Compile After Class Crossover Episode  🔗 **Source:** [Compile After Class Crossover Episode](https://share.snipd.com/episode/9bd7ad82-6058-4c9d-9a8b-fad50ab819c7) ## Highlights - **Be The Pleasant Person Who Moves The Group** - Being the smartest person in the room is less valuable than being pleasant and collaborative. - Scott emphasizes using Transcript: Scott Hanselman I thought that you had to be the smartest person in the room and that you had to like totally grind it out and be the one that has the answers. And I realized that life and as a general rule, is a big group project. And just like, you know, group projects, because y'all are high school students, just like with group projects, there's always like one rando who just shows up for the group and gets The grade, but they didn't do any work. And he's like, dude, that guy didn't do anything. He got to do this is screwed up. You know what I mean? That happens in life as well, but you have to be able to get along with human beings. So being pleasant and easy to get along with is way, way, way more useful of a skill than being the smartest person in the room. ([Time 0:03:10](https://share.snipd.com/snip/ae8bea51-ccac-437d-82ed-d571d6c769d6)) - **Learn How To Learn And Vet Sources** - Learn how to learn, not just memorize curriculum; build a critical BS detector for online claims. - Scott warns against trusting influencers and urges independent research beyond Reddit posts. Transcript: Scott Hanselman When you started is primarily teaching you stuff and it's kind of like this daily lecture from a bunch of old people and then at some point, probably where you guys are like 16, you start Being like, I get it, dude. Like I get it now. And then school becomes less where you think you get it. You don't, you have no idea what you're doing. You, you get less about learning stuff and more about learning how to learn. So what I would tell you as, you know, people that are gonna be juniors and seniors in high school is, yeah, absorb all the stuff they're teaching you, but focus on, can you learn stuff On your own? And, you know, we see a lot of misinformation. You're growing up in a time where politics and health and influencers, any asshole, excuse my French for swearing on your podcast, but any asshole with a microphone can just say some Stuff. And it's like, oh oh, look, look, he's got a fancy room and a bunch of LEDs. He's old. He must be smart. That is not authority. So you as people who are learning how to learn need to be able to say, I don't believe that guy. He doesn't know what he's just a guy. This is a guy with a podcast mic. A podcast mic does not make you the authority. So learning how to learn means you need to have a good BS detector where someone says something on a podcast, whether it's about tech or about AI or about crypto or about politics. And you go, my critical thinking hat doesn't, says this is nonsense. Don't go down the wrong side of the internet. Half of the internet is trash. So you need to be learning in the next two years and in college how to teach yourself and actually do your research. Because going on Reddit is not doing your own research. Oh, I did my own research. No, you didn't. ([Time 0:05:18](https://share.snipd.com/snip/1f248380-5b51-4493-91d6-7f401b15b493)) - **Stop When Your Body Says Stop** - Listen to your body and stop when you're making things worse; take breaks and get sleep, water, and sun. - Scott used a jittery scrolling bug he worsened by grinding as an example to walk away and recharge. Transcript: Scott Hanselman I need to stop or I need to push through this to control your brain? That's a really good question. I have two young sons. I have an 18 year old and a 20 year old and getting them to listen to their bodies. What is going on in my head right now? Why do I feel this way? Why am I angry? Why am I frustrated? Why did I snap at mom? Why do I feel like crap? Oh, I got up at one in the afternoon. I haven't had any water in three days. I ate a Chick-fil yesterday and I feel like crap. It's probably your fault, dad. You know what I mean? That's how you do that. But if you can listen to your body and you go, man, I haven't drank any water in three days and I haven't had any good protein and I haven't seen the sun in a week. Listen to your body. That applies to if you're working on a problem. Literally yesterday, I was working with co- I was getting a bunch of weird layout issues in a Windows app. The scroll was jittery. I was scrolling with the mouse wheel and it was jittering. And I was like, and I was grinding through it, but I was making it worse. And I listened to my body. It's like, I'm done for the day. ([Time 0:07:20](https://share.snipd.com/snip/6d34a3b9-9982-4f05-891b-3c367a21db0c)) - **Ship Podcasts Without Editing Yourself To Death** - When starting a podcast, avoid editing burnout: remove ums, normalize levels, and hire an editor if possible. - Scott recommends paying a student or music-department friend instead of spending six hours per episode editing. Transcript: Scott Hanselman Production part like like we'll probably do this podcast and it's like what is it today it's the 22nd of july is this coming out next week or is it coming out like after school like dude You're gonna spend hours editing this yeah so i would probably say maybe don't i know that you're kids and you have no money but like could you hire someone on fiverr? Could you like pay a local kid? Could you find a kid who works in the music department, right? Like Cam's got music interest. He's got guitars. Somebody out there needs to edit this. So when you start a podcast, don't just yap for 30 minutes. And then in six weeks, I'm like, email Sri. I'm like, bro, where's the show? Oh, life got busy. Get it done. Do the thing. And you don't have to do it perfect. There doesn't have to be an A plus podcast where you talk for 30 minutes and then edit for six hours. Just remove the ums, level the audio, make sure that we're all at the same level. ([Time 0:09:47](https://share.snipd.com/snip/5d005ddf-dfa0-4fbd-9a8b-58a5d5a40a02)) - **Do The Thing You Say You're Going To Do** - Do what you say you'll do and be consistent; small tasks done now prevent long disputes later. - Scott frames punctuality and paperwork (notary example) as adult responsibilities that determine outcomes. Transcript: Scott Hanselman You say you're gonna and i'm interested by the way i want to hear from you guys after your parents listen to the show. And if they disagree or agree, like, yeah, they're going to give you parent versions of what I say. But doing what you say you're going to do. Like, I trust a tree is going to edit this or you're going to edit this and it's going to get published. It's sad. I have had shows where I'll come on with a bunch of folks like you and we'll do a show and the show will never come out because they just, they didn't produce the show. I'm watching on my TikTok. There's two ladies right now that are going viral on TikTok because they lost their jobs. One lost their job because they got the job and on day one and day two, they didn't show up on time. They were three minutes late and then five minutes late and they're fighting for their lives in the comments, but the facts are they didn't show up for work. So do the thing. Consistence. Consistency. ([Time 0:12:43](https://share.snipd.com/snip/6f268ddb-a125-4696-901f-028cb36af849)) - **Building A Diabetes App From Personal Experience** - Building Glucopilot as a diabetic gave Scott product intuition outsiders would miss. - He later sold the app, felt ripped off for $25,000, but it helped him move out of his parents' house. Transcript: Scott Hanselman That's a great point. So that's a good topic where we talk about diverse teams. If 10 people get together to write a diabetes application and 10 people do not have diabetes, I would argue that that's probably not a diverse team because there's no one to work with, Right? Like if I'm working on an app for veterans and no one on the team is a veteran, how can I truly understand what a veteran needs? So as a diabetic writing an app for myself, I fully and completely understood the topic. And then I later sold it and worked with a healthcare company to make it better. So if you're making an app for an audience, make sure that someone on your team is that audience so that they understand. I wouldn't make an app for teens if I wasn't a teen. I would get a teen as an advisor or a board member or a programmer. So I don't know what specifically I did, like specific features that were better, but there was a ton of stuff that is intuitive to me that would not be intuitive to you unless you were Also diabetic. ([Time 0:19:12](https://share.snipd.com/snip/258b84c6-9690-4e67-81de-ae2a1d246608)) - **Prompting Is Not The Same As Programming** - Prompting is inherently ambiguous and not equivalent to deterministic programming. - Scott contrasts nondeterministic LLM outputs with compiled code's repeatability and calls Transcript: Scott Hanselman That's a great question okay so first i'm gonna be a little spicy Prompt engineering is not a thing. Okay. So you and Cam could both prompt the exact same prompt. You will not get the exact same code. That by its definition means that the prompt was ambiguous. The prompt is not a programming language. But if you and Cam each wrote a for loop in C and it compiled and it was the same for loop, it would run deterministically. That for loop is guaranteed every time until the end of time to work. So then let's say we had a million trees and cams all doing the exact same prompt. How many, so a million people all prompting the exact same prompt. How many unique results are there? Are there a million? Are there a hundred thousand? Like what is the ratio of prompt to solution? How many potential solutions are there? There's an old programmer's joke. You may have heard me tell, but I'll tell you anyway, because it's useful. I need you to go to the store. I want you to get me bread. And if they have eggs, get 12. So I tell that to Cam. I tell that to Sri. Cam comes back with 12 loaves of bread. And I go, Cam, what the heck? And he's like, they had eggs. Go to the store, get bread. They have eggs, get 12. It's ambiguous. Do you mean 12 eggs or do you mean 12 loaves of bread? You're like, I don't know. If eggs, then 12 like you did the math you figured it out arguably if you give instructions like this somebody common sense will dictate get eggs and someone else is going to bring back Sri Manakaraja Bread and you could argue that they're both right yeah that is ambiguity that metaphor is like the first metaphor that actually made sense to me in that are you saying that like the last Scott Hanselman 20 minutes of my metaphors have been trash? No, no, no, no. Sri Manakaraja I meant like in that specific topic out of all the metaphors I've heard from everyone, that's the one. Oh, really? Oh, rock on, dude. Yeah. Scott Hanselman Because is that a prompt? It is a prompt. It's not a programming language. Like here's the deal. English is not a programming language. 90% of the arguments you have with your friends are about the definitions of words. Right? You want to talk about politics? You want to talk about this? You want to talk about that? You'll be fighting and then you're going to second. That's not what that word means. And what? It's not? And we've been arguing all this time because we're arguing about the definition of this word. That doesn't happen in programming. Programming is clear. Compilers are clear. The definition of a keyword and the definition of what a word is in programming is absolutely rock solid, scientifically clear. But with prompting, it's not. So good luck. That's... Oh, and another thing. ([Time 0:24:43](https://share.snipd.com/snip/7b5b1c70-a3c9-4124-832f-648f01e3001c)) - **AI Pulls Design Toward The Statistical Middle** - AI tends to produce the statistical average, which drives bland sameness in websites and products. - Scott argues human designers must Transcript: Scott Hanselman Well, another thing to think about is why AI websites all look the same. Let's think about that. So it's trained on everyone's website. And if you take a bunch of things and you blend it up, like, you know, you take like all these different fruits and stuff. Yeah. And you blend it. It's going to end up being gray. It's going to be mid. It's going to be medium. It's going to be the most average looking website. So all of these, you know, opacy looking clawed websites, they all look the same because they are the statistical average website. So then your job as programmers is to push to the right, push it towards awesomeness and creativity and make it special, you know, push it away from mediocrity. ([Time 0:28:52](https://share.snipd.com/snip/f65d8013-a965-4e9b-a8a9-82572a08c31b)) - **Wired Headphones And Ethernet Still Win** - Wired headphones and wired networking still solve practical problems better than trendy wireless solutions. - Scott notes people lose AirPods so often there's a market selling single pods, and he advises Ethernet to fix flaky connections. Transcript: Scott Hanselman So well wired headphones wired headphones like literally i have a pair of lightning wired headphones i saw lebron go into a meeting recently with wired headphones on like airpods they Look like you've got like q-tips hanging out of your ears everybody loses one of them you go from wired headphones where it's like one thing they always work they have no battery there's No pairing they just work to three things the case and two airpods there's even a website dedicated to buying one airpod it's literally called the right pod and they sell loose airpods Think about that someone built an entire business around selling one-off airpods because people lose them so much that's dumb and then their solution was rather than going back to Wired was to make find my friends work with each airpods you can find the case and then find the left pod and find the right pod if you simply attach them to each other, you will not lose them. So yeah, wired, man. Wired headphones. Wired everything. Because right now, Sri, you're having a little bit of trouble with your connection, and you're kind of coming and going a little bit. Wired. If you can get a wire to your router, it will change your life. Sri Manakaraja Yeah, Ethernet cable. ([Time 0:32:10](https://share.snipd.com/snip/33c225c0-1fae-4974-bc2c-de9e10a881a9)) - **Bit Rot Eats Projects Slowly** - Bit rot is the gradual decay of software over years due to missing compilers, dead sites, and expired certs. - Scott's Baby Smash game from two decades ago may no longer build because dependencies and tools vanished. Transcript: Sri Manakaraja So is there like an equivalent of that in software? I'd say like a slow failure that we're not treating with the urgency it deserves because i guess the crash hasn't really happened it doesn't happen immediately oh wow that's a really Scott Hanselman Deep question yeah that's a pretty crazy redid for dude i've never heard that before so diabetics die of tiny little things that eat away at them, little bits of cellular destruction That happens over 50 years. So we have this phrase in software called bit rot and bits don't actually rot, but like I have this video game that I wrote called Baby Smash. And Baby Smash let my, at the time, one-year smash the keyboard and it would make letters and circles. And it was like a fun thing to teach him how to do. Now, that baby is 20. It'll be 21 in November. Does Baby Smash still run? Does it still build? Can you still clone the repo and build it? There was a time when the bits rotted and you couldn't find the compiler. You couldn't find the libraries. This website was down or that website was down. You see this if you visit, you've Googled for stuff and then you find a website and some of the pictures are broken and then some of the CSS is broken. And the next thing you know, the site doesn't, the certificate dies and then it says yeah i can't visit the site at all bit rot i think is a big challenge and maintaining software takes It's like a garden it's like weeds taking over a garden so diabetes and health is the same thing like if you don't maintain your garden you will eventually be overcome with weeds Good Analogy, Sri. ([Time 0:36:59](https://share.snipd.com/snip/6e601a87-8c87-4023-a6af-52cdf0aee4b1)) - **Stack Overflow For AI Agents Is Emerging** - Forums like Stack Overflow are evolving toward agent-friendly interfaces where AI agents can query and post findings. - Scott mentions Stack Overflow building Transcript: Scott Hanselman VP of product for Stack Overflow on my podcast last week. So she's going to be coming out. Her episode will come out like the next week or two. They're making Stack Overflow for agents where your agent, your AI bot can go and ask Stack Overflow questions. And when it finds answers, it can post there. So it's an AI agent run forum. It's an AI agent run forum where they can actually like, they they can do like, today I learned. So there's like agents going up, TIL, da, da, da, da, da. And they can like put their memories so that everyone can learn them. ([Time 0:42:03](https://share.snipd.com/snip/bd668f62-e227-4cee-9c30-fad295efc606))