you'd join me someday.

[DJ]: I'm so proud of you.

[JM]: I did that just for you.

[JM]: I love it.

[JM]: What a great differentiator for them because you know, you're never going to see anything like this come from Spotify.

[JM]: And I think it's great that when someone asks me, Hey, what streaming service do you use?

[JM]: I'm going to say, well, none of them, but if I were to use one,

[JM]: I would definitely use Quobas for these reasons because they have principles.

[JM]: They have posted this long document talking about what they're committing to.

[JM]: Hats off.

[JM]: It's great.

[JM]: OK, there has been a lot of talk recently about large language models and how they are affecting work in many different ways.

[JM]: And I've seen a lot of anecdotal posts and comments talking about the ways in which using them to, say, generate code, for example.

[JM]: can be challenging for the people who are doing it.

[JM]: But this week saw the introduction of a report that quantifies this to some degree.

[JM]: So instead of it just being anecdotal, we're looking at a study of 200 employees at a United States-based technology company.

[JM]: And the study was performed from April to December of 2025.

[JM]: And these initial findings come from Berkeley's Haas School of Business in a report titled AI Doesn't Reduce Work, It Intensifies It.

[JM]: And the general gist of this report is that regardless of how much or little these tools increase productivity, the actual work of managing productivity

[JM]: this process of using large language model agents to generate code seems to be an exhausting exercise.

[JM]: And I'm going to read a brief part of this report in which they said, AI introduced a new rhythm in which workers managed several active threads at once.

[JM]: manually writing code while AI generated an alternative version, running multiple agents in parallel, or reviving long deferred tasks because AI could handle them in the background.

[JM]: They did this in part because they felt they had a partner that could help them move through their workload.

[JM]: While the sense of having a partner enabled a feeling of momentum,

[JM]: The reality was a continual switching of intention, frequently checking of AI outputs, and a growing number of open tasks.

[JM]: This created cognitive load and a sense of always juggling, even as the work felt productive.

[JM]: And I've seen comments...

[JM]: that describe the same basic concept, both before this report was released and in its aftermath.

[JM]: It seems like a lot of people are finding that they are able to get so much more done, but after a brief amount of time doing it are feeling very depleted in ways in which they did not feel before they started using these tools.

[JM]: And I can get the allure, right?

[JM]: Like you feel like you're getting a lot done.

[JM]: You feel like there's all this momentum.

[JM]: And even though you know, you're feeling depleted, and you should probably take a break, do something else.

[JM]: There's just this pull like, okay, one more prompt, so I can get this one more feature built.

[JM]: And it feels a little bit concerning, particularly when we hear stories of companies that are pushing these tools on their employees, that there could be we still don't really know, but there could be a significant productivity boost in time, on average across the board, but at what cost?

[JM]: if the actual work itself feels so depleting and so exhausting.

[JM]: We already have, I would argue, a burnout problem when it comes to creative work, software development, and I don't know how this is going to affect burnout in these industries.

[JM]: And I think it's also interesting, by the way, that this report was published in the Harvard Business Review, which is not exactly a bastion of pro-labor policy.

[JM]: This is a relatively pro-capital system.

[JM]: that produced, or I should say published this report.

[JM]: So if the folks at the Berkeley Hospital of Business and Harvard Business Review are saying this, to me, it carries a little bit more weight, perhaps, than if it came from another source.

[DJ]: It is nice to see research being done on this.

[DJ]: For the purposes of our show, I can switch back into anecdotal data, which I have sometimes heard described as anecdata, which I like.

[DJ]: I can switch back into anecdata because I've been experiencing this exact thing.

[DJ]: So like I saw a link to this article yesterday and I shared it in my work Slack where we have a variety of channels and we talk about these tools and some of them because I can...

[DJ]: already feel this like i'm starting to do more and more of what we've been calling like spec driven development where you spend most of your time instead of writing code as a developer especially a senior developer you spend a lot of your time writing a spec you give it to one of these code generation tools to work through these tools at this point in time are really doing quite an awfully good job of that compared to where they were even six months to a year ago and

[DJ]: which is exciting to see.

[DJ]: And then you review the results and then you keep going.

[DJ]: And occasionally, I feel like more and more, you start making a choice about, instead of you're definitely going to write some of this code yourself, you start making a choice about, okay, I want this to be slightly different.

[DJ]: Should I just step in and change it?

[DJ]: Or should I put in a prompt and be like, please change...

[DJ]: And I'm finding increasingly it's actually the latter.

[DJ]: The number of places where I actually step in and start typing code into a code editor is going down.

[DJ]: However, my mental load is not going down, I would say.

[DJ]: And in some ways it's actually increasing.

[DJ]: And I think this is going to differ from person to person because we already are so different in how our brains work.

[DJ]: I think this is part of why the burnout problem that you just identified is such a pernicious and challenging thing is because we sort of have these expectations for how everyone should work.

[DJ]: That are some baseline.

[DJ]: It's like, here's the expectation for everybody.

[DJ]: And some people really thrive under those expectations and some people don't.

[DJ]: And so I would argue that our goal should be to find ways of working that suit the people doing the work.

[DJ]: And I think these tools can help.

[DJ]: But we're in such early days and there's so much excitement about it that I think it's easy for it to go the other way.

[DJ]: And instead, people are like getting out over their skis, as the saying goes here in ski country, where they're just doing more of the kind of thing that will burn them out or they're switching from something that was sustainable to something that wasn't.

[DJ]: When I've been trying to analyze this for myself, like why does this feel so different, I think part of it is switching.

[DJ]: It's the cognitive burden of task switching, which I actually have a lot of trouble with.

[DJ]: Like I recently moved to a new living space, and I've noticed that my body is not used to where everything is anymore.

[DJ]: Where's the edge of that countertop?

[DJ]: Where did I leave that thing?

[DJ]: Like it's a different set of rooms.

[DJ]: You don't necessarily think about how much of your life runs on like habits and automated behavior.

[DJ]: And so when you do things like switch your living space and now like your keys just aren't in the place where you're used to your keys being, it creates this little mental load.

[DJ]: And those loads accumulate and they're exhausting.

[DJ]: They actually do take a toll on your body because your brain is part of your body.

[DJ]: Just like picking up heavy things and putting them back down, right?

[DJ]: Like your body gets exhausted.

[DJ]: And I think that using these code generation tools...

[DJ]: especially when they go off and do so much work on their own and then you have to come back to the context that they were operating in and review the work and set it off again.

[DJ]: I think that's doable.

[DJ]: I think it might even prove to be highly productive, but it is a different way of working.

[DJ]: And I'm glad to see people acknowledging this because I think the solution is to treat it as a new way of working.

[DJ]: Like I've been starting to think of it as, oh, okay, this is a new skill I have to learn.

[DJ]: Like, I'm already very proficient at programming computers, but now I have to learn how to kind of manage these asynchronous tasks.

[DJ]: And I'm not saying, and I hate that, and therefore I'm going to change careers and become an airline pilot.

[DJ]: I'm just saying that it is a challenge.

[DJ]: Yeah.

[JM]: I think it's interesting that this report came out right around the same time that Anthropic released changes to Claude in terms of how it works.

[JM]: Because now, I believe by default, when you invoke Claude code, instead of harnessing a single agent, it will spawn multiple agents that will interact with one another to perform the task that you've prompted it to do.

[JM]: And I don't have any direct experience with this feature.

[JM]: So I don't know if it adds to this concept of cognitive load and the number of things that you have to juggle or whether it improves it and abstracts it away so that you don't have to do as much of it.

[JM]: And it's just handled somewhat transparently for you by the system.

[JM]: I have no idea.

[JM]: I also want to read you this one quote that I saw in the aftermath of this news because I thought it was interesting.

[JM]: LLM AI programming agents are not good for mental health.

[JM]: It supercharges FOMO and takes procrastination to the next level.

[JM]: I have about 20 large change sets across two computers and five repositories generated with AI agents.

[JM]: ready to be pushed that I just can't force myself to review or take a look at.

[JM]: Meanwhile, there are small changes my apps need that I'm avoiding doing, because I have to be on the future train.

[JM]: Now, obviously, some of this could just have to do with this particular person's reaction to these new tools.

[JM]: As you said, it's a skill.

[JM]: and a technological development that presumably requires some adjustment and adaptation.

[JM]: I do wonder just as an offhand thought, it seems to me that a lot of people that are attracted to software development

[JM]: seem to be good at focus and concentration.

[JM]: That's why they're good at it.

[JM]: So there could be a little bit of a self selection, where the kinds of people that are doing it are the people that naturally are this way.

[JM]: Or perhaps they become this way over time as they hone their craft.

[JM]: But either way, now, they're being thrown a very significant curveball, which is

[JM]: This skill or innate property, whichever it is, maybe it's both, that relates to focus is now about managing a lot of context switching that feels perhaps really chaotic, particularly compared to the way that software development has been done up until this point.

[JM]: And I think that's what this person who said this is trying to communicate.

[DJ]: That's a great observation.

[DJ]: I'm really sad that the person who you quoted feels the way they do.

[DJ]: That sounds really hard.

[DJ]: And as they said, if someone says this is bad for my mental health, I believe them.

[DJ]: I think there are maybe different approaches that one could take.

[DJ]: Maybe that person would benefit from them.

[DJ]: If they were asking me, if I was their coach in this regard, I might suggest titrating their exposure to these tools.

[DJ]: So I came across a tip many years ago before LLM tools were a thing.

[DJ]: It was like a tip for freelancers in technology.

[DJ]: And the person was talking about how there are so many tools out there, right?

[DJ]: And it never feels like the right time to stop doing client work so that you can try to figure out if like this task manager would be a better fit for you than the task manager you're already using.

[DJ]: And the advice was block out time in your calendar, like actually pick a day, like two hours on Friday afternoons or something.

[DJ]: It's tool time where you deliberately set aside time to try out new tools in a very focused way.

[DJ]: And I think one of the benefits of doing that is exactly as you said,

[DJ]: Those of us who thrive on the ability to immerse ourselves deeply in a single task for long periods of time can find it very challenging and disruptive to try to bounce back and forth between multiple things.

[DJ]: And so like that person said, I have all these different change sets that have all been generated, and I'm paraphrasing what I heard.

[DJ]: They're totally overwhelmed.

[DJ]: So now they're not even but they're also not making like the small changes that they need to make because they feel like their attention is supposed to be over here with these tools.

[DJ]: My suggestion would be really box your time for these tools.

[DJ]: And if that means that you can't get through 20 change sets, like even though these tools can generate a ton of work.

[DJ]: But maybe you can only review 5% of it this week.

[DJ]: Well, then just accept that that's fine.

[DJ]: That's what you can do.

[DJ]: And unless someone is really putting unsustainable external expectations on you, like, hey, why haven't you reviewed...

[DJ]: this 200,000 word manuscript of my novel by tomorrow.

[DJ]: It's like, okay, well, that's an unreasonable expectation, right?

[DJ]: Like we're going to see unreasonable expectations here too.

[DJ]: It's like, well, why haven't you shipped a year's worth of software engineering work in a week?

[DJ]: And it's like, well, just because these things can generate code doesn't mean we can ship it all.

[DJ]: And if we can't ship it all, then we're still limited by how much of it we can review effectively.

[DJ]: That's a bottleneck, right?

[DJ]: it's okay for that bottleneck to exist.

[DJ]: I think that's what a lot of us are starting to feel weird about is it doesn't feel okay for there to still be like human bottlenecks in the system if we're all kind of expecting, oh, like now that we can generate 20,000 lines of code in an afternoon,

[DJ]: therefore we should be shipping 20,000 lines of code every day.

[DJ]: Well, we can't.

[DJ]: And we might eventually get to a point, as I've seen other people talking about, where even the act of reviewing code shifts from actually looking at lines of code to essentially specifying, okay, robots, you're going to generate a system.

[DJ]: Here's how we verify it works.

[DJ]: I'm not going to read the 50,000 lines of code that you generate, but...

[DJ]: Here's a verification suite or something that it has to pass.

[DJ]: There's still a bunch of human labor in there one way or the other.

[DJ]: But I think as we get better at deciding which labor to do in what part of the system, these things will start to feel more sustainable.

[DJ]: But in the meantime, I really do want to encourage people to set reasonable expectations for themselves, even if external expectations may or may not be reasonable.

[DJ]: Set reasonable expectations for yourself and do what you need to do so that your exposure to these tools works for you.

[DJ]: And if that means you only use them on like Thursday morning for a couple hours, that's fine.

[DJ]: Do that.

[DJ]: That's much better than like burning yourself out feeling like you're missing out on the future.

[JM]: I think another reason that people might feel like they are missing out is this feeling that the agent shouldn't be idle.

[JM]: If you're working with other humans and you know that someone is blocked because they are waiting for this thing for you to do so that they can continue doing the thing that they're supposed to do next.

[JM]: If you feel like you're the blocker, you feel a certain pressure to not be the blocker.

[JM]: And I think that there may be some aspect of that going on here where we're so accustomed to that feeling of not wanting to be the blocker that when your software generation agent is sitting there idle, either because it's waiting for you to answer some question that's being asked of you or it's finished and now it's your turn to either review what they did or to provide another prompt.

[JM]: you're now the bottleneck.

[JM]: And so I would think that that might be part of the angst that people feel when they're using these tools.

[DJ]: Yeah, great observation.

[DJ]: I really do think you're right, and that's part of it.

[DJ]: And I think the root cause there is a sort of unhealthy and unrealistic notion of what efficiency is.

[DJ]: And this sort of clinging to an idea that there is some way to be perfectly efficient.

[DJ]: And that's a good on its face, right?

[DJ]: That it is necessarily better to be maximally efficient.

[DJ]: Like 100% efficiency is obviously better than 90%, which is obviously better than 70%.

[DJ]: But like a lot of human ideas, just because we can have that idea doesn't mean we can practice it.

[DJ]: And also doesn't necessarily mean we need to practice it.

[DJ]: The reality is I'm pretty sure like no physical system is perfectly efficient and yet that's okay.

[DJ]: And especially like you're right that we do feel pressure, especially social pressure when we're like a bottleneck on our peers.

[DJ]: The fact of the matter is.

[DJ]: Sometimes, as much as you might like to snap your fingers, you might just have to say, hey, you know what?

[DJ]: I'm really sorry.

[DJ]: I'll get that to you as soon as I can, but it might not be before the end of the day.

[DJ]: And usually that doesn't destroy your relationships or cause your company to go bankrupt.

[DJ]: It's just the way it is.

[DJ]: You try to work as efficiently as possible.

[DJ]: But perfect efficiency is not possible.

[DJ]: So I think you raise a really good point that it's the same with these tools.

[DJ]: Like just because this agent can finish a task in five minutes does not mean I'm going to be ready to review its work and move on five minutes from now.

[DJ]: Maybe I'm not ready for a half an hour.

[DJ]: And it's like, oh, but that's 25 minutes during which the robots could have been building some other thing.

[DJ]: And I think that while that's true, it kind of doesn't matter.

[DJ]: It kind of needs to not matter that much because we just can't be these perfectly efficient machines, right?

[DJ]: Like we're not machines.

[DJ]: Machines are also not perfectly efficient as part of my point, but certainly human beings are.

[DJ]: So like I really do hope people can, as we move through our excitement about the capabilities of these new tools, come back around to an understanding of our own limitations.

[DJ]: Like we have those limitations and it's okay to have them.

[DJ]: It does not reduce your value as a human being no matter what, I don't know, the economy tries to tell you.

[JM]: Indeed.

[JM]: One final note before we adjourn for the day.

[JM]: I just wanted to highlight that OpenClaw, previously referred to as ClawedBot and then MultBot, is still named OpenClaw.

[JM]: And I have to say, I did not see that coming.

[JM]: It has lasted for a week and I am surprised.

[DJ]: Yeah, it's interesting.

[DJ]: I wonder if they've reached a steady state yet or not.

[DJ]: I mean, we'll continue to report on this, obviously, in future episodes.

[DJ]: We can call the segment Claw Watch, but then we'll rename it Malt Watch.

[JM]: Claw Watch, that's perfect.

[JM]: There have been a lot of things related to this project over the last week, besides changing names, including, among other things, the realization that some of the things being posted to MaltBook.com

[JM]: a so-called social network for AI agents, were in fact being posted by humans.

[JM]: What?

[JM]: Which is really funny, right?

[JM]: Because sometimes it's done for entertainment value, right?

[JM]: Where a human is saying, oh, I'm going to mess with this and post something funny.

[JM]: And also a human could just post really malicious stuff because the whole point of this so-called social network is for other agents to consume the things being posted on it and act upon it.

[JM]: So imagine the mischief you could create as a human maliciously putting instructions into a mold book as if you're a software program doing it, because that is, after all, the point of this thing.

[JM]: And that raises this question of, okay, so we have CAPTCHAs for proving that a human is a human, but I don't think we have a way so far of proving that a piece of software is indeed just a piece of software.

[DJ]: Oh, like verify that you're not a human is what you're saying.

[DJ]: Exactly.

[DJ]: That's probably actually really easy because you can ask a piece of software to do something that a human probably couldn't, like do some complicated math equation in less than 500 milliseconds or something.

[DJ]: But I think that this is really funny for another reason, which is that for a large number of years now, we've been concerned with so-called bots, like automated posts corrupting the content on human social communities.

[DJ]: And now we're talking about the opposite, which is this is supposed to be a social network for bots.

[DJ]: What are these humans posting for?

[DJ]: How do we stop them?

[DJ]: Up is down and black is white, Justin.

[DJ]: It's official.

Yeah.

[DJ]: The world has turned upside down.

[JM]: And we have mold book to thank for all of this amusement.

[DJ]: Yeah.

[DJ]: I mean, I think it's called open book now.

[DJ]: Claw book.

[DJ]: This really is the gift that keeps on giving.

[JM]: And on that note, thanks everyone for listening.

[JM]: That's all for this episode.

[JM]: You can find me on the web at justinmayer.com and you can find Dan at danj.ca.

[JM]: Reach out with your thoughts about this episode via the Fediverse at justin.ramble.space.