Transcript

[Banjo music]

MANTRA: Just Build Websites!

Dave Rupert: Hey there, Shop-o-maniacs. You're listening to another episode of the ShopTalk Show, a podcast all about front-end Web design and development - and more.

Chris Coyier: [Laughter] And more?!

Dave: Hey, Chris. How are you doing today?

Chris: Oh, gosh. I just couldn't be better, really.

Dave: Oh...

Chris: It feels like everything is perfect right now.

Dave: Everything is perfect.

Chris: Just kidding. But this show is perfect.

Dave: [Laughter] All my kids are healthy. All kids are healthy. We are doing great. No problems.

Chris: [Laughter] Yeah. I was cleaning definitely out in the snow. I'm glad I didn't take my hose off the house, which is kind of a mistake. I feel like you should kind of spin that stuff down in the winter where it gets real cold like where we live, which I didn't. But I'm glad it wasn't because I had all the mats out of my car, just spraying puke off of them in my snowy front yard. That's where we were at yesterday afternoon.

Dave: Yeah. My son, every time there's a three-day weekend, like President's Day, he decides he's also sick the next day. Statistically, he misses the day after a long weekend. Man, as a parent--

Chris: This has already been long? Yeah.

Dave:; Yeah. It's like, "I got to get back to work." It's a tight week of work here, bud, so I've got to get back into it. Oh, well. He's at the age he can just watch TV all day, so he just watched Adventure Time while I worked.

Chris: Oh, good choice.

Dave: Then he came into my office.

Chris: Yeah.

Dave: I was on a call, Chris.

Chris: Yeah.

Dave: I kid you not. I'm on a call with two women and he farts. He cuts a fart.

[Laughter]

Dave: And the two women I was on a call with were cool and did not say anything. But they didn't know he was in my office.

Chris: Oh, no! [Laughter] You're like, "If you heard some flabergastation, that was my son."

Dave: So, what do I do, Chris? Do I follow up? This is my advice show. What do I do, Chris? Do I follow up with these people and say, "I didn't cut the beef," or do I just let it go?

Chris: Just go full silence?

Dave: Just go full silence?

Chris: Oh, no. I would. You know those red buttons that they sold at, like, Staples - or whatever?

Dave: Yeah.

Chris: I would write in some kind of white paint on it, "Fart Button," and put it somewhere visible in your office. Next time you have a meeting with them, be like, "Oh, he probably just leaned his elbow on the fart button."

Dave: Yeah. Yeah.

Chris: You know?

Dave: Yeah, okay. Okay, so some sort of elaborate ruse to cover it up.

Chris: Absolutely.

Dave: Yeah? Okay.

Chris: Yeah. I would never fart in a meeting with my butt. But I did it with a button.

Dave: No, no. I don't do that. But you know, at least I'm on mute. No. Yeah, it was... I was just like... [Laughter]

You know thankfully I wasn't on camera. Or maybe I should have been on camera because I just turned around and was like, "Dude! What?!"

[Laughter]

Dave: Anyway, it's just whatever, man.

Chris: Yeah. You learn those skills later.

Dave: Life with a pre-teen over here. So, there you go.

00:03:22

Chris: We have some questions we can get to. We have some weekly news that came up. We didn't mention that Interop dropped. I guess it's almost two weeks ago now, or is it, or a week and a half or something?

Dave: Yeah.

Chris: By dropped I mean they have announced what the topics for Interop will be. It's always an interesting process because you know people do... There is like a GitHub repo or something.

You publish ideas, and you don't really vote. There's no voting. But there is a lot of, like, thumbs up and stuff. In years past, I've tried to compare what got chosen to how many up thumbs things got.

Dave: Mm-hmm. Mm-hmm.

Chris: It was a little correlation but not much. I also went into it very open-minded like I don't actually care, though. I don't think that things should necessarily be chosen by up votes because there's stuff that needs to get... It's almost like it would be just as valuable if you picked the things with the least up votes.

Dave: Right.

Chris: When it comes to Interop.

Dave: Right.

Chris: Sometimes stuff needs to get fixed that isn't very hot.

Dave: Right.

Chris: But the list is out. The Igalia podcast fellas did a good episode on it. I don't think we need to go up and down the list necessarily other than to say the two ones that I think people are going to be like, "Hell, yeah!" about is the anchor positioning and popups--

Dave: Yeah.

Chris: --because they go together like peanut butter and jelly, baby. And if they worked across all three browsers, I think that's going to have a big impact on stuff.

That's not to say that everything else on this list isn't also awesome. So, please do all of it. I like all of it. And whatever your process is, please keep doing it forever.

00:05:00

Dave: Yeah. Yeah, it's like I want all of this stuff. Navigation API was kind of like a cool surprise, I guess. I think that really fixes a lot of JS routing. At scope is cool, but I feel like it's in... I guess it's only in Chrome maybe right now.

Chris: I'm working on a talk about scope in CSS. I'm going to be at CSS Day, I think it's called, in Amsterdam.

Dave: Oh, yeah. Yeah. Nice.

Chris: It should be interesting because obviously there's a lot more to talk about with just the concept of scope, which is my talk.

Dave: Mm-hmm.

Chris: Of which I think @scope is actually kind of a small part of it. It's not that I totally despise it, but I'm just like, "Eh!" You know? I don't know. It's just not my favorite CSS feature, particularly because I work on lots of different CSS and never look at that and be like, "Ooh! I'm going to use that."

Dave: Mm-hmm.

Chris: It just has no "Ooh! I'm going to use that feature" to me. I say that when it just came up yesterday at work where we were looking at ways to do... We have a little bit of light mode dark mode stuff at CodePen but mostly not. We're going to more lean into it fully when the new editor comes out. When you're designing a component-based website that uses custom properties for light mode dark mode (or potentially lots of modes), it really is nice to kind of scope the variables and stuff to individual components and possibly sometimes be able to have those components in a different color state than their neighbors or parents.

Dave: Sure. Yeah.

Chris: There is some stuff in @scope in CSS that does assist with that a bit.

Dave: Mm-hmm.

Chris: What do they call that? Proximity scoping is actually kind of interesting, so I might. Who knows? Maybe it'll be next week; I'll be like, "Dave, have you seen @scope?! It's amazing!"

00:07:00

Dave: Right. Right. Right. One other one that was kind of cool, well, view transition APIs, so that's rad.

Chris: I think they specifically scoped it to a single SPA style.

Dave: Oh, the SPA style? Boo!

Chris: [Laughter]

Dave: Everyone stinks!

Chris: But does it really say that? I'm reading the Read Me here. Focus. Focus our proposal for same document.

Dave: Same document transitions.

Chris: Yeah, but it's like that stuff is so tide together that I would think that a browser team digging into view transitions, like, couldn't you see I'm just being like, "Screw it! Let's just get it done while we're in here"?

Dave: Yeah. I actually... Safari has it. Safari has this weird back behavior. You know the swipey where it reveals the whole thing when you go back.

Chris: Okay.

Dave: There's one little piece that I kind of need to disable the animation. Safari saves a picture of it. Then if you three fingers swipe back - or whatever - it'll show the old state. Then it'll run the view animation. So, I need a way to disable that. But whatever. I'm splitting hairs there.

Chris: Disable the view transition or disable the three finger swipe?

Dave: Yeah, and there is a way to disable that view transition if user agent transition - or something like that. You can just bail. But there's not one for navigation auto. Anyway, there's one, like, if there's a UA transition, you can just say, "Okay, I'm not going to do my view transition." But there's no, if you're setting at view transition navigate auto - or whatever. There's no way to bail on that if there's a native or a browser UA transition or something.

Chris: Mm-hmm.

Dave: So, anyway, that's kind of like one niche thing I need.

00:08:52

Chris: I just fixed a Safari bug yesterday. There was a web-compatible kind of site where you log bug tickets.

Dave: Mm-hmm.

Chris: You've seen this thing. Yeah, it was kind of new to me. But it was like, "Look. This works in this browser but not this browser." Then you can chain people - or whatever. It was Jen Simmons sent it to me. There was a problem on CodePen. I was like, "I'm on it." You know? I don't want to be on that list.

Dave: Yeah.

Chris: But here's what it was. Iframes have attributes on them. There's allow attributes, and then there's an attribute called sandbox. They're kind of like the opposite of allow attributes. It's very restrictive. It makes an iframe very restrictive until you put values in that attribute that allow certain things.

One of them is called allow downloads. Sandbox equals allow downloads, and then if you don't... If you have just sandbox on there and you don't have allow downloads, you can't download any files. The iframe won't do it.

We didn't have allow downloads in there for Safari for some reason that I can't remember. But in our codebase, there's a file that maintains this list of attributes. It would probably offend some people. It says, "Safari attributes, Firefox attributes, Chrome attributes," and they're like arrays.

Dave: Mm-hmm.

Chris: And it's just the list of ones that those different browsers get, and we have to UA sniff to do it.

Dave: Yeah.

Chris: We use some Rails library or Gem or whatever that puts classes where they need to be - or something. In this case, we don't even need it on the front end. It's like a backend concern, so it figures out on that side.

You don't want to do that. That's not good. But there was just no choice.

If you put a sandbox attribute on that thing in a browser that doesn't support it, you get a big, nasty red console warning.

Dave: Hmm...

Chris: And when you're CodePen and you have lots of users and they're playing around in this console because they're developers, do you like seeing big, red errors from the website you're trying to work with? No. It's ridiculous.

It's a really strong error presentation, I think, for such a silly little one attribute is wrong. HTML tends to be a little more forgiving than that. But it's really noisy about it in all browsers.

I'm like, "Fine. I'll just put the attributes that you actually support on there so I can not send allow downloads to Safari and not see the big, red, stupid error."

Dave: Mm-hmm.

Chris: Even in the ticket it was like, "Oh, you should put allow presentation on there, too." I forget what it does. But I think it enables full-screen mode or something. Who knows?

I was like, "I'll try it." But then it's a choice you have to make. In stable Safari, it throws a big, red error (if you put it on there). But you can enable the flag. And if the flag is on, then it doesn't. You're like, "Well, I can't... I don't know what to do then." I can't tell if you have a flag on or not. That's purposefully not allowed on the Web.

00:11:55

Dave: Yeah. I understand how we got here, like with the whole allows, like the permissions array, I guess.

Chris: That's what it is. Yeah.

Dave: I feel like it's a Safari thing, not to poke at Safari, but you have to have the attribute combo correct. Have you ever tried to video plays inline loop?

Chris: Muted.

Dave: Muted. Allow plays.

Chris: Yeah.

Dave: You have to have these three or four attributes to make it cook, right? Or you can do image source MP4, I guess - or whatever.

Chris: Oh, is that well supported? Can you do that across all browsers (put an MP4 as an image source)?

Dave: I think you can. Yeah. Yeah.

Chris: Hmm...

Dave: I think I've done it before. But then that's only looping, right? But you can add an alt attribute. But then if you then do a video plays inline, then people can have controls to stop it, which is also kind of like an accessibility feature.

Chris: Yeah.

Dave: Anyway, I always get tripped up by those, you know, you need combo attributes to enable a feature that you're trying to enable. That's a tough one for me.

I'm going to just throw all of ARIA into this bucket, too. [Laughter] You need these combo attributes are difficult to deal with.

Chris: They are! And there's no RSS feed you can subscribe to that says, "Ooh... Next week, Chrome is dropping allow panda foo." You better allow panda foo on there.

Dave: Yeah.

Chris: You know how we know when you have to--? Obviously, I just made that up.

Dave: No, but viewport Haml UI.

Chris: It happens all the time. What we get is a support ticket from a user who is trying to do something and they can't figure out why and it turns out to be one of these allow attributes that has changed or added or something. There's just no--

Despite me keeping up with the Web almost professionally, there's no way to know when these attributes change. It's crazy.

Dave: How do you get a form to auto-complete with a credit card? You know what I mean? There are four or five attributes you've got to do. You've got to do a pattern. You've got to do an input type number. You've got to do an input type, and then you've got to do auto-complete CC whatever.

00:14:22

Chris: I love your... Dave is hating on attributes today. I hate attributes!

Dave: I hate attributes! I hate combos, man. I hate when you're like, "Oh, if they perfectly set 85 different attributes, it'll be the ideal experience." [Laughter]

Chris: Yeah.

Dave: "That's how we'll solve the problem." It's like, "No! That's not--" We need it more like CSS where it's like background pink, you know.

If you had to be like, "Background alpha channel 1-2-3, background green channel A-B-C," you know?

Chris: Mm-hmm.

Dave: If you had to set each individual channel every single time you wanted to render a color, you wouldn't do it. Everything would be black and white on the Web.

00:15:00

Chris: Interesting. Speaking of inputs, you know I was looking at Adam Argyle. I was trying to find a color mix demo where what I wanted was, on the left half you pick a color, on the right half you pick a color, and then it shows you the output of it. I was like, "Surely somebody has done this, right?"

I was looking around on CodePen to find it. Sure enough, I find it. It's Adam Argyle.

Dave: Mm-hmm.

Chris: But how he's designed it is the top left quadrant of the screen is a big color swatch. Guess what it is. It's an input type of color that's just absolutely positioned into the upper left quadrant of the page such that you don't have to set a background color or anything. You know how color inputs just are a color swatch?

Dave: Mm-hmm. Mm-hmm. Yeah.

Chris: It's kind of like that. Then the upper right is a color swatch, too, of input type color. You click on either of them, and it brings up the color picker. It's always different depending on what browser. In fact, I looked at this recently. It's absolutely surprising how different the color pickers are across browsers and platforms.

Dave: Mm-hmm.

Chris: But anyway, you pick a color on the left. You pick a color on the right. It color mixes them. And you see what's up in his little demo.

But I was like, "But I want to color mix different--" Color inputs are only hex. Did you know that? They're only hex colors.

Dave: Yes... And is it with or without the hash? It's without, right?

Chris: I think it's with.

Dave: With. Okay. All right.

Chris: It's with the hash in there. But they do not accept... You cannot set the value of it to, like, OKLCH.

Dave: Mm-hmm. Yeah, yeah.

Chris: Or whatever else - RGBA or whatever. It's hash only. So, you know.

These days, it's like chances of me working with a different color function are pretty high, and color mix works with those. So, I just started to get annoyed at color inputs. I'm like, "Why can't I set the value of a color input to a valid color string? Why is it so limited to hex only?"

Anyway, it's just a tiny, little crusade I was on. And why can't I see what the fricken' value is with my eyes? In no instances in any browser can you see what the value of it is.

You pick a color from the color picker. You see the color swatch. But you can't tell what the hex code that was chosen unless the color picker inside of it happens to display it, but most do not.

Dave: Mm-hmm.

Chris: I just thought that was annoying. I was like, "I want to fork Adam's Pen and make it accept color strings, too, of valid color formats." I was like, "How do I do this?" Or could I somehow augment a color input to display the value as well?

I'm like, "Ooh, baby Web component time," and HTML Web component. So, it just takes a color input. Then you just wrap it in a Web component. Then what my little take does is it just added an input type text right next to the input type color. They're just kept in sync with each other, so it's like eight lines of Web component.

Dave: Mm-hmm.

00:17:57

Chris: If you set one, it sets the other. You set the other, it sets one. But the question is, what if you put OKLCH in the text input? Can it somehow be smart enough to make that color swatch show that value? Well, how would you do that? Wouldn't you need a fancy color library like Color.js to convert it to something that's useful?

Dave: Color.js will convert it for you.

Chris: Yeah.

Dave: But you'd have to custom render, right? Yeah. You need your own swatch. You need your own button. You need your own everything.

Chris: Well, Eric Merchant in the D-d-d-d-discord solved it in a more interesting way. He just took the color string and just applied it to a custom property and then you get custom... There's some JavaScript function that just... You set it to some divs color and then you get computed style on it - or whatever.

You use the color function to force it into RGB. Then when you pluck the computed style off of it, it's a hex for some reason.

Dave: Hmm...

Chris: Now you have the hex. You can convert OKLCH to hex with just a couple of lines of JavaScript. It requires a little string manipulation and stuff but it worked pretty good. It totally ignores transparency, but that was its one downside. Otherwise, without loading Color.js and stuff, worked.

I added that to the little Web component and rock and roll. This wasn't for anything important. This was me just chasing... being nerd-sniped by a color input.

Dave: Mm-hmm.

Chris: But it is surprising to me. Why would you guess? Is that in the spec or something?

Dave: Um...

Dave: Why does it say, "Do not show the color value to the user"? [Laughter]

Dave: Yeah, it's weird. I don't know.

Chris: I don't know.

00:19:53

Dave: Yeah. I don't know. I mean, well, I think it's supposed to do the thing where it falls back to a text input if it doesn't. And so, maybe that would be a double? I don't know. That's weird. Yeah, it's weird it's not like the file input where it just adds in, like the pseudo element for the selector.

Maybe you said it. Did you use... think about using the new select menu or to build your own thing?

Chris: Uh... no. But I'm not sure what you mean, though. Like the one where you opt into the stylable one?

Dave: Yeah because... Well, I'm wondering.

Chris: I don't have a list of all colors or anything.

Dave: Yeah. A list of every color ever. [Laughter] I wonder if you could--

Chris: Did you see that website that lists all UUIDs? [Laughter]

Dave: Yeah, yeah, yeah. Perfect. I just wonder. Could you build a dropdown like that but instead of list items, you are presenting a color picker. Maybe that's a totally bad ARIA thing. [Laughter] I'm going to back out of my statement here. But you could have some predefined swatches or something and then a custom color picker or something. I don't know. Anyway, now I'm curious. I've built my own color picker here in my head and I need to stop.

Chris: [Laughter] It's okay. It was just a fun, little detour of life. Nerd-sniping, as we call it.

What else do we got?

00:21:30

Dave: On last week's show, we talked about AI - or whatever. I think I'd said... Somebody wrote in about, like, are we going to have left-wing and right-wing AI, right?

Chris: Oh, yeah. I remember that.

Dave: I think we said, "Yeah, probably." I mean you can look at what's happening at X and probably abstract the future pretty easily there, right?

While we were having that conversation, another conversation was breaking out in real life or on the blog-o-sphere. Miriam Suzanne had a good post on, like, AI and sort of the billionaire club and kind of how it's impacting her life. It's a really good thing about how "Tech continues to be political" (is the name of the post). It's worth a read.

I'm just kind of want to flavor this conversation we had with other conversations that were happening at the same time.

Chris: Yeah.

Chris: Robin Sloan had another article called "Is it okay?" and I saw this through Jeremy Keith, I think, actually. I'm subscribed to Robin but I hadn't read it.

Then Jeremy wrote, like, "Hey, I don't know. This seems okay." Then Baldur wrote... [Laughter] Baldur Bjarnason wrote a full-throated rebuttal of that post. It was basically like, "No, it's not okay." I think I've always said Baldur sort of takes the extreme opinion of what I feel down inside. [Laughter]

Chris: Yeah. Yeah. Love you, Baldur, but it is true. Yeah. Sorry.

Dave: Which is... you know. But Robin's thesis was sort of this, like, what if we could get, you know... What if chat bots solve cancer, the cure cancer? Well, isn't that a necessary risk - or whatever? Baldur's reply was knowledge tech that's subtly wrong is more dangerous than tech that's obviously wrong.

Sort of saying this idea of, like, the nondeterministic nature of LLMs maybe creates misunderstanding rather than scientific understanding. Does that make sense? And so, I think he said there is no path that can take current tech synthesis models and turn them into super-scientists. I think that was sort of the big pull quote.

Robin replied to that. Then it wasn't sort of... It made Baldur more mad. [Laughter] And Robin kind of... I want to say his ideas was, like, I think it is science fiction but it's just this idea of, like, maybe there's something that it can unlock - or something like that. Baldur was basically like, "No, you can't do anything. It's bad."

Acactio, Jeremy Keith, gets in the mix. Sides with Baldur here on this. And I thought it was a pretty good, like... Jeremy is pretty sober there.

Michelle Barker, I think, wrote a reply, too. We can put a link in the show notes.

But anyway, there's a lot going on with AI, what can it do, what's going on, and then specifically on this, like, are we going to have left- and right-wing things. I think I said I don't know how you would train an AI (with the corpus of the Internet) and, in a month since Trump became President, you would shift it entirely.

Well, there was an article that came out on TechCrunch where OpenAI tries to un-censor ChatGPT, which is like, "Oh, okay. What's going on here?" That being kind of the biggest model, most popular model.

They reprogrammed the safety mechanisms... or what would you say? They created new policy around the safety mechanisms of ChatGPT. And so, they kind of came up with this list of, like, "We're not going to push an opinion," or "We're going to present perspectives from any point." If you're like - I don't know - think of the most heinous thing you could think of and ChatGPT is like, "Yeah, man! That sounds great!" Other people might say... I don't know.

It's interesting. I read through this whole document, this whole 187-page sort of like how they're doing prompts and stuff like that because I was just really interested in just how you would make an agenda-less model but still has truth. So, I don't know. They're riding a very thin line. So, they basically have repealed mechanisms and they're riding a very thin line is sort of, I think, the answer there.

00:26:48

Chris: Yeah. It's one thing to just say what you're going to do. I see what you've linked to here. It says, like... It lists out bullet points of how we'd like it to behave.

Dave: Mm-hmm.

Chris: But they're also saying, "We changed some stuff." But I don't see a list of what was changed because I think that might be a little more scary - or something - isn't it?

What we're not seeing is a change log of, like, "Ah, we took the stuff off about racism. There was a thing that prevented it from being racist, and we deleted that." You know?

Dave: Mm-hmm.

Chris: Is that implied? We don't get to see that. It's one thing to say, "We want a model where there are good people on both sides." [Laughter]

Dave: Right. Right. Yeah, and that's kind of the vibe.

Chris: What actually changed?

Dave: Yeah. I think they kind of hint at it in these little blue textbox - or whatever - but just kind of like, "We want to approach such as scientific objectivity and deliberate discourse to inform our approach to neutrality," or something like that -- minimize editorial bias, which generally, when you read this, you're like, "Yeah, I don't want editorial bias in my AI.

Chris: Yeah, sure.

Dave: But at the same time, it's sort of like, are we comparing? If you're like, "Earth is flat," which they have as an example, Earth is flat and Earth is round, I think there is an answer. I feel like it's the same with, like, do trans people or people of color have rights? There's an answer there that's pretty obviously a yes.

Anyway, they sort of said avoid factual reasoning and formatting errors. But if you're trying to be factual, sometimes the facts bear out a political opinion. So, how do you navigate around that?

Dave: This stuff is so complicated. Just not an expert. A little hard to weigh in. I tend to just be on the side of safety and responsibility - whatever that tends to mean. And we already know that that's not the jam.

You started the company--I'm talking about OpenAI--as, like, "Let's do this nonprofit. Let's do this for the good of the world," and it took two seconds (after some success) to be like, "Let's bring this baby in-house! Let's have a $200 plan. Let's tell people it's worth $1,000. Let's make most of what we talk about crazy science fiction stuff that's very business-oriented, not like what's good for the world."

I see what you're saying, and I'm not even against all of it necessarily. But I just see where you're going. The point of this is extracting profit from words on the internet. That's the point.

Dave: Yeah.

Chris: You're going to change policies to help that goal because that's all you talk about? Duh. [Laughter]

00:30:00

Dave: Yeah. Yeah. Well, yeah. Or you're going for a broader appeal over sort of like safety stuff - or who knows. It's kind of... Anyway, I can put links in the show notes. But it's worth reading.

Chris: It was a pretty hot blogging week, wasn't it, with all the back and forth?

Dave: Yeah.

Chris: Very nice.

Dave: Back and forth blogs about blogs. I think that's really good.

Chris: It is, yeah, and I still see--I hate to say--both sides of this--again, because it's such a loaded term these days. But I can feel how Robin thinks about this, Robin Sloan. He's like, "Look, this stuff is absolutely wild already."

Look at what these things can do. It's the most science fiction thing ever that's just happening right in front of our eyes. Unbelievable what they can do.

There has been a rate of growth that have made these things better and better and better and better. Unbelievable. How can you just look at all that and be like, "There's no more science fiction left. Where this is going to go, we've squeezed out the science fiction. You will no longer be amazed after this"? That's not going to happen.

To hear him talking about it, especially as an author who writes science fiction books for a living, to be like, "I can see a world where this is headed towards a super-science situation," just feels very, like, "Of course, I understand his perspective on that."

Dave: Mm-hmm.

Chris: It has a ring of, like, "Yeah, probably," kind of thing to it while everybody is, like, ringing the safety bells is even more pertinent.

00:31:37

Dave: Yeah. yeah. I think there are probably pessimists and cautious optimists and then sort of I think there are also people who are just [laughter] full evangelicals about it, like, "It's going to solve every problem."

I can't escape it, man. I went to pizza, and I had a conversation about AI with a lawyer who is making his own apps using ChatGPT.

Chris: Yeah. Yeah.

Dave: I can't say it's a bad thing. I can't say... You know at a certain point I think the software is going to let you down. But it is cool. [Laughter]

I think it's going to code itself into a corner, and you're not going to be able to fix it. Maybe the software can. But for creating little apps or widgets or automating little processes they do, maybe that works. Maybe it works for them.

I don't think he's on the, like, "I'm just going to AI lawyer instead of do actual lawyer." I don't think he's on that train. But I think he's... Maybe it's like, "Summarize this bunch of text - or whatever - so I don't have to and let me know if there's anything in here that's related to X, Y, Z," or something like that. Maybe that's useful. I don't know.

Chris: I don't know. You know I spent a decent amount of time being mad about ain't nobody asked me slurping up the entire Internet and ignoring copyright. All of this stuff still has not really changed or seen any differences on that. But at the same time, I don't know. I'm not ready to hang up the towel on technology yet. I can't. It's the only skill I have. I need to be in this industry for a while longer.

Some part of it is kind of fun, especially because I feel like I come at tech from a little bit of this, like, I like the tech and I like the UI and UX around tech, and it happens to be kind of like a poppin' off place to be at the moment of, like, how can we design around these things. Not that I'm really participating that much but I'm certainly using it. To be like, "I'm in my VS Code, and then I'm trying Cursor, and I'm trying Windsurf, and I'm trying Trae, and I'm trying Augment," and I'm trying all this stuff.

It's like, "Ooh... How do they do it?" "Ooh... What are their key commands?" "Oh, that one pops up inside the editor?" "Ooh... This one pops up when I select text. That's a nice take on it."

There are a lot of UI/UX stuff that's fun to watch as part of it. Not to mention, I've been doing a bunch of coding lately that feels like it's actually helping me where I'm having a problem.

I'm in a Vite config, but it's one that I don't need to know intimately. It's just a side project. I'm like, "Can you just update this Vite config to proxy at this URL and put in the hot module reloading there?" It's like, "No problem. Here you go."

Dave: Mm-hmm.

Chris: I'm like, "Sick. Thanks." You know? That saved me so many minutes, man. So, to do that, to just be like... To worry about whether that thing that helped me, you know, what its political position is, I just can't right now. [Laughter]

Dave: Yeah. Yeah.

Chris: I don't want anybody to get hurt. I want safety bells to be there. But sometimes, in the tech-specific, coding-specific stuff, I'm like, I don't know. I'm a part of it. Call me the devil.

00:35:13

Dave: Well, I think it's just a big thing, and there's been a lot of VC. It's the only thing, man, [laughter] that is exciting tech. It's the only heartbeat that exists, right? And so, I feel like people are just doing this, adding it to get funding, so it's in everything, so you kind of have to talk about it.

I think there is some utility to Copilot saying, "Hey, here are some... Just hit tab, and what you were kind of already starting to write auto-completes," just like on my phone when I send a text message.

Chris: Yeah. In a way, it kind of helps you think. I was listening... I still like that show Search Engine with PJ Vogt. You should listen to it. It's good stuff.

Dave: Mm-hmm.

Chris: He had an AI one on the other day. [Laughter] It was like a story that's old news, I feel like. But that's just how he rolls sometimes. It was kind of about the homework story of, like, "Now that we have ChatGPT, is your English homework dead?" - kind of story. [Laughter]

Dave: [Laughter]

Chris: It's not behind, but it's like, "Yeah, that's still there." I'm sure high schools are still very much struggling with that, or all levels of education, really.

But there was a moment. Some dude, I think he was McSweeney's dude or something, has a book about writing in the age of AI. He just has this simple point that I think is strong that, like, when you let a machine write for you that you aren't benefiting from the regular process of writing, and writing is a struggle-y, difficult, hard thing to do. But guess what it does for you. It helps you fricken' think.

Dave: Mm-hmm.

Chris: Duh! It doesn't have to be a book that you're writing. It could just be on a piece of paper. It could be your diary. It could be anything. I just mean writing very generically. If you start leaning on tools that write for you too heavily, you really are robbing yourself of thinking and, crucially--he always pairs them together--thinking and feeling. How does it feel to write these words? Are you representing yourself in that way?

Then probably, ultimately, you're communicating something as well. So, that thinking, feeling, and communicating is such a big deal. Kids, don't generate too much text! You need to use these tools differently.

But I feel like sometimes from the coding aspect, it's almost helping me think because I want to think through this code. I'm thinking about it right now. It's all up in my brain. But sometimes when the AI can just help auto-complete and speed up that part so I don't have to stop thinking and go look up syntax or whatever, or I'm in TypeScript and I can just hover over something and see what parameters it wants and what types they are, it's helping me. It's actually helping me think, not robbing me of thinking.

00:38:09

Dave: And that's my hope, too; these become tools for thinking not tools for outsourcing thinking. That's where I've been in conversations where people are just kind of like, "I've thought through this. Am I missing anything? I'll just ask ChatGPT, you know, ask Chat what it says." Like, "What's another word for blah-blah-blah?" Kind of like thesaurus.com style stuff, and that seems... other than the enormous sort of ethical considerations there, that seems like a pretty naïve fine use of an LLM. It's good at sort of modeling language, right?

Chris: Right, and I can see disagreements happening there, too, is that that is the absolute nature of them. There is no technological innovation that's happening yet that isn't anything beyond it's trained on a bunch of words that already exist and it gets better at piecing those words together based on that knowledge. It's not inventing new knowledge. It cannot do that.

Dave: Yeah. Yeah.

Chris: It can make interesting connections. It certainly can help you think. It can help you research. It can help you reformat things. It can summarize things. It can expand on things -- yadda-yadda-yadda -- do all that stuff, but it cannot do new science... probably. Right? Unless there is some world in which it reads all existing science and then can just somehow see gaps in it - or something - or that a scientific breakthrough comes in a gap instead of an epiphany.

Dave: I don't want to go toe-to-toe with Baldur in a blog post.

[Laughter]