DHH: Future of Programming, AI, Agentic Engineering, Vibe Coding & Linux | Lex Fridman Podcast #501

0:00 There are decades where nothing happens and weeks where decades happen.

0:05 And we have seen decades of progress happen in the last nine months.

0:11 If you're not recognizing the gravity of the moment, that's the delusion.

0:17 That's the psychosis.

0:19 Who would not get delirious if suddenly a genie pops out of the bottle and says,

0:24 "You can have whatever you want.

0:26 Every feature you've ever dreamed of in an operating system,

0:30 I can deliver them to you, most of them in five minutes,

0:33 a few in 20, and if we really go hog wild,

0:36 it's gonna take me two hours." I want the fastest car breaking the speed limits.

0:40 I want the diver watch that can go down the Mariana Trench.

0:44 I want the operating system that can install in less than 60 seconds.

0:48 One of the things I always loved about race cars was when I would

0:52 stumble out of the car absolutely smashed and barely able to hold my head up,

0:56 and I'd lay down on the garage floor and just think,

0:59 "Holy fuck, I'm alive." The Overton window does not open itself.

1:03 It opens one nudge at a time by people risking a little.

1:09 A little reputation, a little pushback, a little criticism,

1:12 or maybe sometimes a lot of reputation

1:14 or a lot of criticism or a lot of pushback.

1:17 Creating life with another human that you love is

1:21 literally the peak experience of being on the planet.

1:28 The following is a conversation with David Heinemeier Hansson,

1:31 also known as DHH, creator of Ruby on Rails, CTO of 37signals,

1:37 creator of the new Omarchy Linux operating system,

1:41 best-selling author, race car driver,

1:44 and one of the most outspoken and prolific programmers in the world.

1:49 For the past 20-plus years, for him,

1:51 programming meant meticulously handcrafting beautiful Ruby code.

1:56 But recently, since late 2025, DHH has embraced the AI revolution

2:02 and has openly transformed himself into, once again,

2:05 one of the most outspoken and prolific agentic engineers,

2:10 even though he hates that term.

2:13 So let's say practitioners of whatever programming

2:16 is becoming where AI is doing most

2:19 of the actual programming and the human steers

2:22 the ship with high-level design, vision, and taste.

2:26 Yes, it's true.

2:27 DHH is at times controversial, but he's always fearless,

2:32 brilliant, and fun to talk to.

2:34 This was, once again, an intense, eye-opening,

2:37 and wild rollercoaster ride of a conversation.

2:40 This is a Lex Fridman podcast.

2:43 To support it, please check out our sponsors in the description

2:46 where you can also find links to contact me,

2:49 ask questions, give feedback, and so on.

2:52 And now, dear friends, here's DHH.

2:56 13 months ago, we sat down right here to talk about programming,

3:00 and then everything changed.

3:02 At that time, you were a bit skeptical about

3:06 the role of AI in the process of programming,

3:08 and then we went through this rapid evolution of agentic engineering.

3:14 So, simple question to start:

3:16 How has your view on AI's role in programming changed?

3:21 Are you excited?

3:22 Are you terrified?

3:23 Are you on an emotional rollercoaster ride?

3:27 I am incredibly excited.

3:29 You've been talking about fun a lot.

3:31 There is none of the existential threat.

3:34 That doesn't exist for me as an emotional component.

3:38 It is only there as an intellectual component,

3:41 and the emotional component for me is 100% pure, unadulterated joy-...

3:47 and optimism and amazement that we've made computers do this.

3:52 And I find it so interesting that we talked just 13

3:56 months ago because it's like we talked in different universes, different eras.

4:03 I love this quote.

4:04 I think it's Lenin.

4:05 There are decades where nothing happens and weeks where decades happen.

4:11 And we have seen decades of progress happen in the last nine months.

4:18 I mean, imagine you're there when the Wright brothers take flight.

4:21 Yesterday, the New York Times would write,

4:23 "It's gonna be 10,000 years before we fly," and then the day after,

4:27 we're up in the skies,

4:29 and just a few years after that, there were cross-Atlantic planes.

4:33 The whole world has completely changed.

4:35 What a blessing to be there in that moment.

4:38 If you zoom out and look at all of human history,

4:41 how many humans got to live within the same epoch that they were born in?

4:47 They never saw that complete change of the world and of society.

4:52 And to have been blessed with two of those feels just such a privilege.

4:59 I got to see the internet-...

5:02 from pre-internet to post-internet, and then now pre-AI, post-AI.

5:09 What an amazing run.

5:11 How fortunate.

5:12 Yeah, but the thing is, this feels like a thing that happened

5:16 faster than anything else in human history.

5:18 Correct.

5:18 So I would say somewhere around December, maybe late November-- November 24.

5:24 This is...

5:26 That's the exact moment.

5:26 You know, when they talk about when one nation invades another in a world

5:30 war or something like this, it— this is

5:32 how we talk about AI changing everything.

5:34 Yeah, the...

5:35 And it really just sh- shifted for many great developers.

5:39 It shifted to where AI is doing some, basic autocomplete,

5:43 maybe writing 5, 10, 15, 20% of code to writing 80% of code.

5:49 Or 100.

5:50 Or 100, yeah, especially if it's not a public-facing product.

5:55 To me, that's why even when I look back upon our conversation a year ago,

6:01 I don't actually have different opinions.

6:03 I have the same opinions.

6:04 A year ago, I did not like the mode of AI we were offered.

6:09 It was the autocomplete mode, or it was the AI chatbot mode.

6:13 Now, the chatbot I actually liked, as we talked about.

6:16 Great tutor right from the get-go,

6:18 great way of looking things up on the internet,

6:21 not what was gonna replace me chiseling code.

6:25 But-...

6:26 then we get the agents.

6:28 And the agents start out being curiosities for about five minutes,

6:34 and then they get amazing, and then they get,

6:38 "Oh my God, is this AGI?" And all of that happened since last year,

6:45 even just within the last nine months.

6:47 We have basically these few phases here.

6:50 We have AI in the pre-agentic era.

6:53 I was excited about that, but it was not

6:57 fundamentally rewriting the rules of the game for me.

7:00 It was not completely changing how I worked.

7:03 I was still chiseling code, I just had a little helper, a little sidekick-...

7:09 who could bounce ideas off, and I could look up this information online

7:14 and so forth in a more efficient way.

7:15 It was just a more efficient way to do what I was already doing,

7:18 and it didn't change the emotional connection I had to the computer.

7:22 Then we get to November 24th, 2025.

7:27 Opus 4.5, to me, was the dividing line, where suddenly...

7:32 I didn't even try it on the 24th.

7:34 I think I tried it on the 26th.

7:37 I give it a couple of tasks, and I realize that the quality of the output

7:45 is uncannily close to what I would've written.

7:49 And I remember just leaning back and thinking,

7:54 "What just happened?" "How did we go from this autocomplete mess that I was

8:00 talking to you about in the summer to this just a few short months later?

8:04 How did we get both the increase

8:06 in intelligence and then also the increase in usability?"

8:11 This agent harness question where I don't know if

8:16 Opus 4.5 was that much smarter than Opus 4, which is what we had in the summer,

8:22 but its ability to instrument your computer,

8:24 to use tools, to check its own work, to apply its intelligence in such a way

8:30 that you could get real meaningful work out of it, was completely different.

8:35 And I think this is then the big change that happens for almost

8:40 anyone who paid attention and started playing with it over the Christmas break.

8:43 This is what I heard-...

8:44 from Shopify and other places with lots

8:46 of employees who suddenly had a breather,

8:49 suddenly had a couple of weeks to lean back and just

8:52 look at what was going on, gave it a try,

8:55 and had the same mind-blowing experience that these agents

9:00 were of a different genre than what we had before.

9:04 And then what we get to is, I'm already excited at this point.

9:11 Like, by December, I'm already revisiting sort of all my priors-...

9:15 and going like, "Wow, if it can do this, can it also do that?" "Oh,

9:19 yeah, it can." And again, Opus 4.5 now looks like a retarded model.

9:26 And this is the magic of this progress, is you think you've reached something...

9:32 I remember thinking at the moment,

9:35 "If this is the last model we get, I'll be set.

9:38 I'll be happy." I could live with Opus 4.5 for the next 20 years-...

9:43 and you would hear no complaints from me

9:45 because it was just so incredibly amazing

9:48 to see an agent do all this work in the way that I wanted it done.

9:52 Because it was not just about it being able to solve a task,

9:56 it was also that I could look at its path there and go, "Yep.

10:01 Yep.

10:02 Maybe not there, but almost," and it would give you two notes.

10:06 You'll get to where I wanted to go in the way I wanted to go there.

10:10 You could produce code I wanted to merge.

10:12 You could produce code that actually looked beautiful if it was written in Ruby.

10:18 Rust, different question.

10:19 But this ability for the agents to truly become

10:25 an extension of how I wanted to work was very novel.

10:30 But then we wait just until early spring, and suddenly we get sub-agents.

10:37 We get harnesses that can subdivide the task,

10:40 and something that would take Opus quite

10:43 a while suddenly took a fifth of the time, a tenth of the time,

10:48 because it could get chopped up and suddenly

10:51 you got eight sub-agents working for you.

10:54 But both of those two first phases of the agentic age,

10:59 to me, still felt like I had to drive.

11:01 I could tell it what I wanted, where to look for it,

11:05 and steer it a little bit when it went off,

11:08 and then we'll get there, and I would go much faster.

11:11 But I had to be in the driver's seat.

11:13 I had to tell it what I wanted,

11:14 and I had to be the reviewer, the auditor of what was coming out.

11:20 And then finally now, this summer, with Opus 5,

11:25 Fable, and Sol, GPT Sol, and to a lesser extent, some of the open weight models,

11:33 we've arrived at a new era where I'm not telling it where we're going.

11:37 I'm telling it the problem I have.

11:40 I'm telling it the fuzzy, vague idea I have.

11:44 It tells me where we're going.

11:47 It tells me which path to take,

11:49 and I will still look at it because I'm a curious

11:51 person and I like computers and I like the outcome of it,

11:55 but I really kinda don't have to.

11:58 I've become optional in the part that produces the code that picks the route.

12:04 Remember when early GPS systems came out?

12:09 They were amazing compared to looking at a map,

12:11 but you still wanted to pay attention.

12:13 Is it gonna drive you in the harbor?

12:15 I remember these newspaper articles.

12:17 "Ah, GPS, they're terrible because people don't pay attention,

12:19 they drive in the harbor." When was

12:21 the last time GPS drove anyone in the harbor?

12:23 Like, that just doesn't happen anymore.

12:24 In fact, the cars now just drive themselves, right?

12:27 And this is where we've arrived at now, that I can trust.

12:33 For the domain I'm working in right now, I can trust it,

12:36 and feel completely confident that it's gonna have my back.

12:40 It's not gonna do something stupid,

12:42 and if it does something stupid, it's gonna be able to recover.

12:45 Mm-hmm.

12:45 We should mention that there's all kinds of domains that-- Correct-...

12:49 programmers operate in.

12:50 There's...

12:51 I actually don't know,

12:52 but I think the most common domain in development is, like, web dev CRUD.

12:58 You have a database.

13:00 You have a user-facing interface, and it does something back and forth,

13:04 and it could be internal to just one person,

13:07 to multiple people, to a small number of people, or to the world.

13:11 And I think for that, it's— I mean,

13:14 you could really get to the 100% of code written by AI and really if

13:20 you're a good programmer and you have

13:22 a good intuition about what's happening behind the scenes,

13:25 you can legitimately not look at the code,

13:28 at least that's, that's been my experience.

13:30 I don't know how much,

13:33 prior experience with programming you need to have to kinda

13:36 know that things are working correctly behind the scenes,

13:39 like intuitively by observing the ripple effects, the symptoms of the system.

13:45 But I think for that domain, it's close to 100%,

13:50 and then there's domains like you're also

13:53 operating in, which is writing a Linux distribution.

13:55 There, maybe there's more because you're obsessed

13:58 with speed and all that kind of stuff.

14:00 There, maybe you need to look at the code a little bit more.

14:03 Then there's maybe safety critical systems all the way down

14:06 to operating a nuclear power plant or a self-driving car.

14:09 Maybe you need to look at the code more carefully.

14:12 Yes, but that said, AI is insanely capable at both finding-...

14:17 and fixing security vulnerabilities.

14:20 This was the whole blowup about Fable.

14:22 This model was so capable of finding holes that a an attacker

14:27 could exploit that it was simply not safe to release.

14:30 So the irony here is that when you look at that field,

14:34 it seems like we've reached levels of intelligence that virtually no human can

14:42 match because many of these security

14:44 holes are about stringing combo moves together.

14:47 You find one little vulnerability here that by itself

14:51 might not be the worst thing in the world,

14:53 but then you combine it with four others,

14:55 and suddenly you have RCE, remote command execution.

15:00 Humans who are able to do that are very rare.

15:04 They usually work inside state-sponsored

15:07 organizations or other clandestine operations.

15:11 They're not just out and about finding things.

15:14 So they've gotten so good at that.

15:17 The Linux distribution is why I've gotten 100% AI pilled.

15:21 Because I've been working on Omarchy for the last three months,

15:26 this version that just dropped a few days ago called Quatro.

15:31 And almost right from the beginning I'm working on that version,

15:35 the agent acceleration neared 100%, and in the last two months it has been 100%.

15:41 I have not written-- Really?

15:44 any of the code that's shipped in Quatro by hand.

15:49 I've reviewed the shape of all of it.

15:52 I've reviewed the individual lines of anything that's

15:55 critical in the model layer of the system,

15:57 and I've not looked at a bunch of the UI code.

16:02 I have not looked at a bunch of the auxiliary code,

16:05 and I've not written any of the new functionality entirely by hand.

16:11 But then the web part, actually evolving Basecamp and Hey,

16:17 our professional products that have lots

16:19 of users and are relatively large code bases,

16:22 have proven surprisingly tricky to fully accelerate with agents.

16:27 We just released Basecamp 5 not too long ago.

16:31 That was the first product at 37signals that was really agent accelerated.

16:36 Because we were in this final sprint phase from around February.

16:41 By then, agents were already good.

16:44 And we had this early surge of, "It's solved.

16:47 We can just have the designers do the programming.

16:50 They know what features they want.

16:52 They know what shape they want it to take.

16:55 Let's just— Let them vibe." And we let them vibe.

16:59 And we ended up with a lot of PRs

17:03 that individually perhaps could have been justified for a hot moment,

17:07 taken all together, destroyed the architecture of the system.

17:10 And we actually had to clean up manually, mop it up by hand,

17:14 by human hand, to get back to an architecture that felt cohesive and coherent.

17:20 So we still have a bit of that— that was February, by the way.

17:23 Things are quite different now.

17:25 Well, hold on a second.

17:26 What's the lesson from that?

17:26 Is one of the lessons from that that you have to be

17:29 a programmer at this stage to be able to vibe code?

17:34 To be able to vibe code on existing substantial- I see.

17:38 code bases, even if they're crud,

17:40 if you wanna retain the element of architecture

17:43 that got that system to where it was.

17:45 Now, that's also a point where I've stressed many times

17:49 that when people accuse vibe coders of being slob generators,

17:54 I go right back at them and say, "Have you looked at the average programmer's

17:59 output?" That is some other slob too.

18:02 If you've looked at behind the scenes of many great companies and what

18:06 their code bases look like after there's

18:08 been 3,000 humans through them, they're awful.

18:13 Absolutely awful.

18:14 Can you tell me the intuition you have?

18:17 I seem to just, stepping back and observing the different

18:19 apps that we all rely on, I don't know, Adobe Photoshop, all this kind of stuff.

18:23 There seems to be the progress on development there has not accelerated.

18:29 So why is it, what lessons can you draw from Basecamp,

18:32 like well-established, huge user base?

18:35 Why aren't we seeing, like, super rapid increase in, like,

18:39 new updates, features, all this kind of stuff in these well-established apps?

18:44 Multiple reasons.

18:44 I'll start with the first one that's the most critical.

18:47 As soon as you're having human teams work together on something,

18:52 the bottleneck is rarely implementation.

18:55 It's human bandwidth and communication.

18:59 When you have a product manager and a couple of designers

19:02 and a VP above them and a CTO above them,

19:07 and everyone wants to be part of the shaping

19:11 process because we're all justifying why we're here,

19:14 that's where all the productivity goes to die.

19:19 The revelation I've had working on Omarchy the last

19:24 three months is that to get that magical 10X, 100X, in a few rare cases,

19:34 1000X productivity boost, you have to interact with the agents directly,

19:40 and you cannot intermediate that bandwidth

19:44 with another human because it's simply too slow.

19:49 And on the one hand, that's a bit of a bummer.

19:52 I mean, I like humans, and it's great to work together.

19:55 But it also means we need to temper our expectations

19:57 with what these agents can do if it's humans driving it,

20:02 and you have three layers of approval

20:04 and all the other machinery of a large corporation.

20:08 The implementation part is only a small segment of it.

20:14 The other thing I'd say is that most organizations don't know what they want.

20:21 They don't know how to make it better.

20:23 They're not bottlenecked on implementation.

20:26 They're bottlenecked on ideas.

20:28 They're bottlenecked on vision.

20:30 They're bottlenecked on taste.

20:32 And if you don't have those element

20:36 in excess of your implementational capacity, it doesn't help.

20:41 So you can make a lot of shitty ideas come true, then what?

20:44 Are you gonna ship that?

20:46 That sounds like, I don't know, something coming out of Microsoft.

20:49 That's not what we're trying to replicate here.

20:51 That process is not actually...

20:52 Because we already had this.

20:54 If you step back for a moment and think of the towering organizations

21:00 we have who have had tens of thousands of programmers at their disposal.

21:06 I mean, I'm picking on Microsoft here.

21:08 I love Microsoft some of the time,

21:09 but I'll pick on them in this case because they have had endless resources,

21:14 endless programming capacity for decades, right?

21:17 This is what's showing us that just being able to write

21:22 a lot of code does not produce great compelling software.

21:25 Now, the other thing is we've had this capacity for about six months.

21:33 That's not very long in human life

21:37 cycle capacity of internalizing what's going on.

21:40 And I think it's actually funny that the critique of AI is,

21:45 why isn't it going faster?

21:47 What are you talking about?

21:48 We've not had any other form of progress that has moved as fast as AI,

21:51 and you're impatient because in the last three months we haven't

21:54 rewritten the whole world and made it a utopia of software goodness?

21:59 You're basically saying everybody should be switching to Linux now.

22:01 One of the arguments is, like,

22:03 we can rewrite all the software that's not available in Linux in Linux.

22:06 Linux has been the love of my life for many years,

22:08 but one of the reasons I'm still attached to Windows and now,

22:11 uh, Mac is because of video editing, Premiere.

22:15 We're gonna fix that.

22:16 But-- So the question is, who's gonna build Premiere and Photoshop for Linux?

22:22 And it feels like one person can now.

22:27 100% one person can.

22:29 And so, and I'm allowed to be impatient.

22:31 In fact, me or anyone else,

22:35 impatience is the first step of doing it yourself, right?

22:39 Like-- Correct.

22:41 and there's a community of tens of thousands

22:42 of people that use Adobe Premiere or DaVinci,

22:45 all these non-linear video editors, that know the frustrations.

22:48 Everybody shares them.

22:49 You can look at Reddit.

22:51 You can look at the forums.

22:52 Everybody knows it.

22:53 Uh, I don't know what it is.

22:54 It may be the management.

22:56 It may be the meetings.

22:57 All the things you've always discussed is the, The bureaucracy that's

23:01 inside companies that's slowing everything down in the development of features.

23:05 But it feels like there— if you just let

23:09 some of the developers loose with a bunch of agents, you can fix all that.

23:13 But again, based on everything you're saying,

23:16 maybe the way to do that is to start over from scratch,

23:20 maybe through open source,

23:22 or somebody launches a new company that within large companies now it's hard

23:27 to, to really accelerate into the agentic era by letting developers just build.

23:33 This is the classic innovator's dilemma.

23:36 These companies have gotten so good, so established at the old way,

23:42 and therefore their entire structure, management layers,

23:45 processes are tuned for a time that no longer exists.

23:52 But you can't pivot that.

23:54 These are super tankers.

23:56 It just doesn't happen.

23:58 This is why we're finally getting this upset in the technology industry.

24:04 For a while, I was very upset about

24:07 the duopoly between Apple and Google on mobile, because it felt like mobile was

24:12 the most important computing platform that we had,

24:14 and I could not see a path to unseat either

24:19 Google or Apple sitting on top of that, controlling everything,

24:23 taking their toll booth money.

24:26 But the game has changed.

24:27 This is no longer the most important platform.

24:30 Mobile phones are important,

24:32 but there's also a lot of other form factors coming,

24:36 whether it's glasses or it's earpieces or whatever else have you.

24:41 It's all in play now, but the computing platforms themselves are in play

24:46 for the first time in probably 40 years, if you look at the desktop.

24:51 Linux has been around since '91.

24:54 It's not taken off or taken over on the desktop.

24:58 It's taken over everything else.

24:59 All the devices you have on your desk, your fridge,

25:02 your toaster, everything runs Linux, except for your computer.

25:07 Now, funnily enough, your Android is actually Linux,

25:12 but wrapped so sufficiently that you can't really recognize it.

25:16 But now there's an opening, and the opening is exactly as you say it.

25:21 If there's a piece of software you depended on that bound you to Windows,

25:27 it is completely within reach for you personally to start rewriting it.

25:32 And maybe you won't get to 100% coverage,

25:35 but this is the old joke about Microsoft Office.

25:38 "I only use 5%." Yeah, well, we all use a different 5%.

25:42 Well, what if we all just build our own 5%?

25:45 What if I just took the functionality that I need and just did that?

25:50 That is a completely different challenge and one

25:53 agents are incredibly capable of doing right now,

25:57 today, and I've done it a lot of times over.

25:59 So one of the amazing things I've found with the latest

26:03 agent engaged is I have become a polyglot programmer,

26:07 something I was absolutely not before.

26:10 I was a Ruby programmer—...

26:13 first and foremost, and then I dabbled a little bit in Bash—...

26:17 when I had to.

26:18 And in the last two months, I've written C++.

26:21 I've written three applications that have shipped in Omarchy Quattro.

26:26 I wrote a writing app.

26:29 I was using this app called Typora,

26:31 which is a very nice app-- Great, great app-...

26:33 which in itself is based on another app called iA Writer,

26:36 which was my real love on the Mac-...

26:39 for a clean, simple Markdown writing environment.

26:42 That's where I write all my essays.

26:45 And then I moved to Linux, and I couldn't get iA Writer,

26:48 so I moved over to this other tool called Typora.

26:50 It was a piece of shareware, and it had a lot of features I just didn't need.

26:56 And about six weeks, seven weeks ago, I thought, "Do you know what?

27:00 I only need 5% of Typora," which is already a basic app.

27:04 I literally told the agent to get going, that I wanted it written in C++ and Qt,

27:08 because that would fit well with the aesthetic

27:11 of what I was building with Quattro.

27:14 And in, I think, about 20 minutes, it had the first version.

27:17 I started using it, and it wasn't quite right.

27:20 Within two days, I'd given up Typora,

27:23 and I wrote and have written all of my essays since that moment in Omawrite.

27:30 Can I ask for your advice and your vision about something?

27:33 So what I find with agents, you can actually write something like a replacement

27:37 for Typora that is perfectly customized to you,

27:41 to your needs-- Correct with just one user in mind.

27:45 And then there's a version of that that is- starts

27:50 with one user in mind but expands to a larger audience.

27:53 So I've actually written a large number of software just for me,

27:56 and I find it easier to do and better to get my situation done.

28:01 For example, I have a video editor I've written,

28:04 which is very easy to do just for me.

28:06 I've never, unlike you, built something that's used by a large number of people,

28:11 and I find that a bit scary and intimidating.

28:14 So, like, at this, with this agentic age,

28:16 how do you take a step to actually build something that solves your problem,

28:21 which I think is a beautiful way to build,

28:24 but also expand it to actually useful for other people?

28:27 Like, is there advice you can give, and do you also see the problem of how

28:32 easy it is to just build a tool for yourself?

28:35 Yes, but it is simpler than you think.

28:37 By the time you're done building the tool for yourself,

28:40 you simply tell your agent to put it on GitHub.

28:43 You don't even have to do anything else.

28:45 It will figure out how to put that repo on GitHub.

28:48 It'll write a nice README.

28:49 It'll start using GitHub releases so that you can track things.

28:53 It'll actually be a better software maintainer than you could ever be-...

28:57 because it is far more patient, it is far more diligent in doing the drudgery

29:02 of managing open source software than you are.

29:05 And you can simply get on with the joy of using

29:10 the software that you built and evolving it as you see fit.

29:13 Now, this gets into the big argument we've

29:15 been having in open source for a while.

29:18 Is AI contributions good or bad for the maintainer?

29:23 There's a lot of maintainers right now who are quite upset about

29:27 the fact that they're suddenly getting a huge influx of pull requests,

29:33 of contributions from people who may not

29:37 be the best programmers or programmers at all-...

29:41 towards their software.

29:42 I look at that argument and go, "Are you kidding me?" Here is a vein

29:50 of free contributions that you can take or don't take,

29:57 but you're complaining about the fact that they're there?

30:00 It sounds straight out of- My steak is too juicy and my lobster too buttery.

30:05 What are you complaining about?

30:07 This is the most amazing thing.

30:09 This is what we proclamated for so long

30:11 was going to be the magic of open source,

30:14 that we could all contribute to it, but it was, of course, never true.

30:17 There was only a small subset of people who contribute to open source,

30:21 and that was the very highly skilled wizards.

30:23 So this is a bit of a reformation moment.

30:26 We had this disintermediation between us and the computer,

30:29 and it was this class of clerics and priests called programmers,

30:35 and suddenly they're being disintermediated here by was

30:39 it the 95 theses by Luther in 1500s, nailing them to the door?

30:45 And one of those was very clearly about the intermediation,

30:48 that you should have direct access to the higher intelligence.

30:51 Mm-hmm.

30:52 But don't you need the wizards to keep

30:53 a high bar of excellence in the code base?

30:56 100% you do, which is also something that's always been true about open source,

31:00 which is the second argument that really grinds me.

31:03 I've been running open source projects for 25 years.

31:06 I have literally looked at the output of thousands,

31:09 if not tens of thousands of programmers.

31:11 I feel like I have the statistical

31:14 basis to assert that most programmers, they suck.

31:21 And I don't mean that in the way it sounds.

31:24 That was why I paused for a hot second here.

31:26 I mean, they suck in the sense that they

31:28 don't write the code I want to have written.

31:31 They don't prepare their bug reports with all the relevant information.

31:35 They don't detail their pull requests with the why.

31:39 They don't bother to fill in needed code comments.

31:43 They don't double-check their work.

31:45 They don't write unit tests.

31:47 They don't do all of these things it takes to create good software.

31:50 But do you know who'll do all that stuff?

31:53 Agents, if you tell them to.

31:56 They're very diligent at following your instructions,

31:58 most of the time, sometimes to a fault.

32:00 But it actually means that if you take the median

32:03 programmer and their pull requests towards an average open source project,

32:09 they're already getting outclassed by agents.

32:11 I would rather get an agent-written pull request to one

32:15 of my projects than I'd get one written by a human,

32:18 and it's not just because the quality's better.

32:22 It's also because I feel a lot less bad if I just reject it.

32:27 You didn't even write it,

32:28 so I can simply look at what your agent wrote on your behalf and go, "Eh,

32:32 don't want it." That's always been true in open source, and in fact,

32:38 I think this is one of the problems with open source maintainers.

32:41 They have way too much ball of neuroticism

32:46 and anxiety that requires them to look upon any contribution

32:52 as an obligation on their part to do everything

32:56 to get that to land in the code base.

32:59 That's not true at all.

33:00 You can simply say, "No," or even better, you can say, "No,

33:04 thank you." And if you get into the mode of realizing

33:08 that your project is allowed to exist and evolve according to your vision,

33:14 according to your roadmap,

33:16 it becomes so much easier in the agentic age to decline the contributions you

33:21 don't want because you don't even inconvenience or hurt the feelings of a human.

33:27 It's just a clanker, and the clanker won't mind.

33:29 In fact, the producers of clankers, the labs,

33:33 love when you spend tokens in vain.

33:36 They get paid just the same.

33:39 So, in my opinion, this is the absolute best

33:42 time to ever have been an open source software maintainer.

33:46 Not only do we get this wealth of glorious

33:51 pull requests made by agents with all the boxes ticked,

33:55 we also get to tap into a creativity of people

33:58 who did not have access to contribute to a project before.

34:02 I've been running the Omarchy project now for a little over a year,

34:07 and in the last three months working on Quattro,

34:09 I have merged over 1,000 pull requests.

34:11 Quite a lot of those pull requests were

34:14 written by people who were not classical programmers,

34:17 or were programmers in other domains,

34:20 not Linux operating system or distribution development.

34:25 They were able to contribute their good ideas because of agents,

34:28 and I got to cherry-pick the very best of that-...

34:33 because it was there.

34:35 Otherwise, those ideas would just have lived inside their heads.

34:39 Again, Is- that is not the purpose of open source,

34:42 that we tap into the collective intelligence

34:44 and creativity of the whole goddamn planet,

34:48 and we channel that towards a commons where

34:52 we all benefit from the combined efforts of everyone.

34:56 Now, there's currently, I think, about 400 unmerged pull requests on Omarchy,

35:01 about double what it was a week ago.

35:03 So the acceleration is real, but so are the tools.

35:08 I'm not reviewing every pull request anymore.

35:10 I haven't been reviewing them for quite some time now.

35:13 I have the agents-...

35:15 review them for me-...

35:17 and then they will give me a summary about

35:19 whether something is ready for the decision, the human decision.

35:21 Should we merge or should we not merge?

35:25 I don't need to look at all

35:27 the pull requests that are either wrong, duplicated, bad.

35:30 An agent can sort that chaff away from me.

35:33 I just get to look at the pearls-...

35:36 the good ideas that are ready to go,

35:38 the bug fixes that the agent has validated in a VM on my behalf.

35:44 All the drudgery of replicating and managing

35:46 open source projects is evaporating at lightning speed,

35:50 and we are left with the, with the golden juicy parts,

35:57 the bone marrow of software development-...

36:00 deciding what should this thing do and where should it go.

36:03 Do you think the good ideas,

36:04 ultimately the kernel of the good ideas originated in the human mind,

36:08 so the individual contributors Like, can it be done by agents,

36:14 or is the PR ultimately requires a human idea?

36:18 Like how to improve Omarchy, all those incoming...

36:22 Or can it all just be a pool of agents?

36:24 That was the mode and the way I was thinking about

36:28 agents in the first agentic age from November 24 to February 28.

36:35 I thought all the ideas would originate with the humans.

36:38 They would tell their agents what to build, and off they went.

36:43 I don't think that's true anymore at all.

36:45 I have seen things you people wouldn't believe,

36:47 ideas coming out of models so great that it makes me humble-...

36:55 as a person who otherwise prides himself on having good ideas.

36:59 The agents are incredibly capable of creative thought,

37:04 and folks who are still stuck in the analysis that agents are parrots-...

37:11 just regurgitating the ideas that are already there are delusional about

37:17 the progress that's been made in the last six to nine months.

37:21 So some people listening to you right now will say

37:23 DHH is suffering from the old case of AI psychosis.

37:26 Can you steel man the case that you

37:30 are in fact experiencing in a state of delusion,

37:34 like One Flew Over the Cuckoo's Nest, and can you argue against it?

37:37 I'm in a state of delirium.

37:39 That's what the state I'm in.

37:41 Because I have been working with computers for 40 years,

37:47 and I've never seen the things that I've seen in just the last two months.

37:54 A whole career, a whole life dedicating to the love of computers

37:59 as my main focus for the working hours and then some,

38:05 suddenly completely upended, suddenly completely rewritten.

38:08 Of course I'm delirious.

38:11 Who looking at this reality is not?

38:14 If you are not delirious or at the very least very excited...

38:19 Actually, I guess that's not even true.

38:21 If you're not recognizing the gravity of the moment, that's the delusion.

38:27 That's the psychosis.

38:28 The psychosis is believing that the world is barely different.

38:31 We just have some electronic parrots reciting things to us.

38:37 Now, in my opinion, the reason this argument is actually brought up

38:41 is because people are not able to see the fruits of the progress.

38:45 That goes to your point earlier.

38:47 Where is the amazing software?

38:50 Why isn't American GDP running at 12% year over year?

38:55 First of all, give it a minute.

38:57 It hasn't even been a year, goddammit.

39:01 But second of all, I feel on my own account,

39:05 I have actually brought the pudding.

39:08 Omarchy Quattro launched on Friday.

39:10 It's been downloaded by tens of thousands of people.

39:14 And they like it.

39:15 They like it a lot.

39:17 We should mention that going to Perplexity here,

39:19 Omarchy is a highly opinionated Arch Linux-based desktop

39:22 distribution centered on the Hyprland Wayland tiling compositor.

39:26 It was created by David Heinemeier Hansson, DHH,

39:30 as a polished developer workstation set up with a cohesive

39:35 visual style and many everyday tools already configured.

39:40 Let me put it.

39:41 Omarchy is a beautiful, modern, and opinionated Linux system.

39:45 It's an alternative operating system to macOS and Windows,

39:49 and it's freaking amazing.

39:51 And the Quattro thing is you're pushing towards the age of agents.

39:55 That's right.

39:55 That's the latest version that we just dropped out, which is also funny.

39:58 The Omarchy project is only a year old and a little bit.

40:02 I started it last summer in between sessions at the 24 Hours of Le Mans-...

40:08 where I had watched one too many YouTube videos

40:11 on Linux Rising and got bitten by that bug.

40:15 Started working on this second iteration of my attempt

40:20 at creating a better Linux operating system.

40:22 The first iteration we talked about last time was called Omakub,

40:24 was built on top of an existing system called Ubuntu, and it was fine.

40:29 But the ambition I was able to pour into the new version, Omarchy,

40:34 because I started seven layers deeper down the stack,

40:37 was completely different, but it was still done in a pre-agentic world.

40:41 I wrote all those batch scripts by hand in the beginning,

40:47 and then I got to see the midway point where things changed over.

40:52 I did a couple of versions of Omarchy that had partial agent acceleration,

40:56 and then three months ago,

40:59 it was full throttle, 100%, everything is written by agents but steered by me.

41:04 And that just ended up being a very different experience and a very

41:10 different operating system because I was suddenly

41:13 granted a limitless ceiling on my ambition.

41:17 I could look at any feature in Windows, on Mac,

41:22 on other Linux systems and say, "I want that," and the agents would deliver.

41:28 So everything I wanted was suddenly within reach,

41:32 which meant that I got a little delirious.

41:36 Who would not get delirious if suddenly a genie pops out of the bottle and says,

41:41 "You can have whatever you want.

41:43 Every feature you've ever dreamed of in an operating system-...

41:47 I can deliver them to you, most of them in five minutes, a few in 20,

41:51 and if we really go hog wild, it's gonna take me two hours."- Mm-hmm.

41:54 Have you seen Requiem for a Dream, the movie?

41:57 So the, There's a drug-like effect here.

42:00 I find it personally— We'll talk about many levels of this, but it

42:05 is overwhelming how much power is placed in our hands so suddenly,

42:11 and it's very hard to know,

42:13 it's like with "The Hobbit" or something, what to do with it,

42:16 and I find myself truly overwhelmed with the multitasking,

42:20 On the verge of almost burnt out.

42:23 Super excited about so many things, but ultimately looking back,

42:28 you want to focus on one thing and really ship- I don't have that sense.

42:33 I don't have the sense of overwhelm.

42:34 I don't have the sense of dread.

42:37 I don't have the sense of distortion because I have a mission.

42:42 I'm going somewhere, and therefore,

42:44 I can channel all this new power towards-- Yeah, it's wonderful-...

42:49 a singular goal and outcome, creating the perfect computer.

42:52 And therefore, all my investigations with AI are focused towards that end.

43:00 Now, I also have a day job.

43:02 And we also use agents there,

43:04 and that's also targeted towards sort of specific outcomes.

43:07 But with the Omarchy experience,

43:09 I got to tap in straight in the back of my head and increase

43:18 the bandwidth between ideas arriving in my brain

43:22 and software emerging on the screen.

43:25 It was like going from dial-up-...

43:29 to fiber.

43:30 I think on your podcast, Elon talked about the actual human bandwidth-...

43:36 that Neuralink and other projects are trying to accelerate.

43:39 Like, the bandwidth we're communicating over right now

43:43 is quite low because we're limited by our cognition.

43:47 We're limited by the rate of speech.

43:50 You're not limited by those things when you're working with a swarm of agents.

43:55 Now, let me caveat this by saying, if I'd heard myself talk like this-...

44:03 nine months ago, I would probably have used the label AI, AI psychosis.

44:08 I think it would be fitting because nine months

44:11 ago when people were talking like this, they weren't shipping,

44:15 and I think that's the ultimate difference,

44:17 that there were people who saw this early.

44:19 I was not actually that early.

44:21 As we've talked about, I was rather skeptical,

44:23 and I was using AI in certain ways,

44:25 but I was not the first person who downloaded Claude Code.

44:28 I think Boris released that in end of February.

44:30 I don't think I installed Claude Code until September.

44:34 So there was about six months there where pioneers saw

44:39 these glimmers of what the future was gonna look like.

44:42 It wasn't there yet because the intelligence wasn't there.

44:44 The harnesses weren't there, but they could see the glimmers.

44:47 I couldn't.

44:48 Tobi, Tobi Lütke, the CEO of Shopify and now

44:51 my partner in crime on this Omarchy project in part,

44:55 he's gotten pilled with Omarchy as well.

44:58 He saw these things very early, and he tried to tell me, and I wasn't seeing it.

45:04 I was not seeing what he was seeing.

45:07 You were a hater.

45:09 Yes.

45:09 I was still on Earth.

45:11 And he was...

45:12 He had already boarded the rocket and was heading towards space.

45:16 I remember actually reading his memo to the company,

45:20 the AI memo, and thinking, "Ah, seems a little much." Like,

45:24 you're making a big deal out of something

45:27 that I cannot yet feel is a tangible thing.

45:32 Because where's the output of this?

45:33 I try using it.

45:34 It writes code I don't like.

45:36 It wants to interrupt me all the time.

45:38 Where are you getting this from?

45:39 Uh, so I get it.

45:42 I totally get it.

45:44 If you've not seen what the current quality of intelligence can produce,

45:51 of course you're gonna look at someone and go like,

45:53 "They sound a little nutty," because all your experience up

45:56 until this point would tell you that they are nutty.

45:58 Because if you look at the history of computer programming, for the last 40,

46:03 50 years, people have been promising that regular

46:07 people could write code just by talking.

46:09 We're gonna have fourth generation languages.

46:11 Lisp was once upon a time presented,

46:14 Smalltalk was presented as these environments that would

46:18 let regular people write their own applications,

46:20 and none of it came true in the sense beyond,

46:24 like, Microsoft Access databases or Excel spreadsheets.

46:28 That, those are end user programming environments,

46:30 but it's nothing like what we're able to do now.

46:33 And then, I mean, I'm not even talking about end user.

46:35 I'm talking as me, someone who's been programming for 25 years,

46:40 seeing the output and the acceleration and the quality and then shipping it,

46:45 and I think that's what's gonna change the entire conversation quite quickly.

46:49 The few laggards who are still

46:51 on the fence about whether this is actually useful,

46:54 who are still trapped in a meme from early '25 about mechanical parrots,

47:00 they're simply going to be overwhelmed

47:02 by the evidence that's about to flood over them.

47:05 So I hope to see, projects like Omarchy,

47:10 several of them that, like, really show agent first.

47:15 It's coming.

47:16 All of it is coming.

47:17 In the spaces, in the silly spaces that I already mentioned,

47:20 like video editing, all these kinds of apps that are like Typora-- Yes.

47:24 this kind of stuff.

47:26 It's gonna come, and I've already seen the early inklings of this.

47:30 One of the really neat features about Omarchy is that it has a far

47:37 more robust plugin system that allows you

47:39 to extend your operating system and rewrite it

47:42 really in terms of its user interface and its panel and its features by creating

47:48 your own plugins that can be cloned off of tools we ship in the box.

47:54 For example, say you want a different calendar.

47:57 Omarchy ships with a calendar.

47:58 You click the little clock.

48:00 It pops up a calendar, and it doesn't have iCal support, for example.

48:04 It doesn't consume your appointments.

48:07 There's about 17 implementations of that already, and in three days,

48:12 we had 330 plugins on the Omarchy plugin marketplace.

48:17 I have never seen growth like that with any project I've ever been involved in.

48:24 I've never seen participation that broadly.

48:26 I've never seen that many people be able to create software

48:29 that's meaningful for them- Also be usable for others so quickly.

48:35 All of it is driven by the fact that Omarchy ships a set of skills

48:40 that tells any agent you bring to it

48:42 how to create extensions to the operating system,

48:46 and therefore affording you the vision of the true malleable...

48:53 I was about to say agentic.

48:55 I hate that fucking word.

48:56 And the reason in part I hate that word is,

48:58 first of all, it's become marketing slob speak-...

49:00 at this point.

49:01 Like, it's just slapped onto everything.

49:04 I wish we had a different word that just meant AI doing stuff.

49:08 Vibe coding feels wrong, too.

49:10 It does, because vibe coding to me smells

49:12 exactly like script kiddies did in the early 2000s.

49:15 People applying PHP scripts they just

49:18 downloaded offline they don't understand anything of.

49:21 And while, I mean, that's true in the sense that much of the vibe coding,

49:25 including my own vibe coded projects, and as I just mentioned, I have several.

49:29 I've written things in C++.

49:31 I've written things in Rust.

49:33 Rust.

49:34 I hate Rust with a passion.

49:37 Rust to me is like pouring acid in my eyes when I have to look at the code.

49:42 Wait, wait, wait, wait.

49:42 Why do, why do you hate Rust?

49:44 It is, in my opinion, the ugliest programming language-...

49:46 that has been invented-...

49:47 in probably the last 40 years.

49:49 So it's the opposite of Rails and Ruby.

49:51 It's the opposite.

49:52 But-- You've written it-...

49:54 I've written applications in Rust because

49:56 the output of Rust is actually amazing.

49:59 The memory safety, the fact that it's a system language,

50:02 the fact that it's highly efficient and so forth is incredible.

50:06 So Rust, to me, can both be the most repugnant programming language

50:12 ever devised for human consumption

50:14 and a wonderful platform for agentic engineering.

50:19 These two truths can coexist easily in my head.

50:24 Do you think that there can be a point

50:26 at which we can just call agentic engineering programming?

50:28 Can we just change what programming means?

50:31 'Cause who's actually going to be programming

50:34 the old-school way in a few months?

50:37 I don't think we should reuse the term because for me,

50:39 programming implies an understanding of certain primitives,

50:45 loops, conditions, variables.

50:47 All the constructs of programming

50:50 languages are what constitutes programming itself, not the creation of programs.

50:55 Because you could say in the pre-agentic era, well, your CEO is programming.

51:00 He's hiring a bunch of programmers, he's telling them what to do,

51:03 and out comes software that he can sell.

51:05 Eh, I don't think most people would call that CEO a programmer,

51:08 and I don't think we should call the vibe coder a programmer either.

51:12 And vibe coding, if we define it here,

51:14 is you tell an agent to build software for you.

51:17 You do not look at the implementation.

51:19 That, to me, is what separates vibe coding from programming or, let's say,

51:25 agent-accelerated development.

51:26 But don't you think, so to push back a little bit,

51:29 don't you think a programmer, old-school programmer,

51:32 you, doing agentic engineering is a different kind

51:35 of agentic engineering than a non-programmer doing agentic engineering?

51:39 What I mean is, if you know what a for loop is,

51:43 if you know what function programming is,

51:45 if you know some of the basic good principles of software engineering,

51:48 the way you do natural language-based agentic

51:51 engineering will be different and more systematic,

51:54 and this, the scale and the variety of things you

51:58 can actually build is much bigger than a vibe coder.

52:02 I'm just gonna extend the pro case here

52:04 a little bit just for the sake of the argument.

52:05 Right.

52:07 I actually think for a while it was to my deficit to know as much as I

52:16 know about programming because I was instructing the agents

52:20 to do things as I prescribed them to do, and they were very good at that.

52:24 And in the first agentic moment, let's not call it an era,

52:28 it was something that lasted three months.

52:31 Moment.

52:31 In the first agentic moment, that felt very productive.

52:35 I could get more productivity out of telling the agents

52:38 what to do in the way I would've done it,

52:41 and I used all of my experience as a programmer to do just that.

52:47 Then I was a little late on the next moment,

52:51 and the next moment allowed me, allowed anyone, to describe outcomes-...

52:56 to describe problems to the agents and get better

53:01 solutions than if you had a programmer prescribe the path.

53:06 See, okay, so we're having fun arguing here.

53:08 Well so are you suggesting that there's problems

53:11 for which programmers are worse than non-programmers at agentic engineering?

53:16 100%.

53:16 And the reason I say that, let's define it here.

53:20 The reason I say that is there's a lot

53:23 of programmers who are not very good product managers.

53:27 Software is product management.

53:29 What should it do?

53:30 Who should it do it for?

53:32 How should it do it?

53:33 How should it look?

53:34 What are our priorities?

53:35 What do we start with first?

53:37 What does version one include?

53:39 All of those skills are not

53:43 easily or equally distributed across all programmers.

53:46 And in the agentic era where you are letting an agent,

53:51 an AI, do the implementation, these are the skills you need.

53:54 And I happen to sit in both camps, and this is why I love this moment.

53:58 I still get to live a little bit in the old world where I look

54:01 at the full implementation because there are domains

54:04 and areas where I feel that still brings value,

54:07 and then I also live in the future.

54:10 And as I said, when I created Omawrite,

54:12 I've not looked at a single line of that C++.

54:16 I actually made it a point to myself that I was gonna treat it as an experiment.

54:20 This is 100% a black box.

54:22 I will treat it as though I was any other

54:25 user who had opinions about how their digital typewriter should work,

54:29 and therefore felt like it was a fair experiment as if a writer of any

54:35 other sort who had opinions about how software

54:37 should be done and how it should work.

54:39 As you mentioned, video editors who know exactly where they don't

54:43 like what Adobe Premiere's doing and how it could be better,

54:46 but don't have the capacities to improve it-...

54:49 how that could be.

54:51 The pushback, though, I think what programmers do well,

54:53 good programmers do well,

54:55 is systematic design, like think in terms of systems-- Correct-...

54:59 and rigor.

55:00 Yes.

55:01 And I think still, and probably for a long time to come,

55:04 even natural language-based control of agentic systems

55:07 will require a certain kind of rigor.

55:10 Not overly structured rigor,

55:13 but really explain very specifically the goals of the system,

55:18 how to do the verification, what kind of security testing.

55:22 You wanna do all those kinds of things.

55:24 Although then to push back on that, the system will...

55:27 should be smart longer term.

55:29 I would've said the same thing six months ago.

55:30 Yeah, so this has changed.

55:31 And I think the same thing would've been true six months ago.

55:33 What I have found since, especially even just the last few weeks

55:38 of the agentically accelerated development of Mach-E,

55:41 is that more often than not,

55:45 I have the humility to recognize that the agent knows best.

55:49 You can legitimately say, "Make sure it's secure."- Correct.

55:53 And it would know more about what that entails.

55:57 Now, you're laughing-- What a crazy world-...

56:00 but it's true.

56:00 And in fact-- Oh my God-...

56:02 a good parallel to this is the AGENTS MD/CLAUDE MD files.

56:07 There was a hot moment where it was all the rage about micro- optimizing that.

56:12 Oh, you tell your agent this-...

56:14 you tell your agent that, and you ended up

56:16 with this huge file of instructions for your agent.

56:19 One of the things that Boris, working on Claude Code,

56:23 shared in an interview recently about Opus 5 was

56:27 that the system prompt that they ship for Opus 5, and presumably also Fable,

56:32 shrunk by 80% because the agent not only needed far less human instruction,

56:39 it was actually being damaged by overly prescriptive humans.

56:45 And any programmer who's had a pointy-haired

56:48 boss knows exactly what that is like.

56:51 When the boss walks into the room, doesn't know anything,

56:55 starts telling you how to program, how to code.

56:58 What do you do?

56:59 You sulk.

57:00 You write shittier code if you're mandated

57:03 to do things that are against your better judgment.

57:06 Why would an agent not be the same?

57:08 Actually, to push back against myself,

57:09 I do find the actual skill I'm currently developing,

57:11 which I along those lines is probably what everybody's trying to develop,

57:15 is not to overspecify things, is to get out of the system's way.

57:20 Of course.

57:20 Is to specify enough to provide a kinda high-level vision and guidance.

57:25 It's even better because you don't even need that.

57:28 The fundamental insight of modern software development

57:31 came from the agile software development movement.

57:33 In the late '90s, early 2000s, a group of smart,

57:38 honest, and brave people came together and said,

57:40 "The way we've been trying to do software development for the last

57:44 40 to 50 years has not worked, will not work.

57:48 We've been trying to specify upfront what software should look like.

57:52 We thought that we could ask humans what they wanted, write it all down,

57:57 apply our systems thinking and our rigor to a spec sheet,

58:02 hand that spec sheet over to a group of programmers who would

58:06 then implement just that, and everyone would be happy." Well, guess what?

58:09 No one was happy.

58:10 Because no one knows what they want until they receive it.

58:13 You don't know what a program should do until you play with it.

58:17 So in the agentic age,

58:19 you should resist the temptation to be overly specific upfront.

58:22 Be as vague as you can to manifest something, then interact with the something.

58:29 The way you arrive at good software is you write a little bit of software,

58:33 and then you try to use it.

58:34 It is in the process of using software that you discover what you really want.

58:38 It's where you discover what's important and what's not important.

58:42 And agents are incredible at allowing you to bumble into this experience

58:48 of realizing what you want because you don't know what you want.

58:52 Mm-hmm.

58:52 And I've actually done something similar to what you suggested,

58:55 which is implement different designs and totally different implementations,

58:59 and then I create HTML pages for myself, PHP pages,

59:03 where I vote on what I like more and so on.

59:06 And it does this iterative thing, and it's very nice and fun.

59:09 The interaction with the agentic system,

59:10 if you create a nice interface for yourself, it can be super fun,

59:14 and you're focusing on the design part

59:15 and the most fun parts of the design part.

59:18 Correct.

59:18 And this is what humans are really good at.

59:20 Yeah, taste.

59:21 Differential evaluation.

59:22 You give me three options, I pick one of them.

59:26 Now, humans actually fall off a cliff if you give them 22 options.

59:29 That's the paradox of choice.

59:31 But you give them three options,

59:32 they very quickly within a split second will tell you what they like.

59:36 There are all these ways that exploit the fact that humans

59:41 will make these snap judgments that are actually pre-intellectual, right?

59:44 Like, they're coming from the gut, and then they arrive up in the, in the brain,

59:49 and the brain tries to rationalize why the gut said what it said.

59:54 And if you're willing to let go of some

59:59 of that intellectual pre-processing rationalization

1:00:01 and simply let your gut drive, the agentic age is a revelation.

1:00:06 So this is a fascinating thing to ask you because

1:00:09 you have sort of famously for a long time talked about,

1:00:13 like, detailed chiseling of beautiful Rails Ruby code,

1:00:18 and now you have switched in a matter of months to not doing that.

1:00:23 So, like, what is your current setup look like?

1:00:26 'Cause you're, I assume, are still looking for the beauty, for the chiseling,

1:00:30 for the crafting, but you're just crafting in a different medium.

1:00:35 And at different levels of abstraction.

1:00:37 When I'm working in Ruby code,

1:00:38 and we have a lot of Ruby code because it's in our entire business,

1:00:41 and I'm asking agents to make changes to that code, I still sweat the details.

1:00:48 What I'm coming to realize is that the economic

1:00:53 payoff of that sweat is diminishing rapidly.

1:00:57 The reason why I, for 25 years, was sweating every line of code so judiciously

1:01:05 was because I knew the payoff of keeping

1:01:09 an architecture coherent and malleable was software that could

1:01:14 change and evolve- quickly with a small team, and not exorbitant cost,

1:01:20 and not introducing a bunch of bugs when you change one thing over the other.

1:01:24 That was the driving economic argument for why you should write beautiful code,

1:01:29 'cause beautiful code is easier to understand,

1:01:31 it is simpler, it is more malleable.

1:01:34 That was premised on humans doing the modifications.

1:01:40 I think it is an open question to which degree this still matters.

1:01:44 Now, it does matter, and the reason I say that, at least for the moment,

1:01:50 is that tokens are still scarce.

1:01:52 At this moment in time, we are all token limited.

1:01:58 Well, not all, but anyone who doesn't have endless budgets are token limited.

1:02:03 So therefore, there is great payoff to writing systems that agents have

1:02:08 an easier time dealing with and evolving

1:02:11 without having to relearn the entire context.

1:02:13 Just like humans, if they can make iterations

1:02:16 and changes to the code base without wrecking the architecture,

1:02:20 they can make the next change just as cheaply as the last one.

1:02:24 And this is the classic ball of mud where you

1:02:27 end up with a system that's a ball of mud-...

1:02:32 because it's just put together in a way where

1:02:35 nothing is connected and it's just a real mess, right?

1:02:37 I've seen that with agents, and I've seen it in our own code bases where,

1:02:41 again, the first PR is, like, mediocre of quality,

1:02:44 and then if you add another PR on top of that, and then five more down the line,

1:02:48 it's not very good, right?

1:02:49 So there's still a payoff to that, but that payoff

1:02:52 is premised on our current moment.

1:02:54 I am now able to...

1:02:55 And this is, by the way, where the AI psychosis really comes in, when you try

1:02:59 to extrapolate what nine months from now it's gonna look like.

1:03:02 What two years from now it's gonna look like.

1:03:05 But as an intellectual experiment,

1:03:07 I think the first computer I used was a Commodore 64.

1:03:10 Me too.

1:03:10 Cool.

1:03:11 It had one megahertz CPU and 64K of memory.

1:03:16 Everyone who wrote software for that machine internalized certain constraints.

1:03:21 Well, all the constraints.

1:03:22 That machine was nothing but constraints.

1:03:25 If you put that programmer in a time machine and teleported him

1:03:30 or her to 2026 and gave them a modern computer today and said,

1:03:35 "Now write me a piece of software," they'd be lost for a little

1:03:38 bit because all their techniques and all their heuristics would simply be wrong.

1:03:44 Or not wrong, because efficient software is still beautiful.

1:03:47 They would be out of date with the value that they could create.

1:03:51 The amount of optimization you have to apply

1:03:53 for a one megahertz computer to produce

1:03:55 a video game is just very different from what you have to apply today.

1:03:59 Just imagine that guy that was programming back then on a Commodore.

1:04:02 Just re- remember all those, like what programming felt like and what it meant,

1:04:09 highly constrained resources-...

1:04:10 how slow everything is.

1:04:11 Beautiful in many ways.

1:04:12 Beautiful.

1:04:13 It's chiseling.

1:04:14 But-- It--...

1:04:15 imagine that person coming now.

1:04:18 It's romantic too, though.

1:04:19 Let's also imagine that.

1:04:20 This is one of the things,

1:04:21 whenever we bemoan modern car culture, for example, right?

1:04:25 And then we think like, "Oh, wasn't it better when we all had

1:04:28 the horse-drawn carriages and so on?" No, it wasn't.

1:04:31 Do you know what New York smelled like when we used horses for transportation?

1:04:35 It was an open sewer.

1:04:37 They just shat everywhere.

1:04:38 Yeah, but have you played Red Dead Redemption?

1:04:40 You know how cool those cowboys were-- That is the way--...

1:04:43 on the horse?

1:04:43 to experience the nostalgic past,

1:04:45 when you don't get the all the furry impulses and senses.

1:04:50 So basically programming, old school programming by hand is kinda like

1:04:54 the cowboy culture that we romanticize and make movies about.

1:04:58 It's already happening, the romanticization of handwritten code.

1:05:01 And I have some of it because it was a very romantic era.

1:05:05 I'm grateful to have been alive

1:05:09 for 20 years of economically valuable handwritten code.

1:05:13 That was a, that was a good time.

1:05:17 Where handwritten beautiful code was also valuable.

1:05:20 Correct.

1:05:21 Because handwritten beautiful code exists as much today as it did yesterday.

1:05:26 There are people who willfully write new video games for the Commodore 64,

1:05:32 or for the second Mega Drive, or for other vintage consoles.

1:05:35 In fact, one of my favorite video games of late is ModRetro's.

1:05:40 Do you know Palmer Luckey's-...

1:05:43 side geschäft, which is recreating old consoles.

1:05:46 So he's recreated the Game Boy.

1:05:50 Nice.

1:05:51 It's amazing.

1:05:52 Look it up.

1:05:53 It's really cool.

1:05:54 But what's even cooler than the fact

1:05:56 that he recreated this hardware and the original chunkiness,

1:05:59 but brought it up to date with sort of modern screens

1:06:02 and so on, now they just put out the Nintendo 64 too.

1:06:06 Incredible.

1:06:06 Super cool.

1:06:07 Wild attention to detail, input lag, and all the other things.

1:06:12 But what's also very cool, that one, I have to wave.

1:06:15 That one right there, and hover over it,

1:06:17 'cause then you get to see what I'm talking about.

1:06:19 That is the Chromatic-...

1:06:21 Tetris re-implementation.

1:06:23 Wow.

1:06:23 I played a lot of Tetris on the original Game Boy.

1:06:26 Probably one of my top three favorite games of all time.

1:06:30 And they rewrote it with one change.

1:06:34 If you hit up, the brick slams.

1:06:39 And that speeds up the game by about 400%.

1:06:43 It is an incredible game, and it's a new Game Boy game.

1:06:47 I think the Game Boy came out in- '89 or '88.

1:06:51 It's a very old machine.

1:06:52 The fact that there's still people writing new old games is awesome,

1:06:56 and the reason they do it in part,

1:06:58 now here they're doing it because they want Tetris,

1:07:00 but there are people who do it just for the love of the constraints,

1:07:03 for the love of the romantic notion of writing that.

1:07:07 Just like there are people still riding around on horses,

1:07:10 not to get places fast, but because they like the mode of transportation where

1:07:16 it's a literal biological being below you propelling you forward.

1:07:22 And also, we should say that that beautiful code is great training data.

1:07:25 So in some sense, the beauty that we've built over a few decades-- Yes-...

1:07:31 continues.

1:07:32 We gave birth to this moment.

1:07:35 What a privilege, and I feel that very personally because I...

1:07:39 Almost all of the code I've ever written is public code.

1:07:43 The vast majority of my career has been dedicated to writing open source code,

1:07:47 and therefore, some of that code,

1:07:49 some of those beautiful lines are in the training set.

1:07:53 And in fact, I've heard people do this when they write Ruby code.

1:07:57 They ask the agent, "Write it like

1:08:00 DHH would." And they're pleased with the result,

1:08:02 and that does warm my heart a little bit,

1:08:04 that I helped give birth to this moment in my tiny little small part.

1:08:08 And anyone else who contributed open source code or any code

1:08:11 at all that the agents have had access to over the last,

1:08:14 well, entire time programming has been around as a discipline,

1:08:19 is now part of these new agent overlords.

1:08:22 So you don't have a little bit of sadness that writing

1:08:26 beautiful code by hand is now less and less useful?

1:08:29 Am I sad that if I had to get a job tomorrow,

1:08:33 it's unlikely that I could apply my skills in such a way

1:08:35 that I would be paid to write these manual lines of code?

1:08:40 No.

1:08:42 I did it for 25 years.

1:08:43 Like, I wanna see something new here,

1:08:45 especially if we're all gonna live forever, which I hope we don't.

1:08:48 Maybe we'll get to that.

1:08:49 But, like, I've done that.

1:08:50 That's enough.

1:08:51 That's fine.

1:08:52 I don't feel any more nostalgic about that than I do about the fact that I

1:08:56 don't have to spend my time in the field with a hoe or at the assembly line.

1:09:01 Well, hold on a second.

1:09:03 You are one of the best programmers

1:09:05 in the world at writing handcrafted beautiful code,

1:09:08 appreciators of beauty with great taste.

1:09:11 Correct.

1:09:12 That has been replaced.

1:09:14 I say humbly.

1:09:15 Yeah.

1:09:15 That has been replaced.

1:09:16 Like, you had now...

1:09:18 You're just fast evolving, and you're able to...

1:09:21 One of your really great qualities is you're

1:09:24 able to change your mind and evolve very quickly.

1:09:26 But, like, you had to kill that other person.

1:09:30 I am grateful for every moment that brought me to where I am right now,

1:09:33 and therefore, I have also tried, and maybe this did take some practice.

1:09:39 I will concede that.

1:09:40 Maybe I wasn't always like this, but stoic philosophy, amor fati,

1:09:46 loving your fate, is an incredibly liberating way to live.

1:09:52 And you could attack that and say, "Well, that's a point of privilege." True,

1:09:57 but also I had the same directional point

1:10:01 when I didn't have all this, quote unquote,

1:10:04 "privilege." So I think there's a way to live

1:10:07 and interact with the world where you accept

1:10:09 the things you can change and the things you

1:10:12 cannot change and fall in love with all of it.

1:10:15 Fall in love with the fact that the world today has jumped

1:10:19 two decades forward into technology industry from where it was last year.

1:10:24 So can we talk about the general anxiety that programmers feel,

1:10:28 going through the same transformation?

1:10:31 They've maybe gone to university majored in computer science,

1:10:37 dreamed of being programmers and building stuff,

1:10:40 high salary, and now everything's changing,

1:10:43 and there's deep anxiety about, "What do I do when...

1:10:46 with my life?" Can you empathize with that, and what

1:10:49 the hell are they supposed to do?

1:10:51 Hugely.

1:10:52 Because just because I have this disposition,

1:10:54 I'm keenly aware that that's not evenly distributed,

1:10:58 that not everyone sits with this position and this optimism for the future.

1:11:03 But I do think I wanna separate things a little bit.

1:11:08 I think you're gonna have a hard time coping

1:11:10 with the new reality if the only thing you loved about

1:11:15 programming was the mechanical bits of putting the right logical

1:11:21 constructs together to produce something other people told you to produce.

1:11:28 Because that mechanical process is under threat.

1:11:33 If you are, as you just mentioned, excited about building things,

1:11:38 I don't think you're under threat at all.

1:11:40 In fact, I think there's a great argument for us

1:11:43 needing far more builders than what we have now,

1:11:47 and if you look at the employment stats, it's fuzzy.

1:11:53 It's not clear what's gonna happen at all.

1:11:55 Some stats actually show an increase in openings because the advent of AI is

1:12:02 so dramatically lowering the price of programs

1:12:04 that people want a lot more programs.

1:12:07 This is the classic, the Jevons paradox,

1:12:10 that says when the price of something goes down,

1:12:12 there's gonna be more demand for it.

1:12:14 And this is the example of the ATMs too.

1:12:17 When the ATM originally came,

1:12:19 a lot of bank tellers were very afraid for their job because

1:12:23 suddenly there was a machine that could dispense money from people's account.

1:12:27 Well, what the ATMs did was lower the price of a branch,

1:12:31 and suddenly banks could afford to open a lot more branches,

1:12:34 and we ended up with more bank tellers than we did before.

1:12:37 Now, none of this is guaranteed.

1:12:38 There are also moments where things do change.

1:12:41 The vast majority of people worked in the fields up until the late 1800s,

1:12:47 and then suddenly we got mechanized- harvesting tools,

1:12:51 and we did not need human labor with a hoe and an ox out there.

1:12:56 We had mechanical tools to do that for us.

1:13:00 In that moment, are there people who were threatened

1:13:03 that their job and their livelihood was at stake?

1:13:06 Yes.

1:13:06 Was that a reasonable thing to feel threatened about?

1:13:10 Yes.

1:13:10 The Luddites smashing the weaving machines-...

1:13:13 in England had the same thing.

1:13:15 And I think maybe that's the better

1:13:17 parallel than people working the fields because

1:13:19 these were actually highly skilled professionals doing

1:13:21 a job that they liked on rather favorable conditions,

1:13:25 not slaving outside in the sun all day.

1:13:28 But where we live now,

1:13:30 would we still like for clothing to be this heavily constrained resource?

1:13:35 Like, what if this T-shirt was $400?

1:13:39 I mean, okay, then I'd have two.

1:13:42 Well, my wardrobe probably looks like I have two.

1:13:45 But I think the general progress

1:13:50 of mankind depends on productivity improvements.

1:13:54 And what are productivity improvements?

1:13:55 This is very important because in the abstract,

1:13:58 I think everyone says, "Yay, productivity.

1:14:01 That's good." What did you think, this was all just vibes?

1:14:05 No.

1:14:05 Productivity means fewer people to do the same number or the same job.

1:14:09 Now, the amount of job you want done may increase,

1:14:12 and therefore you get more people, but it also may not.

1:14:16 There may be pockets where a company just

1:14:19 needs a certain set of fixed tasks done,

1:14:22 and suddenly they can do them with a 10th the number of people.

1:14:26 This, by the way, could...

1:14:28 is tragic and difficult in the moment for the individual being laid off.

1:14:32 It's also amazing for the economy at large.

1:14:34 Suddenly you've freed up these resources who

1:14:37 can now go do more productive things.

1:14:38 This is how the whole economy evolves and improves

1:14:42 and how we get growth, and growth is good.

1:14:45 I think this is the other thing to make a strong

1:14:48 defense for, that growth in general is a good thing.

1:14:53 We do not want to wind the clock back to 1920 or 1950.

1:14:58 We should be excited about the fact that we've

1:15:02 discovered new technology that allows us to do vastly more.

1:15:06 And while some of the tasks that may be taken

1:15:10 over by AI were tasks that humans were excited to do,

1:15:13 and they're just no longer economically viable,

1:15:15 a lot of the tasks were just drudgery.

1:15:18 This is certainly true of my programming.

1:15:21 There were moments of programming I thought were amazing.

1:15:24 These were the moments that produced that state

1:15:27 of flow we talked about last time, and they were ecstatic.

1:15:32 How often did they happen out of the course of a year?

1:15:36 If you took your average programmer, how much time out of 2,000 hours at the job

1:15:42 did that person spend in a flow state?

1:15:47 100?

1:15:48 200?

1:15:48 25?

1:15:49 I think all those answers could be plausible for someone.

1:15:54 And taking that drudgery, handing it over to machines,

1:15:58 that is the history of civilization.

1:16:03 Yeah, I mean, the hours spent debugging.

1:16:06 So programming has so many painful, painful hours in it.

1:16:10 And now they're gone.

1:16:12 For the...

1:16:12 I mean, for the most part, it's just fun.

1:16:16 A lot of it is gone in a lot of domains,

1:16:18 and the rest of it may be gone very soon.

1:16:20 So that's where some of the romanticism also comes in...

1:16:25 because pain in the moment is difficult.

1:16:28 It's hard.

1:16:29 It's annoying.

1:16:30 Pain in the rear view mirror is accomplishment,

1:16:34 proud, learning, steps forward, all the things, right?

1:16:38 So we encompass all of this as part of the human experience,

1:16:42 that we hate pain in the moment, but we cherish it when it's five

1:16:47 minutes past and we don't remember the hardship.

1:16:50 We just remember the progress.

1:16:52 So certainly at the societal level,

1:16:54 at the macroeconomic level, growth is exciting.

1:16:56 But the individual level,

1:16:57 there's going to be a lot of anxiety potentially a lot of suffering.

1:17:01 So by way of advice,

1:17:03 if you're like a young DHH or young developer programmer now,

1:17:08 what would you advise they do?

1:17:12 Don't try to anticipate anything.

1:17:13 You will literally go crazy.

1:17:15 Because even the smartest brains in the business cannot anticipate

1:17:19 what two model hops from here is going to look like.

1:17:24 It's an absolute waste of time,

1:17:26 and you will develop an AI psychosis trying to deduce

1:17:29 what two years from now is gonna look like.

1:17:32 Focus on right now, and right now is

1:17:33 the most incredible time to be into computers.

1:17:38 You can make them do the most amazing things if you lean in.

1:17:42 If you maybe for a hot moment force yourself

1:17:48 to learn where the state of the art is,

1:17:52 I double dog dare you not to get excited about what's possible.

1:17:57 And would you recommend building publicly?

1:17:59 I think it's optional actually to build publicly.

1:18:02 I think it's nice to do because then you get to be part of a community,

1:18:06 and there's some camaraderie in open source.

1:18:08 Not some camaraderie.

1:18:10 There's an enormous amount of camaraderie.

1:18:12 Open source is an incredible source of community and camaraderie,

1:18:18 and it's also a great way to stem some of that personal existential dread.

1:18:23 If you're amongst other humans,

1:18:25 you can get infected with good forms of mind viruses,

1:18:29 exciting forms of mind viruses,

1:18:31 the ones that tell you that it's all gonna be okay,

1:18:34 and the future looks bright, and if we build a bunch of things together,

1:18:39 we can do things none of us ever dreamed we could do.

1:18:42 I was just talking to Ryan Hughes,

1:18:45 who's been my main partner on Omarchy almost since day one,

1:18:49 and we were literally talking about yesterday

1:18:51 how when We started on this journey together.

1:18:55 Our ambitions were very modest.

1:18:56 We're just like, "Can we get this distro to not crash at 3:00 AM

1:19:01 because of an Arch package that just got pushed out outside of our control?

1:19:05 That would be amazing.

1:19:06 Then it'd be great." Now we're leaning back and thinking,

1:19:09 "How can we take over the world?

1:19:12 How can we make Omarchy go to Mars?

1:19:16 How can we do everything we've ever dreamed of?" And that level

1:19:22 of ambition inflation is partly because things have gotten better with agents,

1:19:27 but hugely also because we've surrounded ourselves

1:19:29 with others excited about the same journey.

1:19:33 And if you're just sitting in your little

1:19:37 isolated cave worrying about the future, yeah, you're gonna go a little nuts.

1:19:42 I think if we learned anything during COVID is that people will go

1:19:47 nuts sitting in their little cave inside

1:19:50 by themselves worrying about what's gonna happen.

1:19:52 That is indistinguishable from depression,

1:19:55 ruminating endlessly about things you can't control.

1:20:00 I'm sorry to channel some Jensen Huang here, but that's just loser talk.

1:20:07 You don't have to be a loser.

1:20:10 You can choose to lean in and-...

1:20:13 win.

1:20:13 And win is very broadly defined as learning more,

1:20:17 making more, contributing more, being part of more.

1:20:23 And at the end of the day, what are your choices?

1:20:26 You don't have a choice, mate.

1:20:29 The future's coming whether you like it or not,

1:20:31 so you might as well choose to be excited about it.

1:20:34 But I mean, we should emphasize the fact

1:20:36 that planning is nearly impossible with AI.

1:20:38 I mean, most of us plan our life a little bit.

1:20:42 When we're younger, you have, like, hopes and dreams.

1:20:44 There's a reason you go to college.

1:20:46 There's a plan underlying that.

1:20:48 I agree with that.

1:20:48 But there-- I'll concede that.

1:20:50 it is kinda-- This is different.

1:20:52 nuts that it's impossible to really plan.

1:20:54 Because, like, you really don't know, if there will even be a Claude Code,

1:21:00 a Codex, and Cursor six months from now.

1:21:04 Maybe it'll be you'll be talking to WhatsApp.

1:21:06 Maybe it'll be all open call type.

1:21:08 Maybe it'll just all be voice.

1:21:10 Maybe Omarchy will take over the world and it'll be voice.

1:21:13 Maybe there won't be operating systems anymore.

1:21:15 You'll just...

1:21:15 There'll be this glowing orb that just you just talk to because

1:21:19 if natural language is enough to accomplish all tasks in life,

1:21:23 like, what is life about?

1:21:25 Like, what are the major tasks you need to accomplish in life?

1:21:28 If you can just do that with voice, and it writes all the programs for you,

1:21:32 and all is systematized, and what, what are the jobs needed then?

1:21:36 Maybe it's more about the service industry.

1:21:38 Maybe there won't be programmers, like agentic engineers at all.

1:21:42 And so yes, your advice is wise to focus on the moment,

1:21:47 but that doesn't mean that it's just a crazy rollercoaster ride,

1:21:51 'cause we don't know what this December or what this January will look like.

1:21:55 We don't know, and therefore,

1:21:57 we should choose to have faith that it's gonna pan out.

1:22:00 That is my best advice towards fighting anxiety,

1:22:05 is that you have a choice, to a large degree, of how to lean.

1:22:13 Do you wanna lean into PDoom?

1:22:15 PDoom is gonna happen whether you lean into it or not.

1:22:21 You don't have control.

1:22:23 But there's also actual daily schedule of how much learning are you doing,

1:22:26 how much building are you doing, how much uh, diversification you're doing.

1:22:31 And for example, I traveled across rural China recently,

1:22:34 and I did it quite deliberately because you're talking to people in China,

1:22:40 in, in rural anywhere, where time moves much slower,

1:22:44 where there's no discussions about agentic engineering.

1:22:48 And the ti- and I started by going on the, on the Great Wall of China.

1:22:54 That's timeless.

1:22:55 That spans centuries.

1:22:56 And, you know, there was, like, this kind of tension of FOMO,

1:23:00 like maybe I'm missing out on something.

1:23:02 But then I very deliberately wanted to find the timelessness of human existence,

1:23:07 really plug into the fact that, okay,

1:23:10 it always feels like everything's changing uh,

1:23:14 throughout human history but, you know,

1:23:16 the universals are still true, of what matters in life.

1:23:19 And I just wanted to make su- keep connected to that.

1:23:22 You're not gonna miss anything.

1:23:24 This is the part that actually grinded

1:23:26 my gears in the early phases of this transition.

1:23:30 Everyone was so up in arms that if you didn't...

1:23:34 It's loops now.

1:23:35 Oh, no, no, we're done with loops.

1:23:37 It's graphs now.

1:23:38 Oh, no, no, we're done with that.

1:23:39 It's, it harnesses this, right?

1:23:41 They're, they're constantly churning through the frontier,

1:23:43 which in one way is actually very exciting.

1:23:46 This is what happens when a new field opens up.

1:23:48 It's like a new portal, and it's all spilling out.

1:23:52 That's exciting.

1:23:52 But it also means that if I had just been backpacking for the last year,

1:23:58 hadn't touched a computer, hadn't witnessed this agentic moment,

1:24:02 and I just showed up yesterday, do you know what?

1:24:05 I would've been caught up in two weeks.

1:24:08 There's not any accumulation, which is in some ways a great relief.

1:24:13 If you missed the past year, you can catch up to the frontier in two weeks.

1:24:19 If you're a programmer who was out hiking

1:24:21 the Himalayas for a year and you come back, you can catch up in two weeks.

1:24:25 And this is actually the great credit to the progress we're experiencing.

1:24:29 We have so many experiments running simultaneously right now

1:24:33 that are constantly and ruthlessly sorting what works, what doesn't work.

1:24:39 You don't have to remember or even be part of that entire journey.

1:24:42 You can just show up for the results.

1:24:45 But the key step there, when you come back from the backpacking journey,

1:24:49 is to be willing to become a totally new,

1:24:52 different human because- Things are changing so fast.

1:24:56 That's right.

1:24:57 I mean, literally, if you left for the backpacking journey-...

1:24:59 in October last year and came back in April or May-- You

1:25:04 might as well have been in the cryo chamber for 100 years-...

1:25:07 it's like, "What, what do you mean?

1:25:08 We're not programming anymore?" This, it's a it's a huge transformation.

1:25:12 It's startling, and I actually...

1:25:13 I mean, I'm putting some of this on slightly for effect.

1:25:17 I also recognize that it's okay to grieve for a hot moment.

1:25:22 That's not my disposition.

1:25:23 It's not how I do it, but I accept that that's a time-tested

1:25:29 mechanism for coping with a sense of loss.

1:25:35 And it's okay to have a sense of loss that the world that you

1:25:38 knew and maybe you were very fond of is no longer the same.

1:25:43 I maybe I should say out loud and explicitly that I

1:25:45 do have that, the Native American with a tear rolling down my...

1:25:49 Like, there's a sadness.

1:25:51 It's a goodbye.

1:25:52 It's a goodbye to the old world of programming.

1:25:55 That's, it's so fascinating to me.

1:25:57 I've spent so many years-- But it all amounted to this moment.

1:26:00 That is, to me, what makes the entire journey so meaningful.

1:26:05 AI did not arrive from the sky.

1:26:07 It was not alien technology that had blasted through

1:26:09 the universe and suddenly showed up on our shores.

1:26:13 AI has been around for as long as computer science has been around.

1:26:17 In the '50s, they thought the problem was gonna be cracked in about a decade,

1:26:21 the smartest people at the time.

1:26:23 It took a little longer, but the reason it took a little longer was

1:26:27 in part because we took some blind turns on the road

1:26:31 to AI and neural networks and thought that there were other

1:26:35 symbolic representations that were gonna carry the day, and they didn't.

1:26:38 And that probably set us back about 15 years.

1:26:41 So some of it was also we had to, Go through the gaming revolution, 3D games.

1:26:46 If we had not had Quake, if we had not had Duke Nukem,

1:26:50 if we had not had Unreal Tournament,

1:26:52 we'd never have gotten AI because we would never have gotten the GPUs,

1:26:55 and therefore we would never have been able to get AI in the shape it is now.

1:26:59 So everything we did up until this moment was required to get there.

1:27:05 You were part of that.

1:27:07 Holy...

1:27:07 What a blessing.

1:27:09 What a blessing to have contributed in your small part,

1:27:12 either by training or by playing video games.

1:27:15 Literally, you could just choose to think of it that way.

1:27:18 All those times you spent in, for my case,

1:27:21 in the '90s just playing video games, they were-- You were contributing.

1:27:24 Yeah, I was contributing to the AI revolution right there and then moment.

1:27:27 To the progress of human civilization.

1:27:29 No, for sure.

1:27:30 Civilization progresses with the death of the old and the birth of the new,

1:27:34 and it continues in this way,

1:27:35 but it's still the grieving process is a part of it, I think.

1:27:39 Take a moment.

1:27:40 It's okay.

1:27:40 Take a...

1:27:41 And also, you don't even have to accept that it's all dead.

1:27:46 Like, if you are still very attached to the hand chiseling,

1:27:50 which if you'd asked me a year ago,

1:27:53 I would have predicted that I would've been more attached.

1:27:57 I'm surprised-- I'm kinda surprised-...

1:27:59 a little that I've not been more attached.

1:28:01 But the reason I've not been more attached,

1:28:03 I think, is that I found something more fun.

1:28:07 And if it hadn't been like that, if AI had just

1:28:11 replaced the thing I loved and gave me something I hated,

1:28:14 I'd probably be a little bitter.

1:28:16 I mean, this really feels...

1:28:17 It's hilarious and awesome to watch and inspiring to watch.

1:28:19 It's like Picasso all of a sudden like,

1:28:23 getting access to to image generation and getting

1:28:28 super excited and switching in a matter of months.

1:28:31 That's literally what happened.

1:28:32 So, like, it's an artist that appreciated beautiful code,

1:28:34 and all of a sudden you're talking

1:28:36 about having fun-- I think Picasso is actually

1:28:39 a good parallel because if you look

1:28:40 at his early work where he learned his craft,

1:28:44 it was painting realistic paintings, right?

1:28:48 Like, he was training to be a master of the old

1:28:52 ways of depicting reality the best way we knew how,

1:28:56 and then along comes all these other forms of expression.

1:29:00 Cubism comes along.

1:29:01 And all these forms of abstract expression,

1:29:05 he's not decrying that he's not doing Renaissance paintings anymore,

1:29:10 that he's not just painting the perfect depiction of an apple.

1:29:14 He's reimagining the apple to be a freaking square,

1:29:18 and he's getting excited about that.

1:29:20 And I feel that transition, that we've gone from this one mode of expression,

1:29:26 and we've suddenly opened the gates where the, like,

1:29:32 the color spectrum is so much wider.

1:29:34 I'm seeing more.

1:29:35 And it's perhaps because the way I became a programmer.

1:29:39 I did not become a programmer because a deep love of if statements.

1:29:46 I became a programmer because I wanted programs.

1:29:49 I wanted things to exist that did not exist.

1:29:52 And at first, I had to learn how to program to make that happen.

1:29:56 Then I happened to fall in love with programming as a craft,

1:29:59 and then I spent two decades really diving deep on that.

1:30:03 But now I'm back.

1:30:05 I'm back to having an idea and being impatient

1:30:09 beyond belief to see it exist in the world,

1:30:13 and AI has just shrunk that down to almost nothing.

1:30:18 So it's an acceleration of a rediscovery.

1:30:23 Now, I understand that that's not true for everyone.

1:30:26 There are plenty of programmers who

1:30:27 just loved programming right from the beginning,

1:30:29 and they were attracted to the logical constructs and so forth.

1:30:32 I mean, I've been there for two decades, so I get it.

1:30:35 But I also think that you're able to discover new sides of yourself-...

1:30:41 if you let it in.

1:30:43 Again, I keep coming back to this almost regret minimization framework,

1:30:48 to borrow Jeff Bezos's terms here.

1:30:52 Are you gonna look at yourself two years from now and think, "Oh,

1:30:55 I spent my time well being pissy about the present,

1:31:00 about the fact that my industry changed"?

1:31:03 Or Would I two years from now look back upon this moment and be

1:31:07 proud of myself for leaning in, for maybe having a moment of grief?

1:31:12 Again, there's room for all of it.

1:31:14 And then going, "I'm gonna learn how the world spins now,

1:31:17 and I'm gonna make not just the best of it,

1:31:19 I'm gonna make more of it."- I mean, it is true, a lot of programmers,

1:31:24 a lot of friends of mine, me, obviously you, are having fun.

1:31:30 I have had more fun with computers

1:31:33 in the last three months than at any time previously.

1:31:39 Quick bathroom break,

1:31:40 and then I gotta ask you about how you actually interact with the systems now.

1:31:45 Yes.

1:31:47 I gotta ask you about how has your programming setup changed?

1:31:53 So keyboard, voice, what's the IDE?

1:31:58 It's crazy to think about now, but yeah, I used TextMate for almost 20 years.

1:32:02 I used TextMate starting in 2005, I think.

1:32:06 I helped get the first version out, and then I just wasn't interested.

1:32:10 I wasn't in the market for an alternative.

1:32:12 And it wasn't until the switch to Linux that I was forced out of my habitat.

1:32:20 And now with the switch to, what are we calling it?

1:32:25 Agentic engineering?

1:32:26 Oh, I fucking hate that term.

1:32:27 We gotta come up with something that sounds as plain as programming,

1:32:32 but encapsulates the fact that it's with agents.

1:32:34 But-- I still think it should be called programming at this point.

1:32:36 All right, let's just call it programming.

1:32:38 Programming with agents requires a different tool set.

1:32:40 It really does.

1:32:42 And the main change here is that you're going

1:32:48 from single thread programming in your head to parallel processing.

1:32:53 When I was writing code,

1:32:55 chiseling it by hand in TextMate or even Neovim, Not that long ago,

1:33:00 I would just focus on one problem at the time,

1:33:03 and I would methodically work my way through it,

1:33:05 and that was actually the portal to flow.

1:33:07 The portal to flow was deep,

1:33:10 immersion into a single problem, see it through to the end.

1:33:15 That's not how it works with agents,

1:33:17 in part because the agents are at once both too fast and too slow.

1:33:22 They don't give you an immediate reply on something that you

1:33:27 asked them to do that's the same as typing on a keyboard.

1:33:32 So you have to let the agent cook-...

1:33:35 for a bit, and therefore you realize, well,

1:33:39 if I just sit around waiting for them,

1:33:41 first of all, that doesn't feel productive.

1:33:43 Even if the agent, just one of them,

1:33:45 can be highly productive, it does not feel productive.

1:33:47 It does not feel good.

1:33:48 It feels actually like you're a little bit useless.

1:33:50 And maybe I had a moment, when the first agentic moment was there and we,

1:33:55 I was running mostly one agent at a time where I felt like,

1:33:59 "I don't know about this." But you can solve a lot

1:34:02 of hard problems by simply throwing more resources at it.

1:34:05 This is the whole scaling law of AI of itself, right?

1:34:10 That if you parallelize these things and you're not running one agent,

1:34:13 but you're running a handful,

1:34:15 you can feel like you're in a flow state because you're

1:34:21 constantly doing programming work in the sense that you're making decisions,

1:34:25 and you're helping either unblock an agent because

1:34:28 it has a question about which direction to take, or you're ready for a new task.

1:34:32 And to do that, you need a different setup.

1:34:34 I started first doing it in tmux

1:34:37 and just having separate panes and having separate splits.

1:34:41 Basically, a terminal with tabs is a good way to think about it.

1:34:44 You open a bunch of tabs.

1:34:45 I think most humans know exactly how that works.

1:34:48 They don't work in just one tab.

1:34:50 So still sticking to the terminal, CLI.

1:34:53 Absolutely.

1:34:54 So you're not using Claude Code app or the-- No.

1:34:56 Codex app or the-- I love the fact that this agent revolution was

1:35:01 kicked off in the terminal because I was already a huge fan of TUIs,

1:35:05 terminal user interfaces, and the terminal in general.

1:35:07 That feels like a, a really nice place to be.

1:35:11 It's a beautiful place to be.

1:35:13 The modern terminal is just a good-looking place to work.

1:35:18 So I like that.

1:35:20 And then this fact of having multiple agents,

1:35:24 especially once it's not just multiple agents running on your own machine,

1:35:28 but you start running multiple machines.

1:35:30 Now tmux alone is not enough to keep track of it,

1:35:36 and that's why as of late I've switched to this thing called Herdr.

1:35:40 And Herdr is essentially tmux plus agent notifications.

1:35:45 So whenever your agent is done and needs something for you,

1:35:47 it goes ding, a little bell telling you it's ready for its human, uh...

1:35:54 Is it master or servant?

1:35:56 I'm not quite sure always.

1:35:58 But it is ready for a decision,

1:36:00 and it also keeps track of these is it in working mode or not.

1:36:04 So I have this Herdr set up.

1:36:05 I have multiple Herdr setups actually running on individual machines.

1:36:10 I went on this crazy phase just about a month ago realizing

1:36:14 that doing this work on a single machine is not fast enough.

1:36:19 It's like I've discovered multi-core programming, but I only have two cores.

1:36:22 I'm like, "What if I had 16 cores?

1:36:24 What if I had 32 cores?

1:36:25 What if I had 64 cores?" So I instantly went

1:36:28 out and I bought these amazing KVMs called GL.iNet Comets.

1:36:35 And what they do is it's this little box.

1:36:38 You plug in HDMI, you plug in USB and connect it to the computer.

1:36:42 It's like a KVM.

1:36:43 So a KVM is a remote way of controlling a computer,

1:36:47 but what's special about this is just how easy it was.

1:36:50 You connect this thing in, you go to a webpage, log in once, set one password.

1:36:56 Now this thing can hop on your tailnet.

1:36:59 This has been the other revolution of the last year for me,

1:37:04 is discovering these WireGuard networks.

1:37:08 Uh, Tailscale is essentially turning all the computers

1:37:12 you have into a local network wherever you are.

1:37:16 Like right now on my phone,

1:37:17 I have direct access to all the computers in my Malibu office.

1:37:21 I also have access to all my computers in my Copenhagen office.

1:37:26 And I can treat them as though I sat right next to them

1:37:29 without having to punch holes in a firewall or set up complicated VPNs.

1:37:34 And what this does is it just decreases

1:37:37 the friction it takes to get new compute online.

1:37:41 So as soon as I discovered this, I looked at my closet,

1:37:44 and I realized I had a bunch of mini PCs from prior experiments,

1:37:47 and I just said, "What if I just connected all

1:37:50 of them?" And I just connected four of the computers-...

1:37:54 in a closet.

1:37:54 They all had their little Comet,

1:37:56 and suddenly I could run agents on more computers at the same time,

1:38:01 and I could control them all with Herdr.

1:38:04 And it did get to a point where I maxed out my own processing power.

1:38:09 That I think at about...

1:38:10 What do we want to put it?

1:38:13 About four to five machines running, I don't know, three agents.

1:38:17 I have about 16 threads.

1:38:19 That's what I can run.

1:38:21 And the faster the agents run, of course, the fewer threads I can run,

1:38:23 but at the current pace,

1:38:24 I can run about 16 threads-- Mm-hmm, at full acceleration.

1:38:29 And that's part of why I've gone from just

1:38:33 being excited about the agent moment to being delirious.

1:38:38 Because I think as you say, some of it is a little assaulting.

1:38:42 Again, we're on a dial-up bandwidth-wise with our computer.

1:38:46 I mean, it's actually hilarious.

1:38:49 So you think of I'm sitting here a year ago, right?

1:38:52 I'm chiseling my code.

1:38:53 Yeah.

1:38:54 I'm writing my lines.

1:38:55 And over the course of, like, an hour,

1:38:57 I will have written one beautiful controller, one beautiful model.

1:39:01 Like, that's one file.

1:39:03 And I've really worked on it, right?

1:39:05 So maybe there's 60 lines left.

1:39:06 So maybe my bandwidth at this moment is, like, 30 lines an hour.

1:39:11 I think that's even high.

1:39:13 Maybe it's 20 lines an hour.

1:39:14 Now I'm running 16 threads.

1:39:16 I'm producing at some times hundreds of lines

1:39:20 of hour or hundreds of lines of code per hour.

1:39:23 Now, let me pause myself there for a hot moment.

1:39:26 I hate that metric, right?

1:39:28 Like, lines of code is a stupid metric in general to measure these things,

1:39:32 and I think it's right from some in the programming community to ridicule

1:39:37 the AI psychosis when everyone talks about

1:39:39 the number of lines of code they're writing,

1:39:41 and then you ask them, "What did you build?" And they're like,

1:39:44 "Well, I..." And then no good answer comes out, right?

1:39:47 It's just a nice shorthand to-- It is a nice shorthand.

1:39:50 It is a representation or encapsulation of how much output that's coming out.

1:39:54 Now, whether the output is good or bad-...

1:39:57 is not a referendum on that.

1:39:59 But I was producing and producing so much more now, right?

1:40:02 And therefore I'm able to keep all these threads going.

1:40:05 So that's the setup.

1:40:06 It's still Neovim, but at this point, I'm not writing a lot of code,

1:40:11 so I'm using Neovim as a project browser and then as a way

1:40:15 to kick off Lazy Git to see the change log for what's there.

1:40:19 And even that, I'd say if GitHub was

1:40:23 a little faster at showing you your pull request,

1:40:26 the web is probably actually a nicer place to do that.

1:40:28 I don't know.

1:40:29 There's also this other tool I've been playing with a bit called Hunk,

1:40:33 which just produces Diffs in a really nice way.

1:40:35 So you could look at that too.

1:40:37 But I find that when I look at Hunk, I only see the change set,

1:40:41 and when I'm reviewing output from an agent,

1:40:43 I often want to see the surrounding context.

1:40:45 Oh, yeah, so it changed this file, but actually,

1:40:47 what do we have in this other file

1:40:49 that wasn't touched but maybe should've been touched?

1:40:51 So that's why I still like Neovim as a way of doing it.

1:40:54 But it's all happening, by the way, of course, in Amachi.

1:40:58 So it's all happening on Linux.

1:41:00 And this was the other major breakthrough with agents.

1:41:03 Agents love the Unix philosophy.

1:41:07 It loves individual tools that it can invoke through the command line,

1:41:12 and there is no operating system on Earth of the majors,

1:41:16 I'm counting three here, Mac, Windows, Linux,

1:41:19 that works as well with that mechanism as Linux.

1:41:24 Everything in Linux is either a config file or a CLI tool.

1:41:29 Now, that was its main drawback five minutes ago.

1:41:33 This was the reason people didn't like Linux.

1:41:35 It's like it's all config files and CLI tools.

1:41:39 What great irony that the universe has played it upon us that now

1:41:45 the drawbacks of Linux five minutes ago are now its major selling points.

1:41:50 This was one of the things I thought I was

1:41:54 stuck for a weekend with a Mac four months ago.

1:41:58 I thought I had a computer at the place I was going that was a Linux machine,

1:42:02 so I didn't bring my laptop,

1:42:03 and I found out when I arrived I only had a Mac Mini.

1:42:08 So I was gonna make the best of a bad situation here and, Set

1:42:12 up my Mac in some of the ways I've been thinking about with, With Linux.

1:42:17 And you can actually do a lot now.

1:42:20 Homebrew has gotten really quite good.

1:42:22 Homebrew is the missing package manager for the Mac,

1:42:25 and no one has done more to move the Mac forward in terms of ease of setup.

1:42:31 But still something as simple as Raycast, I don't know if you've used that.

1:42:35 That's the- There's no config file that you can just access.

1:42:39 You have to go into the GUI, export a file, then take that file,

1:42:44 I don't know, in your freaking backpack on a USB key.

1:42:48 And then you can import it somewhere else.

1:42:50 You cannot automate the entire setup of your machine.

1:42:53 You can't automate at all the configuration of Mac's default key bindings.

1:42:59 That has to be a manual process where you're clicking

1:43:01 with a mouse like a caveman to set up your machine.

1:43:06 I mean, there's ways around it.

1:43:08 Not good ones.

1:43:08 I looked.

1:43:09 I tried hard- So here's my, been my journey.

1:43:12 Obviously I'm a Linux person, but because of Adobe Premiere, Adobe products,

1:43:16 I'm also a Windows person,

1:43:18 so oftentimes I would use WSL, Windows Subsystems for Linux.

1:43:23 So Linux inside Windows, which is...

1:43:24 in the agentic era is not a good kind of Linux,

1:43:27 'cause it's just-- It's a sandbox-...

1:43:30 it's a sandbox, and you want the Linux

1:43:31 to be unleashed to be able to do everything.

1:43:33 Correct.

1:43:34 And so I have now woken up to things like GrayCat,

1:43:37 and have you know, I'm a keyboard person.

1:43:40 Where's the config file?

1:43:41 What, can I, can I basically do everything where an agent can set everything up

1:43:47 for me and save the configuration so I

1:43:50 can replicate it across systems and everything's automated?

1:43:52 And so I had to ask a lot of those questions, and a lot of them are missing.

1:43:56 They don't have good answers.

1:43:57 But oftentimes you could actually just build the app yourself.

1:44:02 But there are hacks.

1:44:03 Yeah, they're hacks.

1:44:03 There are hacks.

1:44:03 You don't have to live like this, Lex.

1:44:05 There's a better...

1:44:06 Actually, this is a great moment.

1:44:08 Until you have a video editor.

1:44:11 So this is a good moment.

1:44:12 Yep.

1:44:13 Um, I heard it was your birthday.

1:44:15 So I brought a little gift.

1:44:18 No.

1:44:19 We are going to get you onto the agentic

1:44:22 operating system whether you like it or not.

1:44:23 This is awesome.

1:44:24 And I asked-...

1:44:26 my friends at Dell whether maybe they had a machine that, um,

1:44:30 I could get you for your birthday.

1:44:33 Oh, no.

1:44:33 So here's the machine I use-...

1:44:35 which is-...

1:44:36 well, the same one I use, but it's now yours.

1:44:39 It's a-...

1:44:40 Dell XPS 14-...

1:44:43 already set up with Omarchy 4-...

1:44:45 ready to be configured for you.

1:44:47 And once you realize that all your...

1:44:50 By the way, we should almost time this.

1:44:53 You can be up and running.

1:44:55 You have to answer, like, five questions in about one minute.

1:44:58 You could literally do it live if you wanted to.

1:45:01 Oh, Omarchy, beautiful, modern opinion Linux by DHH.

1:45:03 Yeah, let's do it.

1:45:05 Yeah, fill it out while I talk.

1:45:06 So-...

1:45:08 exactly what you realized,

1:45:09 that pain on the Mac was the exposure I got for about a weekend,

1:45:14 and I tried really hard, and I found some decent workarounds,

1:45:17 the main one being that Homebrew now has tabs for GUI apps.

1:45:22 Because it's hilarious to think upon the recent

1:45:26 past where people would go to websites,

1:45:28 like, "I need Zoom." They would enter zoom.com,

1:45:32 find the download link, download a DMG-...

1:45:36 double-click on that, see a weird manual interaction telling them to drag

1:45:41 an icon from one place of the screen to the other,

1:45:45 and that was how you installed apps.

1:45:48 And you'd just do that over and over again

1:45:50 until you had all the apps that you wanted,

1:45:52 and if you were setting up a new machine and you had, like,

1:45:55 12 apps, you'd go to 12 different

1:45:57 websites to download them like a total caveman.

1:46:01 Homebrew has solved a lot of that, but it has not solved all of it,

1:46:04 and has solved nothing in terms of the-- Sorry to interrupt,

1:46:09 but look how fast this is going.

1:46:11 Oh, it's gonna be done in about 40 seconds.

1:46:15 And then you're gonna be inside of Omarchy.

1:46:17 Super is the Windows or Command key on your keyboard.

1:46:20 So which one is the Super key?

1:46:22 Yep, that's the Windows key.

1:46:23 Oh, actually, no, no, you have a special edition.

1:46:25 There are only about six of these computers in the world.

1:46:27 Oh, there's a Super key.

1:46:29 It says Super, and look at where the Copilot

1:46:31 key is on the other end of the keyboard.

1:46:33 That's the Omarchy logo.

1:46:35 Nice.

1:46:35 Okay, cool.

1:46:37 This was Spencer Bull at Dell,

1:46:38 who I've been working with on a lot of this stuff.

1:46:41 They've been incredible.

1:46:43 They, for a moment, let go of the XPS brand.

1:46:46 Yes, they did, last year.

1:46:48 And I was like, "This is dumb.

1:46:49 This is...

1:46:50 XPS is awesome."- It was not just dumb that it got rid of the XPS branding.

1:46:54 They also got rid of the Escape key

1:46:56 and put in capacitive touch buttons at the top.

1:46:59 The mistake that Apple realized like three years ago was terrible,

1:47:03 and everyone hated them for them.

1:47:04 But now they have an incredible team,

1:47:07 and this was actually my first Dell laptop for myself.

1:47:12 Because I've been using Dell on our servers.

1:47:14 Dell was what powered our cloud exit.

1:47:17 We've been running Dell since the mid-2000s, I think.

1:47:20 But I never ran their laptops, because compared to a MacBook,

1:47:24 they just weren't up to the task for me.

1:47:27 The XPS' from 2016, or from 2026, with those Intel Panther Lake chips,

1:47:34 are the first Dells that I would personally actually use.

1:47:37 And it's funny because when I got mad at Apple originally,

1:47:41 I thought, like, "Do you know what?

1:47:42 I should just get a Dell," because we already used them and whatever,

1:47:45 and I got a Dell, and I was like, "Man, I can't do it.

1:47:48 I can't do it.

1:47:49 It's not good enough." And then lo and behold,

1:47:52 this year they just fixed everything.

1:47:56 They fixed the battery life.

1:47:58 They fixed the performance.

1:47:59 They fixed the compatibility.

1:48:00 They fixed the weight.

1:48:01 They fixed the screen.

1:48:03 That's a tandem OLED screen-- Yeah, that's incredible.

1:48:06 which means it's actually better tech than

1:48:07 what you'll get in a freaking MacBook.

1:48:10 It's lighter than a MacBook,

1:48:13 and the chip is finally competitive because Intel spent

1:48:17 a good five years developing the 18A process node.

1:48:21 It took a long time, but now it's here, and it's freaking incredible.

1:48:26 We finally have competition to Apple M chips-...

1:48:30 that can go as long as on battery and can have competitive performance.

1:48:34 And there's a beautiful default theme to this thing, and then what is it?

1:48:37 Super Space opens up.

1:48:40 Actually, you can just push the Omarchy key.

1:48:41 It opens up the menu.

1:48:42 So now you control everything from there.

1:48:44 Oh, same thing as the Super.

1:48:44 Okay.

1:48:45 Yeah.

1:48:45 Cool.

1:48:45 Same thing as Super Space.

1:48:47 So this has been a huge change for me,

1:48:51 that we finally have Linux machines that you don't have to make excuses for.

1:48:57 Because I don't mind making excuses because

1:48:59 excuses are really just another word for trade-offs.

1:49:02 I love that Framework 13 I think I talked about last time.

1:49:06 I still like them.

1:49:07 They still make good computers.

1:49:08 But I think for a lot of people, they didn't want a DIY experience.

1:49:13 They wanted something that just out of the box felt like

1:49:17 it was on par in build quality and whatever with Macs,

1:49:22 and now we finally have it.

1:49:25 So happy birthday.

1:49:27 Thank you so much.

1:49:29 An Omarchy Dell laptop from DHH is the best birthday present.

1:49:33 This is awesome.

1:49:35 Check the screensaver, by the way.

1:49:36 Yeah, it's kind of amazing.

1:49:37 I'm seeing it.

1:49:39 On that tandem OLED screen-- It's gorgeous.

1:49:42 it is absolutely mind-bogglingly awesome.

1:49:46 Look at that.

1:49:47 There's like a retro look to the whole thing.

1:49:51 Totally.

1:49:51 It's like Fiverr pop-- It's like running a bulletin board system in-...

1:49:54 the '90s, which is exactly what I did when I was 14 years old.

1:49:58 So for me, it's a complete throwback and an awesome aesthetic,

1:50:02 combining this sense of modern rising culture in Unix with-...

1:50:08 this throwback of the '90s.

1:50:10 Yeah, like a retro look for an agent first OS.

1:50:13 Exactly.

1:50:13 It's crazy.

1:50:14 Which again, pairs...

1:50:14 This was why I was so excited that-...

1:50:17 all these agent harnesses arrived as TUI's, arrived as CLIs-...

1:50:21 has arrived as a terminal because suddenly,

1:50:25 a whole new generation of both programmers

1:50:27 and regular users discovered the glory of the terminal.

1:50:30 The terminal was actually one of the most,

1:50:34 impressive and productive user interfaces we can have.

1:50:38 Now, we've built user interfaces since then that were easier to learn.

1:50:43 The mouse is a wonderful invention for letting someone who doesn't

1:50:48 know a lot about computers just click around and find things out,

1:50:50 but the terminal is for people who want to learn their systems,

1:50:55 who want to learn the keyboard commands.

1:50:57 So you think the TUI is gonna stick around?

1:51:00 100%.

1:51:00 Now, it's not gonna be the only thing.

1:51:02 Of course it's not, because these labs now have trillion-dollar valuations,

1:51:05 so they need to be able to reach billions of people.

1:51:09 And I don't think we're gonna teach

1:51:10 billions of people the intricacies of the TUI.

1:51:13 So we're gonna have other tools, and we already have other tools, which is,

1:51:17 by the way, one of the sweet things about Linux, the ChatGPT Codex tool.

1:51:23 Yeah, I saw that.

1:51:23 Literally within two hours of that announcement dropping,

1:51:26 I had it wrapped up and ready to go in Omakub.

1:51:29 Already now, if you go to...

1:51:30 You can actually hit the Omakub key-...

1:51:34 and then just write ChatGPT.

1:51:35 It should pop up as, like, install.

1:51:38 And then just hit return.

1:51:41 Sudo password for Lex.

1:51:42 Yeah, whatever you put in.

1:51:43 And then it's just gonna install it,

1:51:45 and it's gonna pop it up, and you have ChatGPT.

1:51:48 Nice.

1:51:48 And that's how a lot of the apps are,

1:51:50 and all the agentic harnesses are built in that way.

1:51:53 Um, Tailscale, as I talked about, if you wanna hop on that, is that way.

1:51:57 Uh, Dropbox.

1:51:57 And the reason all these things are in there is because I use them.

1:52:00 It's because I built this system for me, and I thought,

1:52:03 first and foremost, how do I build my dream computer just for me?

1:52:06 And if I get my dream computer,

1:52:08 I am assured there are other people who are gonna have the same dreams as I do.

1:52:13 Now, you're obsessed about getting the installation to be under one minute,

1:52:17 and a bunch of people asked you online why.

1:52:20 So Why?

1:52:21 Why were you so obsessed to getting the installation to be less than 60 seconds?

1:52:28 First, let me quote Mitchell Hashimoto,

1:52:30 who created Ghostty terminal that's awesome and also available

1:52:33 on Omakub and was also the founder of HashiCorp,

1:52:37 who had this great post on X not too long ago that had the killer line,

1:52:42 "The pursuit of excellence deserves no

1:52:46 explanation." Wanting to make something as good

1:52:50 and as fast and as beautiful as possible has no need for justification.

1:52:58 We should want beautiful, fast systems that are a delight to use.

1:53:04 Now, the funny thing with the install time

1:53:06 is it didn't actually start with a one-minute mission.

1:53:08 I was being modest when it got going.

1:53:11 I thought if I could set up my entire system in 15 minutes,

1:53:15 it would already be such a dramatic improvement over

1:53:17 what came before it that I would be perfectly happy.

1:53:20 And it came off the fact that the benchmarking had been skewed so out of whack.

1:53:25 You talked about the Commodore 64 being your first computer, right?

1:53:28 When you unwrap that Commodore 64 and you hit the on button,

1:53:32 in I think about less than one second,

1:53:35 the basic interpreter was ready to accept your command.

1:53:38 It was literally almost instant.

1:53:40 There was virtually no boot time.

1:53:41 Do you know how long it took me to set up a new Mac that I had

1:53:46 bought for actually the one app that I

1:53:49 still haven't vibe coded my way into on Linux, which is Adobe Lightroom?

1:53:53 Brand-new computer, out of the box.

1:53:57 I set it up.

1:53:58 There's a software update.

1:53:59 Before I was ready to install Adobe Lightroom that I

1:54:03 had it fully up to date, it took 42 minutes.

1:54:08 That's grotesque.

1:54:09 That's absurd.

1:54:10 That's an insult to everyone who likes computers,

1:54:13 that you get a new computer and before you can use it,

1:54:17 you have to spend 42 minutes doing updates on something that was

1:54:21 already installed on it out of the box from the factory.

1:54:24 That is just preposterous, and it's crazy that that's not even the worst thing.

1:54:29 I also got a PC, um, because all these Linux machines, they come as PCs, right?

1:54:34 Right.

1:54:35 And they have Windows pre-installed.

1:54:36 So I got one.

1:54:37 This was just three weeks ago.

1:54:40 Brand-new machine, Intel Panther Lake, the whole shebang.

1:54:44 An hour and 35 minutes from unwrapping the thing until it's ready to use.

1:54:49 How long did you take before you think you were in Omakub?

1:54:54 I mean, it's less than a minute for sure.

1:54:55 A minute.

1:54:57 We can build computers like this.

1:54:58 We have the technology.

1:55:00 We've had the technology since like 1981 when the Commodore 64 first came out.

1:55:05 The fact that we now have a computer that's a trillion

1:55:08 times faster and you have to spend 42 minutes or an hour

1:55:10 and a half setting it up before it's ready to use

1:55:13 is so infuriating to me that it became a passion.

1:55:17 And it's funny Just while we were having a little break here,

1:55:21 I was talking to Ryan Hughes, my co-conspirator here.

1:55:23 We now have, we think, a line where we can create these special

1:55:27 turbo images that are pre-made for specific

1:55:29 computers because it relies on some knowledge

1:55:32 of the hardware, like the Dell XPS.

1:55:35 We're gonna make a Dell XPS turbo image that can install in about 12 seconds.

1:55:41 A full Linux system installed in about 12 seconds.

1:55:44 That's incredible.

1:55:46 That obsession is what gives me this great satisfaction with using computers,

1:55:52 that we do not accept things the way they are.

1:55:54 And this is why this agent-accelerated reality

1:55:57 that I now inhabit is so invigorating.

1:56:01 Because I can look at these kinds of constraints and go, do you know what?

1:56:05 Um, I could have stopped when we were at five minutes.

1:56:08 That would already have been fast enough.

1:56:10 This was the argument on X, right?

1:56:12 Why do you keep going?

1:56:14 Five minutes is fast enough.

1:56:15 Okay, maybe for you, but for me, 12 seconds is a lot more fun.

1:56:20 And I always try to think these first principles.

1:56:24 Okay, in that machine, actually hit the Omarchy, key again, and then do speed.

1:56:31 You'll see a list come up, and one of them is, like, disk speed.

1:56:35 Speed test, network speed, disk speed.

1:56:38 Disk speed.

1:56:38 Do do that one, and then run it.

1:56:40 So this is testing the performance of the built-in NVMe drive,

1:56:43 and what are you getting?

1:56:45 Seven GB a second.

1:56:47 So seven gigabytes a second.

1:56:49 That's how fast that drive is.

1:56:50 Why can we not install a Linux distribution in one second?

1:56:54 The Omarchy distribution is 5.8 gigabytes.

1:56:58 If we can run at seven gig...

1:57:00 Now, of course, you can't actually do that because

1:57:02 the USB key is a lot slower than that, but you

1:57:04 should be able to maximize the underlying physics

1:57:09 and protocols that we're dealing with, and that's your goal.

1:57:13 We gotta get all the way back down to that.

1:57:15 Yeah.

1:57:15 I love that you're doing a Linux distribution.

1:57:18 'Cause I mean, you could be do...

1:57:21 You could be building anything.

1:57:22 But the fact that you're building a Linux

1:57:25 distribution means you're rethinking about what a computer is,

1:57:28 'cause a operating system is the fundamental, like, the foundation of how...

1:57:32 you interact with a computer.

1:57:33 Yes.

1:57:33 And it's been so stale.

1:57:34 Yeah, it has.

1:57:35 The macOS we have today is scarcely different

1:57:39 from the macOS we had 10 years ago.

1:57:41 In fact, it's worse in many ways because

1:57:44 Apple continues to chokehold the control on users.

1:57:48 They don't want them to install software that they haven't already said good-...

1:57:54 to go for.

1:57:55 They don't want users to reconfigure their hotkeys without doing it manually.

1:57:59 They didn't want to change the 500-millisecond animation switching workspaces.

1:58:05 It is an infuriatingly locked-down computer.

1:58:08 Now, to Apple's credit, it's a pretty good computer for being locked down,

1:58:13 but I don't want a locked-down computer.

1:58:15 I wanna own my computer.

1:58:17 Better yet, I wanna mutate my computer,

1:58:19 and this is where the agentic age needs a new operating system.

1:58:24 When you can vibe code whatever app comes to your mind,

1:58:27 you should be able to vibe code your operating system.

1:58:30 You should be able to change anything, how it looks, how it works,

1:58:33 what panels are there, what are not there, and that requires Linux.

1:58:37 As simple as that.

1:58:38 There are no...

1:58:38 None of the other two operating systems can deliver this.

1:58:42 But can you speak to the, I think somebody asked you, like,

1:58:45 "Who the hell are you to think you could do better than Ubuntu?" For example.

1:58:49 Or, or like, "Why are you doing it when

1:58:51 Ubuntu is already pretty good?" And your answer was,

1:58:53 "I thought I could do better." So can you speak to, this was on X,

1:58:58 can you speak to, like, where the craziness comes from, thinking that you

1:59:04 as one person or maybe with Ryan,

1:59:07 you could start building something that's going to be

1:59:09 better than Ubuntu or Arch or any of this?

1:59:13 Well, I'll say that it didn't start quite as grandiose.

1:59:16 When I got going with my Linux journey,

1:59:19 I actually just started with Ubuntu, and I just built on top of it.

1:59:22 Here's a system that already does a lot of the things I wanted to do,

1:59:26 and at least it's Linux.

1:59:27 So can I just build on top of it and make it more modern,

1:59:30 make it actually look good, and have the programs I wanna use?

1:59:32 I got, like, some version of that, and then

1:59:35 I discovered that there's seven layers deeper you could go,

1:59:38 and as, the closer you get to the metal, the closer you get to the kernel,

1:59:42 the closer you get to the individual packages and so forth,

1:59:45 the more freedom you discover.

1:59:46 So as I discovered more and more freedom, I got more and more ambitious,

1:59:51 and I realized, oh, surprise, surprise,

1:59:54 I have strong opinions about how computers should work,

1:59:57 how they should look, what programs should be installed,

2:00:01 and whether they should take 42 minutes to install or 45 seconds.

2:00:06 That's the current world record, by the way.

2:00:09 So if anyone installs Omarchy Quattro after

2:00:11 this and can beat 45 seconds, take a screenshot,

2:00:13 send it to me, and you'll be the new holder of the world record.

2:00:17 When, how long ago did you cross the 60-second mark?

2:00:20 It's funny because it's like that the old limit on, what was it?

2:00:25 It was a four-minute mile or something-- Yeah.

2:00:28 Uh-huh-...

2:00:28 like that in 1952-...

2:00:30 that people thought was impossible for humans to beat.

2:00:32 Then one guy beats it,

2:00:34 and literally within the next few months, three more guys do, right?

2:00:37 So we had this goal.

2:00:38 I originally had the goal it was two minutes.

2:00:40 And then we beat two minutes, and I'm like, "Well, if we beat two minutes,

2:00:43 why can't we beat one minute?" And I remember thinking like, "That'd be crazy.

2:00:47 You can't install a modern operating system in one minute when

2:00:51 they're all..." If the other ones are taking 42 minutes, right?

2:00:55 Like Apple, full of a lot of smart people, like,

2:00:58 it took 42 minutes, so that's gotta be the bar.

2:01:00 No, no, it's not the bar because there's no speed limit.

2:01:03 Like, none of the things you take for granted

2:01:06 are fixed pillars of reality in almost any case.

2:01:09 Until you get down to the seven gigabytes

2:01:12 per second that drive is able to transfer,

2:01:15 you've not reached the limit yet, so you should just keep pushing.

2:01:18 And of course, then it becomes a game in itself and a satisfaction itself.

2:01:22 Every second I could shave off was an excitement.

2:01:24 And when we broke the one-minute barrier, which was really only broken, I think,

2:01:29 three weeks ago or something like that, maybe

2:01:31 even two weeks ago, I just got crazy.

2:01:33 And then I got a swarm of agents to run all

2:01:36 these auto research loops of trying all these different theories on like,

2:01:39 oh, if you change this thing,

2:01:40 if you change that thing, what if you cut this out?

2:01:43 What if you do things in parallel?

2:01:44 What if while you were typing in your username,

2:01:46 we start preloading packages into memory system, we can fire them in faster.

2:01:50 And it becomes a fun game.

2:01:52 And I think that's how you should think about product development.

2:01:54 It should be fun.

2:01:55 You should, it should be frivolous.

2:01:57 It should be trying to overshoot the target and making it

2:02:01 so much better than what it quote unquote needs to be.

2:02:04 Especially now because you can move so quickly and you

2:02:06 can try so many different approaches because of the agents.

2:02:09 Yes, that just makes it more satisfying.

2:02:10 But I think this was always true.

2:02:12 If you look at the greatest Mercedes-Benz of all time,

2:02:14 I think it was called the W126.

2:02:18 I saw this wonderful documentary on it.

2:02:20 It was steered by this guy who

2:02:22 for years had worked in Mercedes' department for safety.

2:02:26 And suddenly he gets put in charge for the whole S-Class project,

2:02:30 and he just puts in all of it.

2:02:32 This is where you got the auto-retending seat

2:02:34 belts and washers on the headlights and everything.

2:02:39 This car was so overbuilt for the market.

2:02:42 It was unbelievable.

2:02:43 But you look at that and go like, "Well,

2:02:45 I want that." In just the same way that you look at a mechanical watch and go,

2:02:51 "It could go down to 100 meters?" Or I think

2:02:55 Rolex has one that can go down to 4,000 meters,

2:02:57 the, the deep ocean, the deep diver ocean...

2:03:00 or whatever it's called.

2:03:01 No one outside of three people in the world,

2:03:04 whoever buy that watch, will ever put it to the test.

2:03:07 But don't you wanna be the kind of person who is

2:03:11 a patron of people who wanna push the human race that much forward?

2:03:15 I want the fastest car breaking the speed limits.

2:03:18 I want the diver watch that can go down the Mariana Trench.

2:03:22 I want the operating system that can install in less than 60 seconds.

2:03:27 So from the individual level, I admire those people.

2:03:30 I think everybody should strive for that.

2:03:32 But also the consequences of that we should mention,

2:03:34 like the pursuit of excellence, even when the metrics seem to be frivolous,

2:03:40 the maybe unexpected consequences of that is other kinds of excellences,

2:03:44 discoveries, innovations,

2:03:44 and all that kind of stuff that has nothing to do with the speed.

2:03:48 100%.

2:03:49 It's discovery.

2:03:50 It's nice to have the metric that you're chasing be the catalyst,

2:03:54 the engine that drives innovation, because you'll just discover.

2:03:57 It's the same reason you go to the moon, you go to Mars.

2:04:00 It's like, why?

2:04:02 Who cares?

2:04:03 Well, actually, the real thing is

2:04:06 you'll discover a lot about chemical propulsion.

2:04:09 You'll discover a lot about how humans can live for a long time in space.

2:04:15 Maybe you'll discover whole new ways of doing solar energy

2:04:18 in space or data centers in space or whatever the hell.

2:04:21 That's right.

2:04:22 All kinds of...

2:04:22 Maybe you'll meet—- That's right- ...aliens, but you never know.

2:04:26 But setting a goal and chasing it,

2:04:30 especially when it's numerical, is really nice.

2:04:33 And setting a grand goal is even better.

2:04:35 When you refuse to just do what's practical

2:04:39 and you strive for what's slightly beyond plausible,

2:04:44 that's when the magic happens.

2:04:46 And we should be going all in on that.

2:04:50 This is one of the reasons...

2:04:51 And in fact, to some degree,

2:04:53 this specific goal was motivated by seeing others do this.

2:04:56 I mean, Elon is the singular individual in this conversation.

2:05:01 The guy is so out there in his goals.

2:05:06 They seem so preposterous at first glance.

2:05:09 I've certainly considered them preposterous on many glances.

2:05:13 And then you realize that they come true sometimes.

2:05:19 Maybe not always on schedule, and so what?

2:05:22 So what?

2:05:22 So we didn't get the self-driving car on schedule.

2:05:26 I gave Elon a lot of shit for that back in '17.

2:05:29 But now it's here, and who would have gotten us here

2:05:33 except for the person who thought it was possible in '17,

2:05:37 who is so impatient with reality that the sheer force is just

2:05:44 gonna propel that reality to the place he wants it to be.

2:05:48 I think that's deeply inspiring,

2:05:49 and we need these kinds of, goalposts to be moved.

2:05:53 We need someone to run the four-minute mile,

2:05:56 so the rest of us realize, "Well, okay,

2:05:58 but I guess I can run it quicker." That's what I try to strive

2:06:02 for, and I really strive for with Omarchy

2:06:04 because it's just more fun to live that way,

2:06:07 to have high aspirations and ambitions.

2:06:09 And I will say I've actually changed my mind on that a little bit.

2:06:12 I spent quite a lot of time speaking the other book

2:06:16 that there are a lot of people who don't actually need those ambitions.

2:06:22 And I also— That's also true.

2:06:24 We don't want 8 billion Elons running around the world.

2:06:28 Would be rather crazy, I think.

2:06:29 But I've come to realize that for entrepreneurs,

2:06:33 it's healthy to have suitable chunky goals.

2:06:37 The idea of just running the Italian restaurant can't just be that.

2:06:43 It should be maybe I just run the one Italian restaurant,

2:06:47 but I make the best damn pizza possible.

2:06:50 I make the most delicious Parmesan.

2:06:55 We have to strive for excellence to muster the motivation to keep going.

2:07:00 And yeah, to dare to think big like you could build a Linux-- Yes-...

2:07:05 distribution.

2:07:07 Is there some stuff you remember about what it took to get us to be so fast?

2:07:11 You mentioned a few things that were, the agents were discovering.

2:07:15 One of the things was to treat the...

2:07:19 lag of human input as an opportunity to preload.

2:07:22 So— Hmm...

2:07:23 there's five questions you answer when- Yeah...

2:07:25 you set up an Omarchy machine.

2:07:26 So it's doing stuff in the background.

2:07:27 We're doing stuff in the background.

2:07:28 And that was one of the things I didn't even consider,

2:07:31 which is, I mean, the oldest trick in the book.

2:07:34 All sorts of video games for a long time have

2:07:36 done this, where you let the player interact with something,

2:07:39 and then you do preloading and so forth in the background.

2:07:42 So it's an age-old technique,

2:07:43 but I just hadn't thought about it for an installer.

2:07:45 I thought, first you ask the user some questions about

2:07:48 their password and their username and their time zone and so forth,

2:07:51 and then you do the work, when, in fact, you could preload that.

2:07:55 Now, the other things were just a bunch of other tweaks

2:07:58 to how the installer itself was doing things in a certain order.

2:08:03 Some of them was also just shrinking it.

2:08:06 The last version of Omarchy was 7.5 gigabytes, and a lot of the time we're down

2:08:12 to now is literally decompressing compressed package files.

2:08:15 That's the bulk of what is on the ISO, and that's the bulk of the install time.

2:08:19 So by shrinking it down to, I think we're at 5.85 gigabytes now,

2:08:23 a pretty chunky reduction, it almost literally translated one-to-one,

2:08:27 and I did some really interesting things there.

2:08:30 So Omarchy uses the JetBrains font, which is an awesome font,

2:08:35 beautiful font, and actually a font I didn't like at first glance,

2:08:39 but then it was the only font I

2:08:40 could get to look perfect in Mitchell Hashimoto's Ghostty.

2:08:45 Ghostty, for whatever reason, did not wanna render my beloved Bitstream Vera

2:08:50 Sans just the way I was used to it.

2:08:53 Mm-hmm.

2:08:53 But it rendered this JetBrains font perfectly, so I switched over.

2:08:58 But it...

2:08:59 The standard package for JetBrains font on Arch is 200 megabytes

2:09:04 because it includes all these variations and all these subtypes and whatever.

2:09:09 And I realized, like, we're not using any of that.

2:09:12 We're using the monospace version in this one Nerd, uh- Yep...

2:09:16 font patched edition.

2:09:17 Do you know how big that is?

2:09:19 I think it's 16 megabytes.

2:09:21 Hmm.

2:09:21 So I simply just came up with a new package, the slim version- Mm-hmm...

2:09:26 of the JetBrains package.

2:09:27 Nice.

2:09:27 And there, right there, I saved 180 megabytes.

2:09:30 I just went through and did that a bunch of times.

2:09:32 We did that with the NVIDIA drivers too.

2:09:34 So the NVIDIA drivers, when they came straight off the, I think the Arch repo,

2:09:39 were not compressed with the most extreme form of compression,

2:09:43 which I think called ZSTD, which is a really slow form of compression.

2:09:47 It's slow to create the archive, but if you're building a package- Hmm...

2:09:50 you have all the time in the world if that means you can then shrink the ISO.

2:09:53 So on just the two NVIDIA packages, I think we saved 200 megabytes.

2:09:58 Mm-hmm.

2:09:58 So what I loved about this was I

2:10:02 imagined myself being a car designer at McLaren.

2:10:05 So McLaren makes some of the lightest super sports cars in the world right now,

2:10:12 and they are absolutely obsessed with shaving

2:10:15 every damn last gram off their cars.

2:10:19 That's why even a new 750S McLaren is just way lighter than the competition.

2:10:25 They use a carbon fiber monocoque,

2:10:27 and that helps them over like Ferrari, for example, that's uses aluminum.

2:10:31 But they also just have this obsession with weight,

2:10:33 and I would see this interview with the McLaren

2:10:36 folks talking to some journalist just going over, "Yeah, we saved, uh-...

2:10:40 370 grams- Yeah...

2:10:42 off that one." And I just went, "Wait, what?

2:10:45 There's someone on a car that weighs maybe 1,040 kilos or 1,380 or something.

2:10:52 They're s- They're worrying about 370 grams?" Yeah.

2:10:57 What obsessed maniacs.

2:10:59 Hmm.

2:10:59 I love them.

2:11:00 Yeah.

2:11:00 I want that in my life.

2:11:01 I want a McLaren-...

2:11:03 just for that reason.

2:11:05 So I thought in that moment as I

2:11:07 was shaving individual megabytes off the packages,

2:11:10 man, I'm kinda like a McLaren car developer here, even though- Yeah...

2:11:13 I mean, we haven't even completed that work.

2:11:15 The package could shrink even more.

2:11:17 But this is actually this other conflict that we've

2:11:20 had with Omarchy and the traditional Linux community is bloat.

2:11:24 Mm-hmm.

2:11:25 This idea that it's actually not kosher

2:11:28 in certain circles to ship with pre-installed

2:11:31 software because who are you to make

2:11:33 decisions about what programs people should use?

2:11:36 And I'm like, "Dude, it's in the fucking name.

2:11:39 It's called Omarchy because the Oma part is short for omakase,

2:11:43 which literally means chef's choice.

2:11:46 I'm the chef.

2:11:47 I'm making the choices, and this is the collection of applications I think

2:11:51 is awesome," which by the way has a video editor.

2:11:54 It ships with OBS- Mm-hmm...

2:11:55 to do your recording.

2:11:56 It's just in there.

2:11:58 Yeah.

2:11:58 OBS is pre-installed.

2:11:59 Nice.

2:12:00 Kdenlive is a video editor, timeline editor, which is great.

2:12:03 I edit all my videos with that.

2:12:04 And then I made my own clip editor.

2:12:06 You talked about the thing with syncing up.

2:12:09 Yeah.

2:12:09 I had this thing with clips, so I would do a podcast or whatever.

2:12:12 I need a clip for X.

2:12:13 Mm-hmm.

2:12:13 And it's kind of a hassle to use Kdenlive or other- Mm-hmm...

2:12:16 timeline editors, so I made Omacut.

2:12:17 Mm-hmm.

2:12:18 Omacut is included too.

2:12:18 Omacut, cool.

2:12:19 It's just for making clips, and you can drive the whole thing with keyboard.

2:12:22 So you can go to just the right time mark- Mm-hmm...

2:12:25 and then you can hit Control+ Space.

2:12:27 Hmm.

2:12:27 It moves the clip line to that line.

2:12:29 Then you go to- Nice...

2:12:30 the end of the clip, and you hit Alt+ Space, and it moves the end line.

2:12:33 Nice.

2:12:33 And then you hit Control+ S, and it saves.

2:12:35 Nice.

2:12:35 I can make clips so damn fast now with Omacut.

2:12:38 It's amazing.

2:12:38 But it has all the software on it.

2:12:42 It has Neovim on it.

2:12:43 It has Herder on it.

2:12:44 It has Tmux.

2:12:45 It has a terminal.

2:12:46 It has a bunch of themes.

2:12:48 It has a bunch of background images, the best ones I could pick out.

2:12:52 That takes up some space.

2:12:53 It should because this is not supposed to be

2:12:56 this just barren landscape where you have to reconstruct everything.

2:12:59 Mm-hmm.

2:13:00 This is supposed to be a productive system the minute you unpack it.

2:13:03 What is it built in?

2:13:05 So it's actually mostly built in Bash for the Linux configuration part.

2:13:10 Right.

2:13:11 Bash is a great language for that.

2:13:14 It is made for that really, for system administration.

2:13:18 It also has its limitations.

2:13:19 You should not build everything in Bash- How are agents with generating Bash?

2:13:23 Pretty good?

2:13:23 Oh, amazing.

2:13:24 Except for one thing, and I have to smack the agents over the back

2:13:29 of the head every single time I catch it,

2:13:31 and remind it to look at the agent's MD file,

2:13:33 because I will have an instruction in there not to make early exits.

2:13:37 The agents love to have preconditions instead of fully expanded conditionals.

2:13:43 I like something that says,

2:13:45 "If this, then do that." Agents like to have precondition or exit,

2:13:51 another precondition or exit.

2:13:53 And then it falls down to the final thing it actually wants to do.

2:13:56 I hate that style.

2:13:57 So except for that, it's gotten shockingly, dramatically, awesomely good,

2:14:03 to the point where when we talked last time,

2:14:06 we talked about how I would ask the agent for a Bash thing,

2:14:09 and then I would type it in myself, and I kinda wouldn't remember.

2:14:11 And then I took the effort to make sure

2:14:13 that all that competence wasn't draining out of my fingers,

2:14:15 so I actually learned how to write Bash myself,

2:14:17 and now I'm not writing any Bash myself.

2:14:19 I have not written any Bash myself for probably a couple months,

2:14:22 because the agents have just gotten so good,

2:14:24 that if you give them the few pointers on style,

2:14:26 and occasionally put pushback on complexity...

2:14:29 This is...

2:14:30 I tweeted about this a couple of days ago.

2:14:32 It blows my mind how often an agent can do something,

2:14:36 say it's done, have the work reviewed by another agent-...

2:14:39 they say it's done, and then I go like, "Eh,

2:14:41 looks a little too complicated for me." And then it'll go like,

2:14:45 "Oh, yeah, you're right.

2:14:46 You're right.

2:14:46 I totally overcomplicated this," and it'll cut it in half.

2:14:50 I'm like, "Couldn't it just do this, like,

2:14:52 from the outset?" And then someone else reminded me, humans are the same way.

2:14:57 If you work on something and someone else

2:14:59 reviews it and they just are like, "Eh,

2:15:01 it looks a little complicated," quite often you're able to go like,

2:15:04 "Ah, you're right."- See, that, that just...

2:15:06 Those little indications to me show that the programmer is still needed,

2:15:12 that intuition, 'cause I think, what, what did you say?

2:15:14 I have a sense of the landscape of the code.

2:15:17 When I build something like Omarchy, I, I do look at the code.

2:15:21 I look at the shape.

2:15:23 And I look at the proportions.

2:15:24 Like, I don't have to chisel everything myself anymore.

2:15:27 This is, again, when I'm flattering myself for...

2:15:29 Not one moment I'm a McLaren audio, uh auto engineer,

2:15:33 and the next moment I'm a da Vinci,

2:15:36 working with his whole studio of students-...

2:15:39 who are doing all the chiseling on the, the fine marble, right?

2:15:41 And he goes like, "Ah, no, the proportions are not quite right." That's

2:15:45 how it feels working with agents now.

2:15:47 That I can give feedback like an editor.

2:15:50 That, do you know what?

2:15:52 I'm not gonna write everything myself,

2:15:53 but I can tell you when the argument doesn't land.

2:15:56 I can tell you when the shape isn't proportional to the problem,

2:15:59 or when it just feels like it's too complicated.

2:16:02 And shockingly often, it just works.

2:16:04 It's like the early days, where the jokes was,

2:16:08 "Don't make mistakes." I don't think you have to tell the agents that anymore,

2:16:11 because now the harnesses are so good, they run automated testing.

2:16:14 They get that feedback.

2:16:16 But you actually do have to tell them,

2:16:18 "Make it simpler."- Do you use voice at all for the interaction?

2:16:21 Or are you...

2:16:21 How much are you typing?

2:16:23 How much are you-- I type the whole thing.

2:16:25 I actually, in Omarchy's built-in Voxtype-...

2:16:28 Audio transcription, and it's actually really good.

2:16:30 I just find that I don't think that way,

2:16:35 and it surprised myself that I don't like to talk to the computer.

2:16:39 Uh, I like when I'm in the car,

2:16:41 but when I'm in front of the keyboard, I just wanna type.

2:16:44 I also just love typing.

2:16:45 Are you using the voice?

2:16:46 I'm an introvert who doesn't like social interaction-...

2:16:49 or speaking.

2:16:50 And-- I think that's part of mine too.

2:16:53 But, like, for me, for some reason, voice is really powerful.

2:16:57 I want to give this...

2:16:58 So I, I run with this thing.

2:17:00 It's called Plaud.

2:17:01 So I, I just hook it up on my, you know on my T-shirt here.

2:17:05 You press a button, it starts recording.

2:17:07 And this is actually more, in terms of flow,

2:17:10 much better experience than behind a computer.

2:17:13 And I, uh, save, well,

2:17:16 usually it's agent gen- generated problems that I wanna really talk through.

2:17:22 And so I write, I speak a long prompt, like,

2:17:28 and a lot of it, it's like stream of consciousness reasoning.

2:17:31 So I'll sometimes just change my mind about a design,

2:17:35 but all of that is in there, and then I have an LLM that processes that.

2:17:39 With Plaud, there's a, obviously an API.

2:17:41 You can grab it, transcribe it, clean up the transcript,

2:17:45 'cause you, you want it to be...

2:17:47 'Cause I'm referring to files.

2:17:49 I'm referring to technical concepts, like files and functions-...

2:17:53 and so on, that you wanna make sure the transcription, the-...

2:17:56 the speech-to-text gets correctly.

2:17:57 So you wanna have a dictionary.

2:17:58 You want it to be aware of the code base when you process the files.

2:18:02 And then you have this gigantic thing that's often 10, 15,

2:18:06 20 minute me discussing the full design,

2:18:08 and I find that to be really, really powerful.

2:18:13 Because it doesn't have the problem overs- of overspecification because of how

2:18:20 much stream of consciousness thinking

2:18:22 the system has about what you're imagining.

2:18:25 And this is really good for the early steps of a design.

2:18:29 Like, if I wanted to build Omarchy from scratch,

2:18:32 the first stages, I talk through.

2:18:34 It could be an hour.

2:18:36 There's been prompts that I had for an hour,

2:18:38 and I'm just running and-- That's amazing-...

2:18:41 and speaking.

2:18:43 I highly recommend people try a voice-based-- So--...

2:18:48 speaking for prolonged periods of time.

2:18:51 I haven't actually-- As opposed to very specific-...

2:18:52 I haven't, yeah, I haven't used it

2:18:53 that, and maybe that's how I'm using it wrong.

2:18:55 I try to just type or speak two sentences, and then I get a little frustrated

2:19:01 that I could've typed that just as fast maybe.

2:19:03 But I have not tried this whole

2:19:05 thing of just talking continuously for 20 minutes.

2:19:09 And especially you because you speak very fast.

2:19:12 Yeah.

2:19:13 No, I could see it.

2:19:14 I think because the reason I've been,

2:19:15 I've been trying to do it for things like replying to emails,

2:19:18 and it got to a point where I thought it could help.

2:19:22 And it sort of did, but then it would make a small mistake,

2:19:25 and I'd have to correct the mistake.

2:19:27 But if you're doing stream of conscious to an LLM

2:19:29 who doesn't care about all this other stuff anyway-...

2:19:32 that probably works way better.

2:19:34 There's two things you have to do.

2:19:35 One, okay, you have the raw audio, transcribes it.

2:19:38 I use ElevenLabs for transcription.

2:19:40 They're probably the best speech-to-text model.

2:19:43 There's still mistakes there, so you need to have a prompt that has a- access

2:19:48 to a dictionary and to a lot of-- How do you build the dictionary?

2:19:52 First of all, there's a dictionary of terms that are relevant to you.

2:19:55 Then have it aware of...

2:19:56 So you have an agent that looks through, when you're talking about code,

2:20:00 looks through your entire code base for key terms that you use often.

2:20:03 Also, when we're talking about messaging, Gmail, and WhatsApp,

2:20:05 and so on, it's aware of all the people and common terms that you get wrong.

2:20:11 You've really gone deep in this.

2:20:13 That, to me, is a problem that needs to be solved-...

2:20:15 because, uh, one of the issues you probably have

2:20:17 speech to text is that it makes mistakes sometimes.

2:20:22 Correct.

2:20:22 Then I can't use it for dictating an email.

2:20:25 Right.

2:20:25 If it never makes mistakes, you-- Then it would be more appealing.