Clean Code : Horrible Performance | Full Interview

Recorded live on twitch, GET IN
/ theprimeagen
Casey
Twitter: / cmuratori
Clean Code, Horrible Performance: • "Clean" Code, Horrible...
Author: / @mollyrocket
MY MAIN YT CHANNEL: Has well edited engineering videos
/ theprimeagen
Discord
/ discord

[@JSHyCS]

Prime you should 100% release some longer form interviews with engineers as a podcast. This was amazing!

[@LuisFKA]

Completely agree

[@MrAustin4399]

Would 100% subscribe to a prime podcast

[@bolodia88]

primecast ftw

[@chupacabraj4539]

PrimeRogen! Lets go!!!

[@rummusLoL]

I would watch the shit out of that!

[@5ONE41]

For some reason I expected this to be a prime reacts video, so when Casey actually replied about the camera crispness I was genuinely freaked out for a second lol.

[@adam7802]

lol me too

[@hgbugalou]

same!

[@jesusmgw]

Please make "The Prime and Casey Show" podcast a reality. It would be 100% pure awesome.

[@empresagabriel]

We need more interviews of Prime with Casey!

[@bestformspielt]

This would be pure gold! Please make it happen!

[@surferriness]

Always Listen to what Jesus says 🙏🫡

[@jesse2667]

💯

[@ScrotoTBaggins]

This would blow all other "coding" podcasts out of the water. They're generally terrible

[@_nske]

Such a likeable and engaging guy, his content goes down very easily! glad you invited him and I got to know his channel. And of course your questions and opinions were spot-on as usual :)

[@ThePrimeTimeagen]

tyty

[@romanstingler435]

@@ThePrimeTimeagen Pre-ordered your biography :P

[@brandenpalmer6741]

This is gold. The worse personal project I ever wrote, one that almost made me give up on learning to code, started as a UML diagram.

[@jan0195]

No waaay hahaha. I literally just finished some diagrams for my next project thinking that I'm being organized and smart. I don't even know yet all the functionalities it will have I just come up with things that came to my head.

[@eazypeazy8559]

@@jan0195 I think, that point here is not to reject the planning before implementation. Actually, it’s good to plan something, to draw something, to visualize. I myself is a visual kind of person, so when I want to plan (or I’m stuck) I like to draw. The point here is to reject fcking “clean code” (TM) (c) dogma. and do things in really optimal way from the performance and developer points of view.

[@antdok9573]

@@eazypeazy8559 The point here is to reject overplanning before implementation... Actually, the point is not even that. It's too vague. The point is to iterate in small steps, clean code or not.

[@eazypeazy8559]

@@antdok9573 yeah and and reject fcking "clean code"

[@aakoss]

LMAO :D

[@davidhayes1639]

the best interview I've seen in a very long time. long fan of Casey and Prime. peace.

[@ThePrimeTimeagen]

Hio!

[@FlashGamer521]

​@@ThePrimeTimeagen This is your best interview. Just incase you thought otherwise.

[@p20ph37]

Casey is a wise dude. His handmade hero series is amazing. I cannot believe it was over 8 years ago!

[@someever]

the amount of stuff you can learn from handmade hero is insane

[@alvarohigino]

​@@sliver170 Hahaha, yeah. So you're forced to make these infinite games like WOW.

[@jupiterapollo4985]

Series is cool, game is kind of trash though. It's still in it's infancy and I don't see an end to the project in sight.

[@fpsoftdev]

@@jupiterapollo4985 That completely misses the point of the series. Who cares?

[@spell105]

@@fpsoftdev How does it miss the point? He says in the first episode the goal of the series to show you how to make a game from scratch. After eight years we got what? A game with truly terrible performance and no end in sight on even a technical level. Muratori is without question a good programmer but an awful game developer.

[@BrandNewByxor]

Great interview. Casey's perspectives are refreshing like an ice-cold glass of orange juice on a blistering summer's day. I love his takes on OOP and "semantic compression".

[@callysibben416]

Your enum vid was my fav. The bit about enum variants actually being structs was super helpful. Thank you

[@chordfunc3072]

Thanks for sharing this, you guys are both great!

[@ThePrimeTimeagen]

i really appreciate it

[@0xggbrnr]

This is the best interview I’ve seen on the internet in at least a year. Incredible.

[@reggieriot1388]

seen a bunch of the clips, stoked to see the whole interview!

[@tictactoc9311]

Dude you are amazing, rarely i saw an ex FAANG actually really enthusiastic about programming

[@tictactoc9311]

@@remirth_bergström I thought he had retired

[@Kriszzzful]

@@tictactoc9311 nope

[@caltissue141]

​@@tictactoc9311 He made a video about "why he quit" but despite that, he still works there lol

[@smorebytes]

Personally I think there's very few passionate and vocal Software Engineers. I don't agree with everything Primo says, but props to him and others for really putting out their content and their opinions out there.

[@razorstone3088]

so glad you posted this!

[@Vitis-n2v]

The performance also correlate with power consumption. If you complete the task faster you get more cpu sleep time between tasks and thus longer functioning of devices on battery like smartphones

[@Divisibly9610]

How did I miss this? I love this guy and followed his Handmade Hero for a long time. Great chemistry between you guys :)

[@tordjarv3802]

An additional thought about planning (comment on the discussion around the 45 min mark); usually planning is done in the beginning of a project, when you know the least about it so over planning might hurt you in the long run.

[@acrosstheocean25]

I don't care what other people think, but I enjoy the long deep dives into specific topics. Loved your video on comparing rust and typescript enums, btw!

[@gnarusg8708]

Thank you for releasing, sir.

[@tee949]

One of the best talks I've seen on the channel

[@CoderDBF]

Casey knows what he’s talking about. I want more of this! You guys should consider doing a weekly podcast together or something…

[@EmiliaHoarfrost]

The name of your documentary should be "The Name Is The Primeagen

[@posteisnoob5763]

love you both. Thanks for this amazing interview

[@olafbaeyens8955]

47 is just the beginning.

[@caltissue141]

At first I thought you meant the video runtime

[@olafbaeyens8955]

@@caltissue141 No his age :-)

[@ubuntubad]

@@caltissue141 same

[@KK-wv7vz]

Of the end?

[@olafbaeyens8955]

@@ubuntubad The best times of your live

[@P-39_Airacobra]

Actually the best interview about software development I've ever watched

[@AungusMacgyver]

This is my 3rd Casey & Prime video of the day. So. Good. Everything they are saying makes so much sense.

[@nclanceman]

Wait, dude was on Dungeon Siege? Dungeon Siege was _awesome!_ I wonder if he worked on the level streaming tech, cause that was a literal game changer back in the day.

[@MichaelPohoreski]

Yes, it helped usher in 3D streaming worlds! (Technically Ultima series did 2D open world streaming in the early 80s) but Dungeon Siege definitely influenced a LOT of people - gamers, designers, and programmers alike.

[@remirth_bergström]

Hell yeah I've been waiting for this

[@karlheinzneugebauer]

Invite him more often!

[@friend0]

great video love how hard you guys are vibing

[@friend0]

you shouldve asked him why he was using notepad in his video tho just cause i bet hed have an interesting answer

[@Rakkoonn]

​@@friend0 Probably just for the video, he doesn't actually use it. He used Emacs for a long time, but switched to another editor someone from his community made at some point.

[@someever]

​@@Rakkoonn yup he uses 4coder

[@goldbarth]

Awesome to watch/listen. You guys give me some extra energy pump for the day and beyond.

[@insydian]

This was amazing!

[@ArkhKGB]

Around minute 40:00 just casually paraphrasing Continuous Delivery: deliver workable code fast let you get feedback fast so you can refactor and improve it as needed. Speed gives you quality, and coding with quality in mind makes you fast.

[@Rohinthas]

absolutely, just got into this channel/streamer and the stuff where he actually goes into the detail is surprisingly insightful for a junior

[@chralexNET]

I watched Casey's video on quaternion double-cover yesterday where he explained how it could be useful in animations. There was a little bit of a hook to it, which is "here is what the industry doesn't do, which would be much better!", but apart from that I thought it was a pretty calm and informative video. My point is that, some people do watch the more technical stuff, and it is up to the content creator to know their target audience, and it just happens that the highly technical audience, who has enough spare time to watch videos, is not that big, but they do watch it.

[@chainingsolid]

got a video title? I'm not seeing it..

[@chralexNET]

@@chainingsolid "Quaternion Double-cover and the Rest Pose Neighborhood - 2006" channel: "Casey Muratori"

[@caltissue141]

loved this interview

[@ShinDMitsuki]

What always astounds me is the massive amount of justifications I see in the comment sections for videos like this, yet the radio silence when people like that twitch guy made a better app than the twitch Mobile app than the official in a hackathon. There are a lot of ramblings about premature optimization but I really believe most people don't actually know what optimization is when that quote was invented. The win prime got with that Netflix code wasn't "optimization", it was actually doing the task correctly, and he still had to fight for the right to do even that. Sure, when he hops into the bosses room and says they need vectorized uis for each major Android processor as well as Apples you can tell him he's going overboard, but do you really think a modern interface should have over 1000ms delay for any action? Again, I hear a lot of hiding behind the veil of the internet for the latency not mattering, or something, but rtt to anywhere in the world is generally going to be less than human reaction time, so even that argument to me makes no sense. Ived piped plenty of data over the net that was still in time critical software, I just had to care about both ends being fast

[@WizardofWestmarch]

The problem is most people think premature optimization is ANY optimization when it should be line level bit twiddling tier madness that can take a day to get 1% when you need that level of improvement. Those are the last resort optimizations not picking the right data structure or algorithm to properly use cache and quickly solve your problem.

[@seanpe8474]

​@@WizardofWestmarch also to add the MVP principle that agile promotes tends to lead to a lacking mentality. agile culture screams for a feature/product to be finished, no matter how bad as long as it works, but when you prioritize completion over practicality you sometimes build something that works but is difficult to fix/optimize. I feel like that's a large part of why people love rust so much; you can build an MVP, but with a good standard.

[@swagatochatterjee7104]

What's the point in trying to optimize for sub milliseconds when you are bottlenecked by network latencies of couple milliseconds

[@sethrenshaw8792]

@@swagatochatterjee7104 Because that code still needs to run on a server somewhere, and when that code runs for billions to trillions of user requests per day, suddenly that "wasted effort" on preventing that extra 1-2ms per request add to thousands or millions of dollars in extra energy costs for the business that wrote the software.

[@phazechange3345]

@@sethrenshaw8792 It's not just that though. When every part of your decision making becomes "but but but but... IO is slow!", those decisions compound. You write thousands of lines of technical debt that cause request latency to jump from 10ms to literally 1000+ms *per request*, and then you also pop off *hundreds* of requests per click. All of a sudden, you have a 25 second per click load time for someone on a wired fiber business class connection that lives 15 minutes away and has a $10,000 PC. And what will that developer say the entire time you call them on this absurdity? "There's nothing we can do. IO is slow bro. That's PREMATURE OPTIMIZATION!"

[@mannycalavera121]

Thank you for posting this

[@dovos8572]

41:00 "the first version has to be perfect" or "my first version works, so why should i change it" is one of the biggest traps in anything you can do, but mostly known in Author circles. in reality the first version you make is NEVER good, let alone perfect. you can always edit/refine it and get way better results by doing so. there is a reason why big successful author take at least 2/3rd of the needed time to rewrite and edit their story. if not more. it is sadly one of the most unrealistic habits we get teached in school and one that gives young people the most trouble. everything you do/write has to be as perfect as possible in the first try/version because you don't get the chance or time to change what you did.

[@equivocator7727]

Fantastic interview to watch, tons of fun! I'm curious which video Casey was talking about where you reacted to his video. I looked for it but couldn't find it.

[@Paintballman251]

I likes the rust vs typescript enums video! I appreciate you having made that anyway

[@xpygx1599]

What a good interview. Yall are great. Love you guys.

[@lisathewhitewolf]

It can help if you are as animated as possible when explaining things, and like go up and down in intensity, you know? But I know that's tough with technical things. I'm a brand new baby coder and trying to learn from whateveri can find for free, so thank you for whatever you put out 😊❤

[@jerodev2]

I love this! I'm realy into optimizing code, even miniscule micro optimizations. However, at work it's: "You've got one hour to make this work, Go!"

[@heathenmonk1249]

Great Interview! I hope more comments help on getting more content

[@CertifiedDynamite]

Casey in high school really reminds me of myself to the point when I had literal flashbacks of those times :)

[@davidthegood]

You're exactly right about the YT clickbait. It works, and if you don't do it, you don't get views.

[@marcobrien4137]

That was awesome. Casey is the man.

[@tedbendixson]

Oh hell yeah! I will always be grateful to Casey for teaching me how to program games

[@riittap9121]

I seldom watch anything over 30 minutes. If it's longer than that, I only listen to it with my headphones while doing something else (cooking, laundry, cleaning, gardening). Tutorials are the kind of material that needs to be watched and thankfully they are usually chopped in shorter videos and combined to a playlist. For me, this channel is mostly for listening. It's a mix of entertainment, technical advice and career/life coaching.

[@ExpertOfNil]

This was an awesome interview

[@Trokkin]

Great interview. And the point of performant applications is so soothing to hear. I hate using online services just cause of the loading times. That's why I use Client mail apps not google, Obsidian not Notion, even VSCode live share not some web collaborative ides.

[@aoeu256]

The problem with slow websites isn’t just JavaScript code, it’s the slowness the network.

[@GeekMasher]

Great interview!

[@Nate77HK]

I think the distinction between advocating premature optimization and just being aware of performance-killing habits is important. Because, frankly, highly optimized code can also be incredibly inflexible. That's another difference between the webdev space and video games - web applications tend to grow over time but most video games don't double or triple their code over their lifetime.

[@MH_VOID]

I'd think it's more the video games are largely swapping out code rather than just adding new code, in addition to many changes just not really being code (i.e., stat adjustments all the time for buffs/nerfs each patch, skins and voice packs and other assets being added forever all the time, but not really entailing any code in most cases)

[@CottidaeSEA]

​@@MH_VOID A lot of those things are largely pure data. Unless there needs to be very specific interactions, a skin/model can be packed with all animations and run in the renderer just fine. The issue comes when the game adds something like an item which has unique effects. I'll take League of Legends as they have a recent example. There is a biscuit item which is supposed to give mana permanently when used or sold. The issue is that if they were sold and you undid it, the mana either increased or decreased permanently and was infinitely recreatable. That is because the item is the only one which has an effect on removal.

[@GonziHere]

Tripling your source code is almost never warrantied. The issue is that the abstractions actually make it hard to keep the solution DRY (think of all of the boiler plate code). On the other hand, you could simply write a solution to your current, actual, contained problem and when you'll add a new functionality later on, you'll also refactor the first one for some code reuse.... Like code on the project absolutely should grow logarithmically, not linearly.

[@CottidaeSEA]

@@GonziHere DRY can sometimes lead to obscure abstractions and difficult to maintain code. Sometines it's just better to repeat yourself if you aren't doing exactly the same thing.

[@sharos404]

I noticed this abhorrent inefficiency of other apps very clearly when I bought an Android game called Slice & Dice. The app seems insanely optimized and doesn't use any battery at all, which is not very usual for mobile apps. Most mobile apps usually just turn your phone into a pocket heater.

[@Snorehog]

Funny because on M1 Mac macOS it ran terribly last time I played 😂

[@Snorehog]

Sure! I have a pc too for games. Still this game isn't a 3D game etc so it should run perfectly fine on newer Mac. But the developer probably used a framework that didn't prioritize working well for Mac. Multiple operating systems is complicated, and yes Apple haven't prioritised making it easy to make games for Mac. Still, would love to play this game on iOS and Mac hehe@@ninsix

[@dmitripogosian5084]

One my phone sudoku decreases battery life by a factor of 3. Sudoku - where there is nothing to compute and nothing is moving ....

[@mentis4046]

The Video about clean code does make sense but its not only black and white. Clean code was not meant to make performance gains but to trade some performance for better readability/maintainability. It all depends on the goals of your software and if you can afford those performance losses. I would say that most companies are fine with less performance if you can easily add more features to it and fix/find bugs more easy.

[@Anon.G]

Your claims are unsubstantiated and flat out wrong

[@mentis4046]

@@Anon.G just Google why you should use clean Code. Usually you find "reader focused". Clean Code is for humans not machines. Machines dont care how complex your code is, your coworkers do.

[@Ehal256]

​@@mentis4046users care when your application takes 20 seconds to start up when it should be instant

[@MrGman166]

@@mentis4046 The statement "Machines don't care how complex your code is" is even objectively wrong especially when talking about a program that is going to access things like virtual function repeatedly during runtime and the compiler cannot optimize the operation and has to perform searches for the method and its signature in a large DDD "Clean Code" codebase. This does not even account the amount abstractions that are used represented as classes or phantom interfaces been used that will drag down the compile time of the program for a large application that make the testing process even more sluggish. You need to retouch basics of what the compiler is doing to the code as well as even just do a simple debug for virtual functions or dynamic dispatching that even disprove that statement. Also stop trying to google for "why you should use a clean code". If you are a programmer or software dev/engineer, it's your job to apply, measure and test the paradigm or approach against use cases and arrive at the architectural decisions based on the performance measurements and parameters of your testing. The fact that you had to do that google search means you don't even know why you are using that methodology and you are justifying software practices on the basis that someone else reasoning satisfies it without any kind of testing or measurements you have done on your end.

[@TheExtremeCube]

Caseys objection to that is that it's not founded on any facts. Did you actually measure if your clean code solution has improved the onboarding of new people, or reduced development time? Its just taken as a given that clean code is better for that without any proof.

[@JimBrännlund]

I'm immediately sceptical given his video has comments turned off.

[@muzikjay]

Such a great discussion.

[@realdomdom]

love the title

[@kaansouth8789]

More conversations with Casey pleaseeee!

[@whiitehead]

This is crazy, I thought this was a react video then Casey started responding

[@vesk4000]

Awesome!

[@TechMaster404]

i like how youtube reconized john blow as an english composer from the 17th century

[@babyboie20]

This vid explained a lot of Prime’s ideology for me lol, now I have a better understanding of why when I like or don’t like something. Really good one vid!! Also, love the guess too. I feel smarter just listening in lol

[@slowpnir]

Fowler's method is about keeping complexity at bay and the API surface at minimum. Casey showed, that when number-crunching, "clean = bad". After that he concluded "clean = always bad" out of the blue. Let's wish him good luck in designing big systems with will not fit even in L3; and the complexity that requires 3 JOINs per request.

[@dandymcgee]

Casey + ThePrimeTime = absolute content gold

[@rossvold]

i love this

[@PGDJ88]

Holy shit my friends and I spent around 18 straight hours playing Dungeon Siege in a basement in high school. Love that game.

[@kradoste6268]

This is a classic example of “the exception does not break the rule”

[@victorian1707]

We love Casey

[@baxkill]

DUNGEON SIEGE?!?! LETS GOOOO!! DUN DUN DUDUDU DUN DUN

[@riolly]

This is what I need.

[@vibrantneon.]

That first minute and a half hit home what I dislike about online. People don't consider a calm thing worth listening to, without a hot take.

[@desmondbrown5508]

I'll say that one of the worst things about polymorphism for me (especially in OOP code with DI) is that it can become quite ridiculous to even keep track of what the hell anything is/does. Like sure you SAY x thing does this... but really it calls y thing over here which calls z and their all interfaces so I need to find out which interface actually has the bug/problem. And yeah... it gets really out of hand.

[@_jagger1407]

exactly, I don't see how "clean code" with abstractions, polymorphism and everything hidden behind access modifiers, somehow contributes to "readability". If anything, it's a detriment to readability AND performance. Maybe maintainability? But keeping all the abstractions and having all these arbitrary rules just sounds less maintainable too, idk

[@scotthinton4610]

I recently did this to myself. I thought "oh wow, this C++14 DI framework in a single header is really cool, and I don't have to worry about my constructors anymore". I've probably spent as much time trying to figure out WTF the DI lib is doing (including inferring behavior, yikes) and why things aren't working. I regret reading Clean Code.

[@Thurrak]

"Dungeon Siege wasn't particularly huge"... Bro that game was the hugest thing to me as a kid. Thank you for working on that game.

[@vladimirzagursky144]

Love that! I wish I could give you more likes. Thank you so much!

[@ritchielrez1680]

Bruh really I never thought this collab would be possible!!

[@bananesalee7086]

I feel like an extra nerd for knowing the cited video before finding this interview, i'm wasting my life

[@DimaJP]

was nice to hear - so as always - the truth is in the middle

[@thiswallz]

You are so smart, pls continue with your smart stories when you save millions and millions.

[@ifstatementifstatement2704]

I also think when I implement something how to do it the best way possible. The thing is that after implementing it drawing on my entire experience so far as a programmer, I usually cannot see a way to do it any better and leave it at that.

[@RetroGenetic]

Dungeon Siege is still one of my favourite games to go back to every once in a while, and hearing RAD game tools brings up memories 😂

[@hbobenicio]

solve & test first, abstract last. measure & optimize then, repeat until enough. ship it, done.

[@ebn__]

I wasted about 8 of my best years on EverQuest 2, thank you Casey.

[@nchuchelov]

yeah my recent discovery was that im cpu limited in battlefield 2042 on my system

[@MyAmazingUsername]

I always think about a design before implementing, to make sure it can be extended. Otherwise you're just writing bad code. But I agree that if you spend time writing a bunch of big frameworks before writing actual code, that is going too far.

[@khatdubell]

Actual interview starts around 6:30 for the impatient, like me.

[@emmanuelomeogah2634]

Wish I could like this multiple times 👏

[@Eidolon108]

gonna buy 10 of the same pants and T-shirts now, thanks for enabling me

[@SpookySkeleton738]

The word "pessimize" is used elsewhere! Clang has a warning -Wpessimizing-move for when someone uses std::move in a way that prevents copy elision.

[@akademiacybersowa]

18:50 when my GPU blew and I was left with just Intel iGPU, The Witness was the only rather modern playable game in my library xD

[@t1nytim]

What I find interesting about the content argument, and using something like Netflix as an example. I decided (number of years ago now), I would only pay for so many services. And the ones I would pay for, would be the ones with the best UX. As I knew I was gonna be going to the seas anyway, for stuff not provided by the services I was paying for.

[@nightbat13131]

"hey guys I think something is wrong" is different than "hey guys look at this test that shows something is wrong"

[@someever]

this interview was awesome

[@pdr.]

On the topic of not knowing the best code structure the first time through: this is the exact reason I prefer writing programs that are cobbled together from loosely coupled components. Not only does that allow each part to be written in the best language and stack for the job, but if any component sucks, just throw it out and rewrite it, using a different stack if necessary.

[@JamesMiller-ih3eu]

Banger

[@nikonyrh]

Haha auto-play transitioned me to this interview from a Cypress Hill live show, and the tab was playing on the background. I thought this was a post-concert interview! "And I'm just like guys, I talk calmly all the time, and no one cares." And I'm like "yeah yeah makes sense, and I like the band's aggressive style anyway".

[@satellite964]

Can't believe you didn't ask Casey M about ZINC OXIDE!

[@rickwoods5274]

41:00 yes yes yes yes yes. I have been forced into big-design-up-front too many times and I've been burned every time. I've learned to just write the code I think I want while "designing" but not saying anything about the code.

[@TJ-hs1qm]

I think the endeavor is nobel but hopeless ... Jevons Paradox, named after the English economist William Stanley Jevons. It states that technological improvements that increase efficiency in the use of a resource will tend to lead to an overall increase in consumption of that resource, rather than a decrease. This is because the increased efficiency makes the resource less expensive, which in turn encourages people to use more of it.