Scott & Mark Learn To...Have Taste

So when is taste an opinion where it's just like, well, Mark's fancy and he has an opinion and we should all listen to Mark versus taste? Like You know, you'll you'll see a movie and someone will say, I didn't think it was that good and it won more Oscars than any movie in history and we saw it at your house and I'm arguing with my wife, like she didn't like it. And it's like

Talking about. I know. We don't need to talk about the movie, but how can you not like this movie? She's like, it wasn't my thing. So when does art or taste become Subjective versus objective, like you know, Mona Lisa, eh, David, eh. Yeah. Oh man, you just opened a giant can, didn't you? Let's let's do it. Let's go.

And which is a very fortunate thing when they want reviews of things. Sometimes they want reviews of code, sometimes they want reviews of architecture for their systems. And in the doing of that, Presumably they we have a mountain some amount of taste. I've had a couple of teams have regular weeklies with me where we go through the product and we make sure that it works the way it should work because they have said very kindly with the compliment, you have good taste. I get good.

Taste. Let me separate uh something out apart here because I think when you taught you you mentioned the Mona Lisa or a movie, there's Button placement in a Criticism versus I view that as a little bit different than product. But because you're there is uh personal taste there. Like Mona like in the movie is her personal taste. It's not her judgment of it as an artistic Work. Right. But but the whole planet loves the movie or the art or the thing. Like is there can someone just be contrarian?

Well people have their own like some people don't like action movies, a lot some people don't like horror movies. Some people have different sense of humor. Scandalous. You know, uh I d I d and even when it comes to how the app should behave, how the UX should behave, when you click on something, what should happen? There is some personal preference there. When I'm making decisions for things on Sys Inturls tools.

Mm-hmm. It's my personal preference. It just happens that a lot of people like those personal preferences I have because I base them on intuitiveness and ease of use. But ultimately you wrote them for yourself. And now everyone uses them. Like when you wrote the video trimmer, I gave you crap about some of the UI stuff. And you're like, no, I'm doing it like this.

Yeah. This for like let's let's take Zoom it and the just to continue on that one because I think that one's a great example. When I made Zoom it, I I told you I was frustrated with an existing tool that came bundled with a laptop. for zooming the screen and it was very clunky. The zooming was extremely clunky and drawing was extremely clunky. And I'm just I wanna make these natural inline gestures that are make make it look seamless.

And so that's where I came up with the interface for Zoom it is to try to make everything just keyboard driven, simple, easy to remember, easy to type. And that's been the whole philosophy around Zoom its functionality. And despite the fact that it's now overloaded and become a platform on to its own. But It is the everything app.

Yeah, that's in the the a feature branch right now. Yeah, of course. But specifically just to even continue with that a little more, we did that preceptor experiment, the small scale experiment where we had a group of people, developers at Microsoft, go implement the Zoom it video tripper feature. Video trimmer, exactly, right. Just by itself, right? And yeah. And a lot of them are like, here it is, it's done. And Yeah, you did it without a visual spec. You just said do it like that.

And then you and I looked at them and are like, wow, this is ugly, hard to use stuff. There was some good solid effort. Yeah. You're being very Very generous. I do too, but I think um that highlights a difference in taste that you were talking about. Mm-hmm. Like universally I think people would find those those interfaces that we saw clunky and hard to use.

Yeah. And there was also yeah, there was multiple levels to the taste. There was the UX exp the actual user experience. And then there was the functionality and like expectations of what would happen when I clicked on this or I dragged on that. So the the the holistic view on it was this is clunky at best. most of them And so why is that? Is it Why am I able to in y yeah, you looked at mine are like it was mostly intuitive, right? And it's like what you expect out of a video trimmer.

So I presume it's because you've used so many video trimmers, you've been exposed to so many things that you have an experience on it. You've used computers and you get where things should like feel right. It's that I think also I'm like, what would I expect if I moved this or what would I want to be able to do? I want to be able to move that here and then that has this effect. But you've used video trimmers before. Did you inform that based on experience trimming videos?

somewhat, yeah, because Because I think that the biggest problem is that people don't see the big picture. I always use the term 20,000 foot view. I've been using voice access on Windows. You know how I do all my typing with voice, but I've also used voice access, which is a feature in Windows where you can actually tell it. Page up, page down. So I was reviewing your document today in the big open source meeting and it's like 70 pages, right?

So I'm sitting here and I'm I'm scrolling and I'm like, I'm gonna use voice access. So I get my my kettlebell and I stand in the middle of the room and I go page down, page down, page down, and I'm reading this document. And then I realized that the bar takes up the entire top part of Windows. Yeah. Which means that when I throw the mouse into the upper right corner looking for the close button. It is no longer the close button for the window. It is taken over.

By the voice access thing. And that seems like a small thing until you realize I've been throwing the mouse in the corner for thirty years. Are you saying the voice access person didn't have taste? Is that what I What I'm proposing is that that person thought about making voice access the best, which is different than making Windows the best. And their muscle memory, which I caught

Immediately, within two minutes, I threw my mouse in the upper right corner and the close button wasn't there. I threw my mouse in the upper left corner, the system menu wasn't there, and now I'm frustrated. They just didn't see that because they've they're focused on their thing. Yeah. Um Yeah, I think I actually uh hyper focused without looking at the big picture. We've talked about that before. What does it take to be a good architect? Or a system synchrone. And I would argue it's exposure.

Yeah. Yeah. I you know, we talk about pre the preceptorship and the context that it has with nursing and things like that, and there's always in an ER show or, you know, any T V show like the Good Doctor Are you gonna give me your throat tracheomy? No, no, no, no. Somebody does a thing and they go, Oh my goodness, how did you do that? Oh, I saw it once thirty years ago or I did it once in the medical field. You're referring to the last episode of The Pit.

I didn't I haven't seen them all. I'm three behind. All right. I'm like three hours behind. Because that actually happens. Are you serious? Did they do it? Yeah. Well not not the pen thing, but they but the like I saw I did I saw it before. I well I saw it once thirty years ago and here I go. Yeah. See that? Oh, was that the part where the ladies, they were like, oh my goodness, that was amazing. And she's like, Oh yeah, have you done that before? She's like, No, but I saw it once.

Oh, that was good. Yeah, like so much of computing is just having seen stuff and then filed it away. So then either people aren't seeing enough stuff, which is an exposure problem, or they're not filing it away. And being able to pull it pull it back from from the arc. I think another way of n you know, the term that we'd use is experience. Yeah, I guess it's just experience just hanging around and paying attention for years.

Hard to acce you know, I guess you can accelerate that with like, you know, intense exposure, directed uh methodical exposure. You can accelerate experience. To some extent.

My wife has been a computer user for her whole life. We are the same age within months. You've met my wife. She is a smart person. Mm-hmm. She's working on her PhD. Couple of times a week, she loses files. So either that's the browser Windows Explorer OneDrive's fault. Or it's her fault. Now I'm I know my wife, so I'm gonna propose that it's a user interface problem. Yeah. I never lose the file. So that's because I was there while the water was getting hot.

And she got thrown into like this hot pot because for ten years she was not doing anything in office work. And now she's come back and she's like, I don't know where the file is. Where is the file? All right, well, it's in the winter. I don't know if you do the, you know, save to OneDrive and it's just like it picks a random place and it's really.

Scanned I scanned I used the OneDrive app on my phone and I scanned my like hotel thing and I was like, upload I'm look look at me, I'm very clever. I'm scanning my PDF of my thing and like and then I'm like, I don't know where it went. It's in the root of the one drive. It didn't even pick the last place I put stuff. Like the only thing I ever use it for is scanning my parking thing for the airport and my expenses. That PDF is gone. Yeah. Now I'm looking for it.

Okay, so good taste. How how accelerated can it get though? Because I was talking with this university recently and they wanted to know how quickly you could make a senior engineer. Well, this is where I think that you can't you know a boot camp, let's go on one extreme. It's a boot camp. Are you gonna have a senior engineer come out of a boot camp? Absolutely. Have a super soldier come out of boot camp. No. You need a lots of sustained repeated exposure to the to variants of the same thing.

Right. And this is the mythical man month right there. Right. You know, nine people can't get a black belt in one year. It takes years. You have to just you need more sleep sleep. Yeah. Explain this to my 20-year-old uh who is doing his uh anatomy and physiology final this morning, and he wanted to know how he could learn more quickly. I okay, when's your final? And he's like, today. Okay.

And I was like, Well, let's talk about space repetition and how you should have thought about that six weeks ago. So now he knows about space repetition. But he wanted to know if there was some magic thing that he could be done in the next six hours. And it's like, no, but there isn't.

Yeah. So going to this, how do you even create a senior engineer question, which uh we're r we're grappling with when a senior engineer is more than just prompting AI. Exactly. Because you don't without having the taste and experience, you don't know what you should be prompting or how you should be evaluating. I met with somebody last week, good friend, longtime friend, and we were having this debate over AI coding.

And he's like, People aren't gonna need to look at the code. You know, AI's gonna get good enough. And I'm like, I just don't believe that. And we talked about this in a previous episode where Mm. I micromanage because I'm sculpting, and there's no way that I could fully specify something and have any expectation that AI is going to come out with everything that I was thinking implicitly or would want.

If you fully specified it, you would have used keywords in a programming language and then you would just wrote the thing yourself. Exactly right. Yeah. So um I I push it here and there and, you know, guide it and I know where I want to go, but I don't know exactly how I'm gonna get there. The zoom it video trimmer was a great example. I didn't know exactly how I wanted the thing to look and behave until I started.

you know, made one, had had the AI make one for me and then go, Nope, I don't like that. I don't like the buttons there. I don't like the buttons that size. I want this here. By the way, I want it so that when you drag this it releases at the certain point and all of that was me playing with it. And we talked to somebody, you know, quite famous in the industry, a legend, that said, I get my boost just by fully specifying everything.

You know, you talked about the fact that if you there's no way to fully specify everything unless you're talking about make a compiler. Yeah. And that AI is gonna fill in the blanks with who knows what. Yes. And back to my thing about how I I do all my dictation. I think I'm getting a boost by doing dictation because I have more words, more context, more input, more specificity. And I think early in career people are getting less good experiences.

lesser experiences because they're leaving so many things up to the the dice roll. Yeah.

Yeah. Specificity matters. So then why am I going to school? Why do I go to college then for four years? Is that just four years to bake? And s get exposure and learn how to I think it is well, the breadth of knowledge that you cover there, which is grounding. Like if let's take a software engineering as a profession and why does computer science degree matter?

Yeah. I strongly believe computer science degrees matter. Not to say that people can't self teach and be good software engineers, but And and it also depends,'cause this person I was talking about last week, we've I finally Coaxed and prodded him. And he doesn't have that much experience doing I assisted coding. He has some, but he's he the stuff he's built is like, you know, web plus database.

Uh and I'm like, okay, so'cause he's like, what you're gonna want are people with philosophy degrees. They're general, they're critical thinkers, that you don't need software engineers or people that understand computer science to be software programmers. And I'm like Well, that might be true for web interface on top of a database. Because that's a very simple app, you know, right down the main line. But anything more complicated than that, and then you start to run into

Lots of complications. Uh this ref reference checker program I told you about, there's fifty million I mean, not fifty million, I'm exaggerating, but there's there's a dozen little questions about how it should behave. Dozens. When there's an error, how does it show you the error? When should it show you er there under this condition? Um should it override it with a warning? Should it, you know, like I had to work with AI through that.

And I didn't know exactly what I wanted until I'm like, no, that's not that presentation of information isn't telling me what I want to know. Yep, change it. That part there, there's a million different combinatory all the combinatorics. Separate from AI augmented programming, if you leave it unsaid, it is unsaid.

And it is going to be filled by whatever the uh default part of the switch statement is doing. You know what I mean? Like whatever it is is gonna just happen. I don't know, I didn't specify. Sorry. And that is amplified by AI. Being nonspecific, being unclear in the ambiguity loop is asking for unexpected things.

And you know, I g I go back to when I was at Carnegie Mellon in the eighties and they formed the Software Engineering Institute and people were like, We're gonna make um a science out of programming. And really codify what it is. And and that's failed. It's never succeeded at saying what is exactly software engineering and coding. Because it the state space for it is Unbelievably large. It's not like, you know, let me make a beam like civil engineering, let me make a beam for a building.

Software is way more complicated. And we start with levels of abstraction, like uh upload this file, OCR it. present it, index it, and then offer give me the ability to save it to disk. That sounds really simple. But underneath that, there are dozens and dozens of corner cases. and architecture decisions and user interface taste decisions. Which makes me wonder though, why everything is working. And I think that's because for simple websites like your friend is making.

The statistical mean mostly works. Yeah. And for edge case stuff like your trimmer or your uh panorama thing where there's like not a lot of prior art. you're getting uh am almost an amplification on the edge cases of just like it either works if it's very narrow or it's just wacky and doing weird stuff on the outer on the other edges. 'Cause when you've some of the stuff you've bumped into is bigger than edge cases. Yeah.

It's just it doesn't have any idea what you're trying to do. And you gotta be laser specific, which means then now waiting for it to compile. Is you waiting for the prompt to do it? That's the new waiting. Half my time is just waiting for the thing to turn, which is starting to become frustrating.

It it is really and it's starting to. It's been frustrating. But have you seen this one? I've no I've seen this one just in the last few days, uh, where I tell it, hey This is not behaving right. Fix it. And then I'm watching the chain of thought as it's working through it and it's like, well, this doesn't fix right. But you know what? While I'm looking at this, I also see this other thing that should be changed. And I'm like Stop. Like I didn't tell you to do anything else. Yep, exactly.

Yeah. I keep coming back to my kids and how old they are and we just came back from uh from Howard with the eighteen year old who is uh doing a uh admitted students' day and figuring out uh you know how he's gonna go and I'm like, what's this kid gonna learn in the next four years? And is this school ready and like You know how I've mentioned to you before that we've got this kind of wave of broken COVID kids kind of going through the system?

I'm wondering what schools are gonna be ready to explain thinking. in the world of AI, uh, versus schools that are just gonna not even talk about it for four years and they're gonna pop people out that are gonna have to sit down in twenty thirty And learn how to think like this. And if they're gonna pull us out of retirement to explain it to them. Yeah. Well I think there's gonna be a lot of schools turning on chat GPT degrees.

Yeah, I've talked I've said this before, I'll say it again. Someone right now is vibing a BA in English. Yeah. If not a full BS in computer science. And that is gonna bite them in the ass. Yeah. Because we said this to our kids early on. You cheat, you're only cheating yourself. Imagine that for calling. Oh my god. So bad.

I hope that we can make I mean people have been reading the paper. Like I've been having great meetings. I'm meeting another I'm gonna meet with some other folks th later this week. I've talked to like five different universities. People get it, but no one's moving fast enough to uh To catch this. Yeah. We're trying to get people to move faster, but

Oh man. Yeah. We'll we'll have a meeting after this podcast and I'll tell you all about it because it was a really good meeting. Um, okay. So let me ask you this in closing then, for this episode, is there value in a degree because there's a lot of really interesting paperwork that's just saying it's so expensive in the States. Yeah. Why bother? I absolutely think there's value in a degree. And I'll tell you why. First of all, what does a gre degree represent? It's structured Yeah.

It represents a gauntlet that you ran through. And um where there's a certain certain expectations around the base level of and knowledge that you got out of it. And you can you might specialize in something and that tells Whoever is looking at your degree that you did some extra training in some area and and knowledge acquisition. And so it's shorthand for evaluating somebody. It's a shortcut. Yeah, that's that's fair. I mean they look at your passport and they go, Yeah, that's that's that guy.

Yeah. Great. Well you ran that gauntlet. I know that's a rigorous gauntlet. I know that they've got high standards and comprehensive training with instructors that know what they're doing. That's why the top schools have that reputation. Yeah. Yeah. But that's the thing though, right? Like if you can get a top twenty school, great, but then there's the

twenty two hundred schools that I went to, but I took eleven years to finish my four year degree and I didn't even get out of school until two thousand and three. But I did it for myself. It took eleven years, but I wanted it done because it was like I don't want to Quit now. I'm so close. I kept telling myself I was so close for like 11 years. Why did you get a PhD? You could have just uh

Finish uh well two things. My br my father brainwashed me. He was an M D and his he was like get as much w pick something you like. It doesn't have to be medicine and get as much Education on it as you can. And don't go into workforce until you've done that, because otherwise you're likely to never come back.

That's pretty good advice. Actually, I mean if you got the if you like if you get the privilege to like go to a masters and then go to a PhD and they they they're gonna keep helping you do it, then yeah,'cause you're never gonna come back, right?'Cause like I'm gonna retire and then go back to school and learn Spanish. Too late. I could have used that twenty years ago.

Yeah. So that that that and and also I just wanted to learn as much as I could about computers. So that was Well, I mean hopefully it'll work out for you one day. Ha ha ha.

Uh if you made it this far into this episode, I'm interested in if you think chat that uh there's value in a degree because I know that a lot of you who are watching it who maybe work at Microsoft with us, uh, you may have gone to a top school like Mark, you may have gone to a community college like me, or you may have just done it on your own with uh, you know, a a notepad. Uh one of the best programmers I've ever met has a uh BS and GR. Yeah.

Which means he's got the structured thinking. He just didn't do a computer science. I just had a uh dinner with Jeffrey Snover, who's retired but now w working as a Harvard fellow. Uh but uh he never gr got a degree. How is Jeffrey? He's doing great Yeah, of course. Lovely.

That's good. I have to text him and say hi. He's such a fun guy. Yeah, he's a spicy one. I love that guy. Okay. Well, that's a show because we want to be responsible and not just yap for for for an hour and a half. But if you made it this far You are in the top one percent of humans on the planet. You are better than everyone else. Yeah.

So comment in the chat that you are better than everyone else. And again, it cannot be overstated. Share the show, tell people about the show, because so much of our fragile egos uh is dependent on engagement and views. And uh we'd love it if one of these went viral someday. Way yeah. It's speaking into the void, you know, is That's I guess I'm the void in that context, but Yeah you are. I think people Four or five people enjoy the show deeply. Yeah. I'm looking for it the sixth.

Well it's not smartless. Yeah. It's not it's not Conan O'Brien needs a friend. Come on. We could have a guest on though. Could you bring someone famous? Did we get Conan? No, like I don't know, Bill Gates or somebody. Like call a neighbor, talk to get Linus on the on the horn. Yeah. Actually, if you think we should have guests on the show. We should try it.

Should try it. Producer Rob can open another Windows window. We could change the whole UI just into like GNOME or Linux just for Linus to come. That way he wouldn't be offended by putting him in a Windows window. Throwing it out there. You know people. Adiós. You know more. Yeah.

He could come on the show. Okay. We'll think about that. Maybe we could do two things. Here's an idea. If you made it this far, chat, you're even in the top 0.1% of people. What do you think about guests occasionally once a month? And a live stream with QA. Scott and Mark learn to ask me anything.

This is part of we we had a brainstorming session a couple of days ago with producer Rob Ark to talk about how to get uh dozens more people watching the show. Yeah. And those were two of the big suggestions. All right. Let's give it a try. Let's go.