James Long: It’s not too far off. I was frustrated around the time that I started this, in 2015-2016, with not feeling I had a good understanding of my finances. I thought I had saved up enough money for this thing, but when I look at my bank account, why don’t I have as much as I thought? And it’s a surprisingly hard – , you think you could just manage your transactions and then categorize them, and then look, and then there’s your answers… But there’s enough complexities, I guess especially as you have anything complex. If you have a mortgage, or you have some kids, or you’re married, or you just have – you have one thing that is a reoccurring bill that you just don’t think is as high… I don’t know. There’s things with reimbursements; you bought a $100 thing off Amazon, but then you returned it… Where does that money go back into? There’s enough things that add up to this vague buffer of things that is a mismatch of expectations in your money.

I think I was frustrated at – I kept just watching my bank account and thinking I had certain amounts of money, but then when I actually really sat down to break it down, it’s like, “Oh, I actually have to pay this bill next month, with this much, which means I don’t have enough now to do this other thing.” It’s a surprisingly hard problem, which is I think why a lot of people just don’t really budget, and they just kind of do spend, and then they watch their bank account.

I think I am a very detail-oriented person, and so for my personality I did want to just know more. And I think that was the crux of it. I was frustrated at just not knowing… That, combined with - I typically have a high bar for software. Kind of that typical developer thing, where I go and I use a bunch of apps, and I just don’t like most of them. Because you log in to Mint and it’s terrible, for all sorts of reasons. I think it’s even gotten worse since then. Now it’s just like super-simple, ads everywhere… Lots of bad things. And it works for some people. Totally great. I’m not putting down any other people’s usages of Mint or YNAB, or anything like that. Totally great.

I did try YNAB for a while. This was actually before – it’s funny… I think this was a year or two before YNAB actually came out with their current version, which is on the web. Back then it was actually their primary app was their desktop app, which was YNAB 4, which is now there’s like –

Gerhard Lazu: Yeah, for sure. Yeah, that’s the other thing. If you have a single instance, then things get a bit tricky. Maybe you have like a blue/green thing going on where the new one gets brought up, and then when it’s healthy, it gets promoted to live. So that makes sense. If you have a bunch of them, obviously you don’t want to rotate everything. And if you have like 100 instances, I think it’s a bad time to rotate all of them maybe, potentially, depending on what they do. And imagine if you have like a couple of secrets updating at the same time, or things are very in flux, then you have like – basically, things are coming and going all the time, and that puts a lot of pressure on the CPU, on disks, on whatever the case may be. So you have like these storms of applications restarting… But yeah, you’re right, that’s like a problem for – it’s a good problem to have, let’s put it that way. It’s a good problem to have. [29:09]

So I know that this keeps coming up a lot, and I’m wondering your thoughts about committing encrypted secrets into Git repositories. A lot of people do that, and say “Oh, this is amazing! Bitnami Sealed Secrets! Yeah, this is great!” Or SOPs from Mozilla, or age… And I’ve used all those tools, and I think they’re okay, but I think there’s something fundamentally wrong with that approach… And I’m wondering if it’s just me. Or maybe I’m wrong, I don’t know. What do you think, Rob?

Bill Kennedy: So after I finished Go in Action in 2015, I had PTSD. I couldn’t write. If you go look on my blog – I’ve written over 100 blog posts since I started. If you go look at my blog, there’s almost a year gap of me blogging, because that book completely traumatized me; the process of having to – I would go to a bar every night… Not drink liquor, but when I walked in, they had my coffee, and I had a space where there was power. And I basically spent six months just to get the book done. It took about 18 months, every night after work, go to this bar and sit down and knock out three or four hours. And I was traumatized, I couldn’t write again.

And then I started writing again about a year later, the blog posts. And it was a struggle. The anxiety levels I would get trying to sit down and write was – they were tremendous. But I fought through the anxiety, and I started to be able to publish again blog posts. And I said, “I’ll never write a book again, because this was too much, I’ll never do it again.” I’d rather stay in the blogging medium, it’s enough for me.

I met Hoanh An about a year ago, he’d taken my class and he’d written all of these notes. The notes were pretty good, and he had a whole repo of code, too; a lot of it was the same from the Go training, but he had kind of made some of it his own. And he had this manuscript that he had written, and he was selling it, and people were buying it.

So quietly, I kind of just downloaded it and I looked at I said, “This isn’t bad. This is a good kind of starting point. This is the book that I want to write. I just don’t have it in me to write it, again, because like I’m still traumatized.” At the same time – and this is interesting… I have people coming to me saying, “Hey, he’s selling your material. Hey, he’s selling your material.” And at some level, maybe; but all of my material is open source, everything I do. It’s like, “Take it, make it your own. Do what you want.” And people think you’re crazy, because now people are going to compete against you. Compete against me; if you’re better than me, then you deserve – like, none of that was bothering me, at all. But I reached out to him and I said, “Look, you’ve got a good skeleton here. I think we can really narrow in the focus of what you’ve got here and put it more into the language that I speak in the class.”

So if it wasn’t for him doing that work, this book wouldn’t exist, because though I’ve wanted to write this book, I just couldn’t get myself to the point of starting it. And even over the 10 months that it took me to basically take what he had and rework it to what we have now and add to it, anxiety levels were off the chain. Like, I could work for a week and then I couldn’t work for two or three on it. And then I’d go back and work for a week, and then I couldn’t work on it. I think it wasn’t until I got about 60% done that I started to see the end of the tunnel, and I focused heavily on it. And even now, I don’t know how I got it done. I’m glad I got it done, but… I did.

Now, I decided two things with this particular book. It was really all the material in written form that I speak during the class, so I wanted this to be the notebook. Like, these were the notes that everybody had been asking me for, for like the last five years for the class. So there it is.

So I did a couple things differently here. The whole book is written in I, as if it’s me writing notes to myself. So you read that book - other than some blog posts I added to the end - it’s all I, I, I. I really wanted that kind of feel for it.

The other thing on the self-publishing side was primarily because I wanted to sell this book cheap, not to lessen the value of the book; to make it accessible to everybody, and I wanted control over that. So if I went with Manning or O’Reilly or whatever, two things would have happened; the book wouldn’t be 10 bucks online and 25 bucks on Amazon. It would have been twice as much, and I would have lost that “I” feel; they wouldn’t have gone with that kind of flow. You know, they have their book size, their styling guides, how they’re going to do it, their editors are on top of you there, and I don’t want any of that. So we decided to self-publish.

I spent more time trying to figure out how to have a copy of this book that could be online and in print where I can manage it. If you see – I also wanted the book to feel like a notebook. Not the traditional size, with a smaller font. Maybe because I’m getting old. This is the geriatric Go book here. So you get this book, it’s big, it feels like a school notebook, the fonts are large… I don’t know, I think it’s just – it’s kind of interesting to just have something different too.

Quincy Larson: Yeah. So we think 50 years out, 60 years out. I think in terms of what is likely to be accomplished. Okay, so if you go all the way back to the the first printing of the Gutenberg Bible, the first mass-produced book in history, it was more than – it was like 600 years ago. It was just with the advent of the printing press. And this book came out and it brought literacy to everyone in the 1450s. 1450s. 600+ years ago. 670 years ago. And still, we have illiteracy. I don’t believe that AI is just going to magically change illiteracy. I think that there are a lot of people who just – it’s still hard to learn to read if nobody teaches you how to read; it’s very difficult. I hear all these things of people who have unschooled their kids, and they haven’t necessarily put effort into teaching the kids… There’s just this idea that the kids will naturally learn how to read. And some kids might, but a lot of kids won’t necessarily learn how to read, and they’ll end up not having reading level nearly what their peers, and it’ll be a huge kind of crushing weight on their shoulders, like “I’m not good at reading.” And they’ll carry that with them. And I feel like it’s the same thing with technology.

People certainly have this with math. “Oh, I just suck at math.” They never learned how to do math, and they always feel bad about themselves, and they’re always intimidated whenever they’re reading a book and there’s some math, or something like that, or whenever they look at some equation… Even the most simple equations, when they’re walking around the science museum and they’re just “Oooh…!” They get this anxiety. And same thing with software. “I don’t really know how a web server works. I don’t really know what garbage collection is doing.” It’s just, in the back of your mind you’re just kind of unsure of yourself.

[01:30:00.20] And so if over the course of 600 years there are still people – when they have had this relatively abundant thing that they could read, mass market… 600 years… It’s going to be a very long time before technology literacy is solved, and before everybody can just feel confident that they understand how recursion works, and things like that. And I do think that these are things that people in Star Trek – like, I always look at Star Trek and I realize that it’s flawed in the sense that it’s just kind of this vision from the ’70s, and then it got interpolated through the ‘90s, with Next Generation, Voyager, and D6-9, which is the best one, according to Andrew Brown, big Trekkie and massive contributor to freeCodeCamp… He’s created 50 certification preparation courses on our YouTube. But I realize that it’s just kind of a sci-fi vision of the future, but it makes practical sense. A lot of the things in that show, I can see that that’s where we’re heading. We’re heading toward computers where you just talk, and you can use declarative speech instead of imperative speech. And the machine, it just gives you a lot more affordance in figuring out what needs to be done. And if you watch The Expanse, there’s lots of AI… But it’s not everywhere, and they’re not talking to AIs, and stuff. It’s just like doing the little things like recalibrating the turrets on the ship, and stuff like that.

And I think that 100 years from now, if we want to talk 100 years - let’s just choose a big, round number - people are still going to exist. We’re probably not going to be all walking around in exosuits. I will be impressed if we’ve terraformed Mars at all by then. I just genuinely don’t think that there’s the political will, and the budget, and all that stuff, and I think that people greatly underestimate how difficult it will be to colonize space, and stuff like that. I think realistically, we’re still going to be here. Hopefully, we’ll have done some stuff to mitigate climate change and mitigate job loss due to automation, and stuff like that… But people are still going to need to learn. People are still probably going to need to work and figure out a way to make money. I don’t believe that UBI is coming to save us, universal basic income, or anything like that. I think fundamentally, people are going to still need to learn and they’re going to still need to go out and do things to provide for their families. That has been how it has been throughout civilization, throughout the history of every civilization, whether that is the hunter-gatherer tribes… People have needed to go out and figure out how to keep the calories coming to keep the body going. And I think institutions come and fall, but we’ve got these human animals, 100,000 years of human tribes interacting with one another, and warring, and things like that… But I still think that they’re going to need to learn. I don’t think there’s going to be a magic matrix thing where you just “Teach me how to fly a helicopter” and there’s a program.

Techno Tim: I did infrastructure for work, but then, as you mentioned, got into software development. I’ve been a software developer for a long time now. So now infrastructure is now my hobby. So everything I do at work is cloud, 100% cloud, 100% virtual. I don’t touch anything except for my laptop.

So I always have this itch for an infrastructure. And it comes from two things. One, the media server that I mentioned earlier, but two, just having an access point and a router at home. anyone can look at their access point router and say either they love it or they hate it, you know? And if they say they hate it, it’s probably rented from whoever their ISP is. If they say they love it, they put some time into it and bought their own. And that’s kind of where I like to play, is like a little bit in the networking, and somewhat in compute and storage… So even if I weren’t totally in, as I am now, I’d still be doing this kind of stuff.

Jerod Santo: And I can see where he would be offended if we looked – we did some of the math on Changelog & Friends last week about how much WP Engine is putting into WordPress versus Automattic putting it into WordPress… And if you’re under this financial strain, and you have so much resources going to the open source side of the thing, and your competitor, who’s winning in the marketplace it seems - very successful, WP Engine; of course, bought by private equity, and so Matt has turned private equity into the enemy… They might be an enemy. I can’t really judge that from where I’m standing. But under all of that, you have a competitor who’s kind of maybe eating your lunch all of a sudden, and you look at them not contributing back, which is just kind of a… It’s not a law. Obviously, it’s open source; you can do or not do. But it’s kind of a social contract, of like “Hey, we’re all building on this common foundation. You should also invest in the common foundation.” If I was the creator of that thing, I would be mad; I would be offended. I’d be like “Come on, guys. You’re eating my lunch, and I’m building the software underneath your platform.” So I understand his motivations 100%. It’s his actions that I don’t understand. And yeah, maybe, like you said, it’s greater social impact of the zeitgeist, as you said earlier.

Jerod Santo: So it makes sense if you want to be a) scalable, and then b) 100% bought into the idea, versus having a founder – you’re a proxy for the founder, in many of these cases. You can make your own damn videos for your own damn products, right? So what are you working on then? What kind of stuff – I know we want to talk about your vision projects, you have some AI stuff going on… What’s exciting to you right now?

Luis Villa: Well, I’m gonna give you my lawyer answer to that. Those of you whose GitHub accounts do things other than commit to other licenses - which is pretty much all I do these days with my GitHub account. We’ll have better notions of code ownership as a cultural practice among programmers, like who’s responsible… I do want to talk a little bit about that one, but let me put a pin in that and come back to it.

The basic system since at least the ‘60s in the US - I’m not sure exactly the timeline in the EU, but I would imagine similar - is that… Well, actually, let me go back even further. Copyright is intended to protect creative works. So what do you have to do to get copyright in a thing? And I’ll explain what copyright is in a second, but let me start with what you have to do. And what you have to do is you have to write down something that’s creative. “Write down” can be broad, right? It can be sculpting, or – but you have to take it out of your head and put it out into the real world in some way. That can be typing it in a computer, it can be, like I said, sculpting it into a sculpture; sculptures can get copyright. It can be a work of art, so it can be an oil painting, or whatever; it can be a Vim poster… Honestly, these days my development environment is Word, but I used to be an Emacs guy…

So that is the key thing, is you are doing a creative thing, and it can be mediated by tools. And Kris, this gets to your point about the AI and where is copywriting there… It can be mediated by a typewriter, or a paintbrush, or I believe - that we don’t really know for certain yet - it can be mediated by an AI. But you are doing some creative something, and turning that into a fixed thing.

Alright, so what happens once you’ve done that? Actually, before I get into what happens once you’ve done that, because I think there’s an important exception… In the US at least, that creative – what does it mean to be creative is not zero. It’s pretty close to zero, but it’s not zero. There’s an important case called Feist vs. Rural Telephone, and the whole thing in that case is literally, telephone books aren’t creative, and so therefore they don’t get copyright. Because what’s the point of a telephone book? The point of a telephone book is to literally just mechanically go through a town and have phone numbers for everybody. So it’s hard work, but it’s not creative. And in the US, at least, you have to have some kind of creative something.

So if you do like a phone list of the 100 most awesome people in New York City - that’s creative; you had to select – one of the ways which you can be creative under US Copyright law is selection. If you pick those 100 people, then hey, you’ve done something creative, and your list of 100 people is copyrightable. But if you’re just “Every single person who lives in Manhattan”, that’s not creative; you don’t get protection. And that plays into questions of databases, and ultimately, I think - and we might not have time to get to this today, but the question of the models themselves. Because there’s both the output of models, what’s the copyright on that, and the models themselves. We don’t actually know if they’re copyrightable. That may be too esoteric; you might have to invite me back for another one for that.

[22:25] But okay, so you’ve created this thing… So now what do you do? So now you’ve got copyright. What does copyright let you do? Copyright lets you control what others can do with it. It lets you decide who gets to use it, who gets redistribute it, who gets to modify it, within the certain limits. But it’s pretty strong.

So the limits include what’s called First Sale doctrine, which is, “Hey, I sold it to somebody. They can usually sell it to one other person.” First Sale doctrine made a lot more sense in the era of like books. That’s what creates used books stores, is First Sale doctrine; it means that I bought the copyrighted thing, and now I can give it to a used books store and they can resell it. In the digital age, First Sale doctrine is a little more complicated… But suffice to say, that’s one of the limitations.

Similarly, fair use says, “Hey, if you’re using this for education, if you’re using this for nonprofit purposes…” I’m oversimplifying a little bit here; the tests around fair use can be a little complicated. Critically, in our digital age, fair use in the US has expanded quite a bit to include what’s called transformative use, which is to say, “Hey, you’re doing something super-new, super-different.” Courts are often going to allow that in the name of sort of not impeding progress.

So for example, Google Book Search is in some sense the biggest copyright violation in all of history, because it’s literally copied systematically millions of books, made these digital copies. But then a court said, “Well, but actually, it’s so different. It’s so great.” And they put strict controls around, you know, you can only get a few pages at a time, and authors can opt out if they want, after the copying has been done… So like Google Book Search is a good example of what transformation means, and potentially analogous to what Copilot is doing. But we don’t know.

The flipside of this is that we just had court cases – we had a court case a couple of years ago about the song Blurred Lines, some of you might have heard… And courts there actually said that even just sort of copying the style of the artist could potentially be a copyright infringement, which was a big surprise to a lot of lawyers. A lot of lawyers are still unhappy about that case.

Next week there’s going to be – or no, tomorrow morning, actually, maybe… There’s going to be a case about Andy Warhol doing – a photograph of Prince that Andy Warhol transformed into one of his Andy Warhol canvases. And the Supreme Court – it’s a little weird, but I think that case might actually have a lot of impact on artificial intelligence… Because we’ve all done, we’ve all played with Stable Diffusion, or Midjourney, or OpenAI, or whatever, to create foo in the style of bar. Well, if bar is still alive, and still has a valid copyright, maybe that’s a problem. We don’t really know yet.

I saw a research paper yesterday that said, “If you prompt Copilot to do code in the style of –” I’m forgetting the guy’s name. Petrov, I think… A top Python programmer - that you actually get fewer vulnerabilities in your code if you prompt Copilot with the name of a top maintainer. And the flipside, the paper’s author was honest enough to note that they prompted with their own name, and the number of vulnerabilities went up. I thought that was nice and humble of them.

So style is an issue that could potentially come up in code as well. That was a very long-winded answer to your question, Natalie. I apologize.

Adam Miller: Yeah, so I knew my ultimate goal was to – it was maybe selfish, but my ultimate goal was “Hey, I want to make the bikes that I want to ride.” I’d made these fat bikes, fat bikes were great, but they weren’t the bike I wanted to ride every day. I was a total bike nerd, and that meant I wanted to ride everything from handmade titanium, gravel bikes, to the ultimate goals, full-suspension, carbon fiber mountain bikes; that’s kind of the pinnacle of engineering, it’s sort of the most popular style of bike within the overall high-end bicycle industry… And that’s what I liked to ride. I was lucky enough to be able to ride bikes all around Colorado and Utah, go to Moab on the weekends and riding in the desert…

A good carbon full-suspension mountain bike is the pinnacle, so I knew in my head that’s what I wanted to create. And I knew I wanted to make the best one, and I didn’t really know what that meant. So I decided “Before I start doing that, I’m gonna make the other bikes I want to ride, which is handmade titanium bikes.” They’re really kind of artistic, super, super-niche. I mean, a small segment of the industry. I found a manufacturer in China that was a mom-and-pop shop, there was about 18 employees there… And I met with them in 2015 when I went over there kind of just to do a tour of old contacts and new contacts in China and Taiwan as I was between my businesses… And I’m at this factory and the bikes were just absolutely gorgeous. And I thought “Alright, I’m going to design some bikes and work with them.”

We built out a whole test lab for ISO certification and testing, and really kind of took this higher level carbon engineering oversight to design these titanium bikes. It was totally my inner bike geek coming out. I was like “I’m just gonna make these bikes that I want, and we’ll sell some, and it’ll be a super-small brand, but it’ll kind of be my way to get my foot back in the door as I’m working on the ideas and the development for the carbon full-suspension bikes.”

We launched the brand, and it was me and – and it was small, and it was supposed to be small. We launched with three models. I found a couple buddies in Utah that I could pay 100 bucks a bike to build them and pack them out, to help me ship them out the door… But it was me with a couple of contract workers for about a year and a half. And it was really fun, because I got to kind of do my own thing and really think through everything… The design of the bikes, my absolute favorite part, and then creating a website, and a brand. I called it Why cycles. It’s a weird name. The idea was not that that brand was ever going to be big. It’s handmade, beautifully crafted titanium bikes that I designed, and then we partnered with this incredible manufacturer to make. They have lifetime warranties… Just super-comfortable, nice, well-made, high-end titanium bikes. And it was an absolute blast.

[38:29] And then at the same time I started getting more detailed about the carbon bikes I wanted to make, and I decided to call it Revel Bikes. I had that name written down for some other naming exercise a few years ago, I always kind of liked it, and so I bought the domain name Revel and started thinking about it. I knew it was going to be carbon full-suspension bikes made for riding in Moab, and all these other kind of iconic mountain bike destinations… But I didn’t want to just make a bike just like our competitors, just like Specialized, or Trek, or Yeti, or Ibis, or Santa Cruz, or Pivot. And so the real secret sauce of any full-suspension mountain bike is the suspension platform; the way the rear wheel moves up and down over rocks.

So really, there’s about four main suspension systems out there in the market; three or four of them have patents on them still, and there’s several other small, small variations out there… But there’s Horst-link, there’s DW, there’s VPP… These different sorts of patents and designs that different engineers have come up with to optimize how the rear wheel moves up and down. Basically, you want your bike to be as smooth and comfortable as possible, just like a car or a motorcycle or anything. Suspension is the secret sauce of all that; it’s how your bike goes around corners, it’s how it goes over bumps and rocks and roots. The difference with a bicycle compared to a car or a motorcycle is that efficiency really matters. You don’t have a motor putting out this power. You have a human who’s pedaling and rotating and putting power down inconsistently on pedals. So if you have a suspension system that’s really smooth and comfortable when you’re going in a straight line, but then you start pedaling and it bobs up and down, you’re losing a bunch of efficiency. So with bicycles it has to be really, really fine-tuned to optimize that human power efficiency, and the comfort when you’re going fast over different terrain.

And all these different companies that come up with their secret suspension system that’s the best of the best, then it’s the best because of all these reasons. But I didn’t want to just go license one of these other patents that all these other companies use, because that’s kind of boring, and then it’s just kind of a copycat system.

Nader Dabit: Yeah. I think it’s important to 100% be clear about this not being some panacea (or however you pronounce that), that there are trade-offs. I think that’s one of the problems that you see in the Web3 space a lot of times, that people are only acknowledging the positives. And I think that that is actually not a good thing. And what you end up having - it’s people that see these discussion from the outside and they think everyone is just full of crap, because all they’re talking about is the good stuff. So yeah, it’s extremely important…

And just to clarify on the number of countries from that thing I referenced earlier… I did a quick Google search, and it said that there are at least 72 countries in the world where women can’t open a bank account, and that was written in 2019… So I don’t know the exact number, but yeah, it’s still a problem. Anyway…

In addition to some of the stuff I mentioned - yeah, I’m really interested, I’m excited about play-to-earn gaming. You’re starting to see a lot of people in the gaming industry that are really high up, like the head of Zynga, the head of YouTube Gaming, and a handful of others publicly talk about them starting to implement these ideas in their games, via NFTs, via cryptos, and maybe via some other mechanisms… But the idea is there.

When you see Axie Infinity going from almost nothing to over two million daily active users, 500 million dollars a month in transactions, 95%of that money sticking with the actual users, not going to the centralized platform - it’s a very compelling case, I think, for a lot of people to take notice.

[51:49] I mentioned accessible and stable currencies - we talked about NFTs for a little bit… We didn’t really talk about them a lot. They’re very controversial. I think that the fundamental technology that empowers and enables NFTs is really interesting. I’m not the type of person that thinks that selling and buying images on a marketplace like OpenSea, which by the way just surpassed Etsy in volume, which means that a lot of other people are interested… But it’s not maybe the most interesting thing to me for that use case. But I do think that what you’re gonna see is people using NFTs as ways to put real-world assets on a digital marketplace and make them accessible to the world, just like DeFi has.

One really exciting example that I am actually diving into and trying to figure out myself and be part of if I can are people that are doing fractional real estate. Let’s say you want to be a real estate owner in New York, but you’re making like 100k a year or 50k a year - you’re essentially just completely locked out. You have no way to really participate in that market. What people are doing now, and it’s already been done in certain parts of the world, but I think we’re gonna see more of it, is that people are buying a building or a real estate via either a DAO or some other type of organization like that… And they’re taking it and they’re fractionalizing it in the form of ERC20 tokens; they’ll attach that piece of land to like an NFT and then they’ll fractionalize it… And then anyone can buy and sell ownership into that piece of property on an open, liquid market.

So let’s say you have $1,000 to invest, you really think that this part of New York is gonna blow up in the next 2-3 years based on you living there and understanding the community… You can actually just take $100 and just buy ownership into that in an instant, and then resell it the next day if you want to… But also anyone in the entire world can do that. And I think that that implementation of what NFTs might offer is interesting.

I also think of the idea around creators and people being able to buy into a community and have some type of community ownership for a few different use cases like fundraising, like people just monetizing their work, and things like that are interesting.

One really good example of this is this artist - I forgot his name, but I tweeted about him not too long ago… He’s actually a JavaScript developer, but he created this generative art using JavaScript, and he created an NFT collection, and he sold over 3 million dollars in NFTs over the course of like a few minutes, and he gave all of that money to the JavaScript community that wrote all of the open source libraries that he built his career upon.

Now, why wouldn’t he just ask people for that money, if he really wanted to do a fundraiser? Well, it’s kind of hard to get people to hand over three million dollars. But when you give them an NFT that they can put and sell on a liquid marketplace, it provides some financial incentive for both parties… Because not only are you just handing over money to this platform or to this person, but you’re getting something back in return. If you had bought one of those NFTs at the time, it’s probably worth maybe double of what you bought it for… But it’s kind of like a win/win situation. If someone wants to help and participate and buy into helping someone else succeed, and then they’re getting something that they can kind of hold on to, a value out of it. And I think that, again, ownership - you’re gonna hear this idea of ownership over and over again… That’s pretty interesting to me.

And another thing around NFTs is definitely the environmental impact, and I think we’re gonna talk about that when we talk about the trade-offs, and also get more clarity as to what that discussion is.

Paul Vixie: I think that that problem would be avoided. Earlier I talked about how in the early days of the web we didn’t have crypto that was fast enough to support – the crypto hardware fast enough to support HTTPS, and so we just sort of didn’t do it… Until you needed it, and then you bought some hardware to assist you with that. And nowadays, I’m thinking about a number of different NIC vendors who make PCIe x16 cards that you can stick into your server, that will do a lot of the offloading for you. Check sum computation, segmentation, reassembly, and so forth.

[00:42:05.27] Now, if you need to operate at 100 gigabits a second, or coming now soon 400 gigabits per second, if your CPU has got other things to do than shoulder every octet through the bus, then you can get a very smart NIC, and driver support for it, and - you know, as we all know, Moore’s Law will continue giving us its annual gift, and the time will come when you don’t need that hardware anymore, and you’re just doing that with your CPU, because you have so many cores, so much cache, and so forth.

So I think any protocol, in order to succeed at all, would have to be the kind of thing – it’d be very difficult in the short term. But in the long run, it’ll just be the way everything works. And so I don’t see that hardware support is going to be called for in any of this. What will be called for is some hard decisions. I mentioned earlier that fragmentation kind of doesn’t work, and it got worse on IPv6. It worked a little bit in IPv4. It doesn’t work at all now. And packet size [unintelligible 00:43:14.10] on our WiFi is Ethernet cells. [unintelligible 00:43:18.20] And that is 1500 octets. And I knew one of the people whose name was on the Ethernet patent. He was my mentor at a mini computer company back in the late ‘80s. His name is David Boggs. And I had an opportunity to listen to him talk about the old bits, talk about being at Xerox and inventing Ethernet, and [unintelligible 00:43:45.20] And so I’m in a position to know secondhand that the intent was the packet size would continue to grow, so that as we got faster at networking, we would also get larger packets. And he was genius in a lot of ways, in this and every day, but his idea about this was every time the clock rate gets you 10x, in other words you go from 10 megabit to 100, to 1000, to a gigabit, or I guess a gigabit to 10 gigs, and so forth - every time you get 10x o’clock, you should probably give about a third of that to packet size, so that the number of packets in a given unit time doesn’t also go 10x. Your packet count and your packet size each go about a square root of 10 [unintelligible 00:44:39.00] And had we been doing that all this time, a lot of things would be simple, that are currently very hard. We certainly wouldn’t hear that fragmentation didn’t work if the packets we could send had gotten larger over time. But they didn’t. And the reason they didn’t is that the Ethernet market relies on backward compatibility. When somebody adds 10 gigabit networking to their office network, they don’t make everybody switch at once. They just say “New ports will be 10 gigs, but the old ones will still be one gig, and we’re going to run a network bridge, a layer two bridge to connect the old one to the new one.” And that won’t work if the packet sizes on the new ones are so big they can’t be bridged backward to the one that made the market exist in the first place.

So Ethernet is effectively trapped at 1500 octets for all time to come. Yes, a lot of us have turned on what we call jumbograms, so 9100 bytes, which turns out to be a very convenient size [unintelligible 00:45:44.00] So if you’re running jumbograms, your NFS is going to be faster, your [unintelligible 00:45:53.15] is going to be faster… Everything you do is gonna be faster. You just can’t use that when talking to people outside your own house, or your own campus. Because there’s no way to discover whether your ISP can carry packets that big, or will at the far end.

[00:46:11.03] So because we don’t have that, because we didn’t do what Boggs inevitably thought was the intelligent obvious thing that everybody should do, give one third of your [unintelligible 00:46:20.29] to packet size, anything we do with DNS in the future is going to have to take that into account. And that in turn means we’ll be making the assumption “Well, I guess we could probably send about 1,400 bytes plus room for all the headers and stuff that was added”, and now we’ve gotta find a way to connect several adjacent packets together, so that we can do essentially application-level fragmentation. Or else we’ve got to deal with the handshake overhead. So I predict, knowing the IETF culture as I do, if we start now, then within no more than four years we will come to an agreement on that sinful issue.

Cory Doctorow: Yeah, sure. I guess it depends on what you mean by the free market. So the term free market really dates back to Adam Smith. And Smith did not mean a market free from regulation, he meant a market free from something that economists call rents. And this is a complex and subtle, but really important idea. And really to kind of get your head around what the relationship rents have to capitalism, you have to read, or it benefits you to read an essay that praises capitalism to the skies, and especially as dynamism and imagination, and its ability to grow and find new value. And that essay is Chapter One of the Communist Manifesto, which is Marx and Engels just going like “Holy shit, capitalism is amazing.” So it’s quite remarkable how many people who get their underwear in a twist about kids on college campuses reading Communist Manifesto, which is also the book from which we get the word capitalism. Capitalists didn’t come up with word capitalism, communists did… They get really freaked out about it, because Marx and Engels are really excited about what capitalism can do. They just want to change the – they want to see if they can preserve that dynamism.

So Smith, when he’s writing about free markets, he is contrasting two different economic systems, one of which is colloquially called feudalism… More properly – historians always write to me when I say this, so I’m going to say this now; properly, it’s not feudalism, it’s manorialism. Manorialism is like feudalism, except if it’s a feudal system, the king can order you to raise an army. So manorialists didn’t have to raise an army. If you had like a land on which the peasants sewed barley, you didn’t have to have a standing army ready to go at the king’s command. That’s manorialism.

[01:09:46.04] So under manorialism, which everybody calls feudalism except for a few weirdos who sent me email, the major source of income was not profit, but rent. So the rent was a sum that was due every single quarter, irrespective of the performance of the asset that was rented from you. So think about like if you own a building with a coffee shop in it, the rent is the same, whether the coffee shop does well or poorly; you are owed that rent. And that rent is distinct from profit in lots of different ways. One of the important ones is that it’s not subjected to competition. So if the peasants on the next plot of land over - and I should say here that entrepreneurialism and feudalism peasants were bound to the land. You weren’t allowed to leave the land you were born on; you had to work it and pay rent to the lord until you died, and then your kids had to do that, and their kids had to do it. So you’re a captive tenant. When we say tenant farming, that’s what we mean; tenant farming.

So these rents were not subject to competition. If the tenant farmers on the next lord’s plot did better, you didn’t do worse. Your peasants didn’t go out of business. If they did better, I guess you could hike the rent, but you didn’t have to. And you didn’t own the means of production. The landlord doesn’t own the espresso machine in the coffee shop, and the landlord who owns the land doesn’t own the grain threshing apparatus. They just own the land. They own an asset that other people work with their capital. And in this case, the capital would be an espresso machine, or a scythe, right?

The capitalists hate the feudalists, because the capitalists want to turn the serfs off the land. This is a process called proletarianization. So they wanted to deprive them of the guaranteed income, but also liberate them from the forced labor, and say “Off you go. Figure out a way to make a living.” They want to rent that land and use it to grow sheep, which will be inputs to an industrial process, the textile mills. This is the Industrial Revolution. And then the fact that these workers have nowhere to go means that they’ll come and work in the factories, and spin the sheep’s wool into textiles. So that’s the process.

Now, the more the rent is on the land where the factory is, or the land where the sheep is grown, the more peasants are bound to the land and required to pay rent, the less money the capitalists get, and the less dynamic the system is. Feudalism is static. The way that you grow your share of the income as a feudalist is you invade another country; that’s why feudalists had to have a standing army. You don’t get it by being a better lord than your neighbor. The requirement for being a lord is emerging from a very lucky orifice. It’s not like being good at business.

And so capitalists, they are really subject to competition, because the thing that creates this dynamism that Marx and Engels are geeking out about is that capitalists are fighting with each other to find ways to increase productivity, to generate more dollars per hour, per worker. And the capitalist that do best put the other capitalists out of business, and absorb their market share. And so there’s this constant churn, where capitalists, they can’t sleep. This is Elon Musk’s hardcore work environment. You’re sleeping on your desk, because if you miss 15 minutes in your commute the next day, that’s 15 minutes that your competitor who sleeps under their desk is going to use to steal a march on you and put you out of business. So capitalists are always competing.

Now, John Steinbeck supposedly once said “Socialism never took hold in America, because workers see themselves as temporarily embarrassed millionaires.” And no one can show where Steinbeck allegedly said that, so he probably didn’t… But also it’s probably not true. There’s certainly like a pretty good labor history in America, which is also enjoying something of a renaissance right now. A lot of people who don’t think of themselves as temporarily embarrassed millionaires. But it’s totally 100% true that American capitalists see themselves as temporarily embarrassed feudal lords. That the goal of successful capitalist firm is to cease being capitalist, and become a rentier; someone who collects rent. To own AWS, not to use AWS. To own Amazon, not to be a seller on Amazon. To provide the Uber app, not to drive an Uber. These are rent extraction systems. 30 cents on every dollar for every app, it’s because I own the app marketplace. I don’t care, I don’t have to. I’m the phone company. You have to pay me rent.

[01:14:26.01] And so we have lived through this period of about 40 years, in which rents rather than capital, rather than profit, have been ascendant. And the difference between a feudal and a capitalist society is not whether there’s only rents or only profits. There were profits under feudalism, and there’s rents under capitalism, obviously; there’s landlords. The thing is what happens when rents and profits come into conflict. If your productive company gets hauled into a patent court in the East District of Texas by some patent troll, who’s pulled the wool over the USPTO’s eyes by saying “I have a system and a method for doing stuff with computers”, and then they sued you for $300 million, and you have to give them $300 million - that is the triumph of rent over profit. They make nothing but lawsuits, you make things that generate the $300 million that you give them for the lawsuit.

And so when rents are ascendant over profits, as they have been in a steady trajectory for 40 years, we tend to forget that rents aren’t profits. And of course, all the rentier companies are also profit-making. Google makes phones, Apple makes phones, Amazon does a whole bunch of stuff that is stuff; they make movies, they make whatever, right? They have stuff, they have coders, they make things. But the majority of their income comes from rent. Amazon Prime is actually free for Amazon to operate, because the 51 cents they take out of every dollar for every dollar made on the platform by third-party sellers pays for 100% of Amazon’s own shipping. So all the shippers that have to use Amazon, because they’re the only – you know, if you don’t sell on Amazon, you might as well not exist… All those shippers have fully-subsidized Amazon’s own use of its logistics system. Amazon uses that for free, and also competes directly with those companies, which is nice work if you can get it, right?

And so if you want to live in a capitalist society, one in which innovation and excellence and productivity gains are the biggest predictor of your success, rather than the orifice lottery, rather than just like claiming the land before anyone else gets to, rather than being the first person to build the cloud that then everyone gets locked into by their tooling, and their other switching costs, and then you capture the regulators so that making interoperable tooling is impossible… If you want to be Unity, where you just rely on the fact that reverse-engineering Unity and making a runtime that would allow people to recompile their games to run without your engine is a crime, and so you can now start extracting rent from people who make games with your engine, you can just like do Darth Vader NBA shit, where it’s like “Yeah, I’m altering the deal. Pray I don’t alter it further”, then that’s feudalism. If you want to live in a capitalist system, where Unity doesn’t get to swoop in and like bankrupt you overnight, because they decided that you owe them rent on a thing you already bought and thought you owned, then you have to eliminate rents.

Austin Parker: So just by way of background, I used to work for a company called LightStep, that a couple years ago got bought by ServiceNow, and it was actually super-interesting… I stayed there for a couple of years, and it was super-interesting getting to talk to ServiceNow customers. Because if you’re familiar, ServiceNow is a big enterprisy place, and it sells to big enterprisy places. Fortune 100, massive multinational orgs. And talking to them about observability was super-fascinating, because they would be talking about like “Oh yeah, we have incidents –” It’s like, their incidents would run for weeks. It wasn’t like “Oh, this is an incident, and we fix it in an hour or a day.” It’s like “No, this has been going on for like two months.” And I’m like “Whaaat?”

Drew Wilson: [00:49:48.07] It ran out of money. It ran out of money. That’s the practical reason. The real reason is, is we approached the market incorrectly, and we didn’t have the cash to correct. By the time we learned how to correct… And anyone who’s in that industry, they know this stuff just takes forever. These feedback cycles take forever. So if we had landed our series A, we would have been good.

So we raised a pre-seed, and then I raised a seed round, and then I ended up spending all my money, way too much money on this bridge round, essentially. It wasn’t really a bridge round. It was more just an open access to Drew’s bank account to pay for the company, to try to get to this A. So what we did is we had this investor - super-cool guy; became a billionaire a couple of years before we started, to the tune of like 4 billion bucks… And had a bunch of money, started buying up all this Western-themed stuff. And by Western, I mean like horses and cowboy stuff. So he bought up Country Magazine, all the country stuff… He bought up all the American rodeos, he bought up all the cable programming for Western-themed stuff, he was building a Western-themed movie studio, because he’s a former movie studio exec… And so somehow I got to be on a Zoom with him, and pitched him what we were doing, and he was into it, and wanted to invest… Mostly because he wanted to create a Western-themed bank under the brand name that he’s buying all this other stuff under. So we’re like “Yes, we’ll do that. We’ll white label Letter and we’ll have this other thing for you, too.”

So that was the deal. He was running it all himself. And then he brought in a team to run that brand that he had made, because he didn’t want to run it anymore. And the team he brought in was headed up by the former CEO of Hulu, the guy who sold Hulu to Disney. And so then I was interfacing with him and his team, and their idea was to create an Apple TV Plus or a Disney Plus, but for Western-themed content. Well, it turns out they couldn’t secure anything; literally anything. They couldn’t secure anything. So he’s like “Hey Drew, we don’t know what we’re going to do here at this brand… But we do know that it’s not going to involve a bank if we can’t figure out how to make money. So sorry, we’re no longer going to fund your bank.” And so this was me trying to get this deal across the line for like seven or eight months… And it was just like, it was always right there. We had paperwork sent to them, they would revise, send back, and just like… I’m like “It’s so close. Just sign the freaking paper.” But then it never happened, and that was early 2023 when they pulled out. And that is the exact same – the same month they pulled out, it’s the same month as Silicon Valley failed, and everybody started pulling money out of there. And we felt that at our startup bank too folks weren’t super-comfortable with the startup bank anymore, and moved their money around.

So yeah, we got kind of hosed there… But we did get to use that guy’s money to lend to people. We created our own in-house lending platform, where you get a loan in a few seconds, and it was 100% programmatic, meaning there was no humans in the loop at all… Which is crazy. And I don’t know of any other thing or program that’s done that, but we did that, which was fun. So that’s why it didn’t work.

Samantha Coyle: Yeah, that’s a good question. So the thing about the Wait For It - I want to say Go package, because it just comes very natural to say… But it’s not technically a consumable package. And so that’s a thing with this repo, if you do choose to use the Wait For It package (I guess, I don’t know), it’s only using package main, which for all of us gophers that means we can’t consume it, unless we modify it. And so what that meant for this project in particular is, unfortunately, I had to copy-paste their code.

So it was nice, because we’ve got all of this wait logic, and of course, giving proper attributions to the author who originated the repo, but it was really unfortunate, because it wasn’t the Go natural way of creating open source packages such that other gophers can consume them, and contribute, and so forth. So that had a few side effects.

I mean, for one, it meant extra code for our team to maintain, and of course, modify, as we found little things here and there.

So one of the big things that I did was I copy-pasted in the logic, so that we could bootstrap all of our main.go’s and all of our services to work with the logic. And as Neethu said, it was copy-paste modify, and then a minor modification to all of our server’s structs to define all of our dependencies. So we can say, “Hey, wait for this service, wait for that service.” So it made it really clean, but it made it to where I also had to add a wrapper around their logic, so that way it was more idiomatic to look at and to consume.

[21:56] So I think they just had functions called wait, and so I added a wait package, naturally, and I added a “4dependencies”. So it was wait.4dependencies… Which is very natural and very easy to understand what’s going on. So a few modifications and one other learning.

And this was so funny, because – so I brought in the logic, added my wrapper, and my additional method, so it was pretty and nice to work with. And then our tech lead and Neethu, they were like” Hey, what about an error case? What happens if the service never becomes available? What do we want in that case?” And so I think it’s really interesting thinking about “Do we retry? How often? What makes sense for our use case?” And everyone’s different…

So what I found out in just bringing over the logic was that I didn’t translate that aspect from the repo. So with the initial bringing in the wait logic for our services, it just hung… If your depending service never became available - and I don’t think that’s something you necessarily want, especially in, say, a production-ready environment… So I actually had to go back and add that timeout logic, because in the Wait For It repo that’s on the CLI side of things; not necessarily translating 100% in the wait logic that exists right now.

Derek Collison: I think that’s a great question… And like I said, I took a stance for probably almost 20 years, of “Oh no, just use whatever you’re gonna use.” I really wouldn’t even engage with people, because it would be a two-hour conversation. I would still probably lose the argument at the end, so to speak.

What’s happening now is that things are kind of coming back to it. So just to level-set - one of the things that I care deeply about with messaging systems is a couple things… And it’s not about messaging, especially not about the message broker from the ’90s type stuff; we need to not think of it like that. We need to think of it as a ubiquitous technology, kind of like network elements.

I was around when there was a single cable, and if you ever kicked the terminator cap off the end, it took the whole network down… And I saw the birth of hubs, and switches, and smart switches, and top of rack, and all the crazy network elements that we have… And just as well, a modern system needs to be doing that similarly. But to start out with - the first thing it does is it says “I am gonna do addressing and discovery not based on an IP import, but on something else.” You can call it a channel, a topic, a subject - I really don’t care what you call it.

Now, again, for a while, people were like “Why? This doesn’t make any sense.” But in today’s cloud-native world, I would argue what everyone is doing today makes no sense. We struggled so hard to change servers from pet to cattle, and yet we’re still saying “Oh, I wanna talk to Johnny’s service, so I need to know the notion of an IP port.” Now, I know what the audience is probably thinking, and I’ll walk through an example of what this really looks like in practice.

The other thing too is that messaging systems, when they do that abstraction, a lot of people call it pub/sub. And we still call it pub/sub, but again, we’ve gone away from that, because it’s kind of got a bad rep. But what I mean by pub/sub is that if the technology can understand multiple messaging patterns - one to one, one to N, m to N, and then one to one of N, meaning I can automatically say “Hey, I wanna send a message and only one of you in this set will receive it.” That’s kind of what NATS does at the very basic levels…

And folks always ask and they go “Oh, I just don’t need any of that stuff”, and I say “Okay, a couple things… One, decoupling is good.” Pets versus cattle was legit; let’s make sure our connected technologies follow the same path, and don’t say “No, no, this one’s special”, whether it’s an individual server, or a load balancer, or a special host… It just makes no sense in a modern world. So we push down on those things and we say “Okay, got it.”

[16:18] The last piece of advice I always give people from the ‘90s is “Never assume what a message…” - and a message could be a request, it could be data events, all kinds of stuff. But “never assume what the message is gonna be used for tomorrow.” So everyone kind of looks at me and says “What does that mean?” and I say “Okay, I’ll give you a really simple example.” When we talk about NATS these days, with a very fortunate, growing user base, and all kinds of crazy interest from customers and clients these days, is - modern architectures, distributed architectures are built using connected patterns, and there’s really only two. There’s lots of nuances to it, but there’s two. It’s either a stream or it’s a service.

The service is I ask a question, I get an answer, and a stream is I’m just sending a piece of data, and then I don’t care. And to level-set, distributed systems, even up to a couple years ago, were dominated - 98%+ of everything was a service interaction. I’m not saying it has to be synchronous, but everything was “I’m asking a question and getting an answer.” HTTP is “I ask a question and I get an answer” type stuff. So I said “Fine, on day one you know who’s gonna be answering that question.” So you coded up so that I’m gonna send a question to Mat, Mat’s gonna respond back with an answer. And you’re doing it on your laptop, and you use HTTP, or gRPC, or whatever, and you’re like “That’s all I need.” I go “Great.”

Let’s not even get to the point of anyone else being interested in the message. Just to start with “Okay, let’s go to production.” Well, we need more than one map. Oh, crap. Now we need a load balancer. Well, now we need to put stuff in, and do DNS… And that’s fine; production can handle that. I don’t have to worry about that. So then they do health checks, and then they have to figure out rerouting, and all this stuff… And all these big companies have playbooks on exactly how they do it, and they all look very similar, but they’re all slightly different.

Now let’s say someone says “Hey, for compliance we need to be able to watch all requests and we need to record them.” Now all of a sudden you’re gonna have to put a logger, and you don’t want the logger in place of a request-response, which is a massive anti-pattern I see being proliferated these days. It’s like “Oh, no… Put a messaging system in between it, and store it on disk, and try to retry, and stuff like that, in line with the microservice, in line with the service interaction. I’m hyped up on Red Bull, but that’s the dumbest thing I’ve ever heard. It’s like Google saying “We’re gonna write all the log entries before we return your search results.” It’s just foolish, no one would ever do that. But there’s a need to say “Hey, I need to be able to write these things down, and someone else is gonna look for anomaly detection, or any type of policy enforcement, whatever it is.

So look at that system when you’re getting to the end of the day, and now let’s say we actually wanna spread it out from East Coast to West Coast, to Europe, and you need Anycast, and DNS, and all kinds of crazy stuff and coordination on the backend of state. They become really complicated, and they’re trying to get around the fact that everything is just point-to-point; it is naturally a one-to-one conversation. Whereas with a messaging system, you write it, you run it in that server, let’s say, or whatever, but think of the NATS server is extremely lightweight. It can run on a Raspberry Pi, in your top of the rack switch, you can run it in your home router, you can plug and play and put these things together in any arbitrarily complex topology that spans the globe… And that’s another discussion on distributed systems that aren’t really distributed.

So you set it up on your laptop and you run the NATS server. Its Docker image has been downloaded 150 million times… It just runs. For example, a subsystem of GE doing nuclear stuff runs our server for two years at a time, with no monitoring, no anything. And when they come in to change things inside of that nuclear reactor type stuff, they shut it down and figure out if they wanna upgrade it or not.

[19:59] So it’s lightweight, it’s always there, it’s ubiquitous, it just works. So now all of a sudden you write the same program, you’re sending a request, getting a response, doing it on your laptop… I would argue you’ll take the same amount of time, or possibly less, but it’s on the same level. You do have to run the server.

But now when you go to the next level of production - I need more mats. Well, just run more NATS – I mean, maths, not NATS. “Oh, well do I have to put in deployment a framework like Kubernetes and a service mesh?” No. I don’t care how you run them. Run them on bare metal, in a VM, in a container, in Kubernetes… It does not matter, run them anywhere. The system automatically reacts.

By the way, you haven’t configured anything on a NATS server yet, ever. Now, all of a sudden you’re like “Okay, well what happens if I wanna do compliance and I need to watch all the requests coming in?” Just start a listener for it. It’s got all of those patterns built in, so there’s nothing changing between me who’s asking the request and Mat who’s giving the response; we have to change. As a matter of fact, we don’t even have to be brought down and restarted; we’re just running along and people can come up and bring up anomaly detection, logging… All kinds of stuff.

So as you keep making these systems more production-ready and complex, you realize that messaging gives a huge win over the other ones. Now, the other ones are kind of known. People know how to put up load balancers, and know how to do logging, and know how to do all this stuff… But when you see something running on a service mesh and you haven’t even sent a single request check and you’re spending $6,000 a month… And I can show you, we can do 60,000 requests a second, with all of the servers latency tracking, all transparently, doing it for real, and it runs on a Raspberry Pi… That also translates to OpEx savings, which is a big deal.

NATS has always been known for how fast it is, but most people tell me “But we don’t need to go that fast. We don’t need to have a server doing 80 to 100 million messages a second.” And I go “I know”, but if you think about it, if you take that same thing for your workload and put it in the cloud, you can save 80% on your OpEx budget.

So do people need messaging systems to build stuff? Of course not, because everything for the most part is built on essentially HTTP, which again, is an interesting one to me, unpopular opinion… But I know why we did it that way, and we don’t have a reason to do it that way anymore and get it stuck, right?

The notion of client-server or request/response in the old days was the requester and the responder were usually inside the same network firewall… Not firewall specifically, but essentially we’re inside the company. And everyone started to say “Hey, I want the requesters, whatever those things are, to be able to walk outside of the corporate network.” So all of a sudden people started doing this, they go “We can’t get through the firewall”, and the firewall people, or the old DB people, they go “No.” It doesn’t matter what you ask them, they say no.

So people kind of went “Wait a minute… Port 80 is always open. We can piggyback off that and circumvent this whole thing. So we can just do request-response on HTTP or HTTPS, and it works.” And it’s true, and I remember doing some of those tricks myself. We’re not in that world anymore; that makes no sense whatsoever to build a modern distributed system off of technology that existed for something totally different and was a workaround around security and firewall systems.

Break: [23:23]

Bryan Liles: Yeah, actually. Since I’m on an open source podcast, I wanna talk about my open source. Actually, it’s really relevant to this whole thing. I’m working on a piece of software called Octant. I wrote it over the last year, and the long story short is that we wanted to solve the problem of how can a person who is not familiar with Kubernetes sit down and understand what’s going on with their workloads? It’s a really heavy topic. So I created a proof of concept last year, and we went through and we got it all approved, and now it’s a piece of software. But that’s not really what I wanna talk about.

I wanna talk about the power of open source, and this whole concept of a 10x developer. I like to think of myself as a 10x developer, and I can just imagine people stopping this conversation and they’re like “That guy Bryan - what is going on?” So this is why I think I’m a 10x developer. And no, I’m not telling the binary joke; that’s not even funny.

My job as a developer is to inspire others to do things and to write software that allows others to do things. But the real testament of a piece of software, whether it’s good software, or even mediocre software, is if someone can take your software and do something with it that you’d never even thought of doing.

The reason I brought up this piece of software of Octant, where you can actually go and view your Kubernetes clusters, and we’re putting more features in there, like terminals and all sorts of things… But what I’ve found is that I’ve put a plugin system in there a few months ago on a weekend - because why not? Plugins seem pretty cool… But being at this conference this morning, I found 4-5 people that came up and said “Oh wow, your software is pretty cool”, and I’m like “Yeah…” But I’m like the developer who’s saying “Well, I could have fixed this, I could have fixed that…” They’re like “No, no, no, no. It’s amazing. And I wrote a plugin using your plugin system, which is crazy, because we didn’t document it.”

[16:04] So people are creating software based on software that I wrote, so I’ve actually inspired people to do things, and someone has come and created something that I could not imagine with my software. That’s what makes me a 10x developer. I’ve made more than ten developers more productive, so guess what - 10x. I’ve 10x-ed myself. And that’s what I actually wanna leave everyone with - this whole myth of the 10x developer; your job, if you can make two developers more productive… So if you have two times two, that’s four, so guess what, you’re now a 4x developer. So if you can do it for 10, or 100, or 1,000, or in some cases at high levels, like a million, because it’s possible - that’s how you actually scale yourself.

And here’s the crazy thing - am I a good developer? I think so, because I practice. Am I the best developer? Probably not, because someone’s practiced harder than I am. But we can still all be developers who are multipliers, and open source is actually one of the greatest ways of doing that, especially whenever you’re using permissive licenses and you’re allowing people to build things that aren’t even possible without your software. And that’s crazy. So I just wanted to leave everyone with that, and that’s actually why I do all of this. I’m here to inspire on the stage, but I’m also here to inspire in software, and just be inspirational all the way around. So that’s my positivity moment for the afternoon…

Ryan Singer: There’s a sense of – you know, if it didn’t ship at the end of the six weeks, something went wrong. That’s not normal, that’s not business as usual. And then we’ve got two weeks of cool-down at the end of every six weeks. That originally started because we recognized that we needed some time to just look at bugs… Because there’s a certain amount of bugs. Most bugs don’t matter, and bugs aren’t special, as David likes to say. There’s just other work that you could maybe do or not do. They’re also, in a way, just the edges of what works.

[28:02] If the core of what works is valuable, then the bugs are just there sometimes; it depends on how competent the execution of the core flows is. We assume always, because we have competent people who have enough time to think, that the core things are properly done… But nonetheless, having that two weeks of cooldown between cycles gives programmers totally – they’re not guided with what to do at all. They’re totally free to just pick off that thing that they wish they had done differently, that bug that was brought to their attention that they would like to solve, the new library that they might wanna use but they haven’t had a chance to tinker with yet etc.

Then also those two weeks of cooldown give us an opportunity to put heads together and meet about what to do next. They give people a chance interdepartmentally - which is a goofy thing, because we’re a fairly small company… But still, when does the programmer get to talk to the data person, if they’re heads down for the six weeks trying to build something? So that’s cooldown.

During the six weeks, a huge part of the notion of that cycle is that it’s uninterrupted time. So the team is given a project, and they’re totally left alone, to do it however they see fit, and to organize themselves how they see fit. So that’s the basic structure of the cycle. And then if you have that in place, then questions arise - how do we put work into a cycle so that we can actually have confidence that it’s going to ship? This puts us into a risk mindset. If I give a team work that is totally underspecified, like “Go build a calendar”, a calendar could do 100 things. Which tenth of a calendar should they build first? They could build the best ever RSVP and invitation and schedule alignment system; kind of go down like a doodle direction, and make that the best ever thing… They could make the best ever high-fidelity native interaction for drag-and-dropping events on a monthly grid of cells. Where do they go?

So if you underspecify the problem, your risk is huge, because you’re giving way too much open space to the team. On the other hand, if you overspecify the problem and you say “You will do this task on Monday and this task on Wednesday”, then it’s all gonna blow up in your face, because you don’t actually have the information to make that judgment upfront… Like we talked about with the imagined versus discovered work.

So the challenge then is “How do we shape the work in such a way that everyone can see what the rough, basic solution is, without totally being specified? And how can we address what might be the major rabbit holes in that work, before we commit to it and give it off to a team?” So you don’t wanna give a team a ticking time bomb, or a hole they’re gonna fall into… So there’s a certain amount of high-level design that we do, and a certain amount of technical stress-testing. “Architecturally, does this make sense? Do the APIs that we expect to use actually exist there?”, this kind of a thing. Performance-wise, if the designer has an idea that this project is gonna work like this, but then somebody who understands how all the jobs are gonna have to get queued could say “Oh, that might be an issue”, whatever.

[31:46] So there is this pre-work, and the pre-work that happens. Before work goes into a cycle, before work is kind of qualified to go into a cyle is this shaping work. And this shaping work isn’t done by everybody; it requires a cross-discipline melding of “What is strategically valuable? How do we figure out the main affordances and moving parts and flows from a customer-facing standpoint?” And then working backward, can we actually envision the technical solution enough that it doesn’t feel like we’re throwing a big mystery over the wall to a team, and expecting them to solve it. So that’s the shaping.

Then the work that we do in the shaping phase - it doesn’t automatically directly go on to some kind of a conveyor belt or queue. There’s no backlog. Just because you shape something doesn’t mean that now it’s gonna go to a team, rather it means that you’ve done some value-add work on a raw idea, on an unshaped idea, so that it’s now a better potential bet. So you’ve improved your odds on the success of shipping this thing… But there’s still a separate question, which is “Is that thing strategically valuable for us to spend time on right now?” Or is the amount of time that you shaped it for.

Let’s say I’ve got some idea for some group notification thing in Basecamp. I could shape up a six-week version of that, and then put it together into a pitch and be like “Here, here’s something I think we should do. I don’t think there’s any rabbit holes in it. I think it’s clear how the team could solve it. Let’s do it.” And then we could have some push and pull, and somebody else, Jason or David might say “Eh… If we could do that in a week, I would gladly ship it… But I don’t wanna give up all these other things that we’d rather be doing for the next six weeks to get that.”

So you have this question of like “What actually matters?” That’s where we get into betting. Betting is very different than planning. Planning has a general connotation of certainty, and betting has a general connotation of risk. With betting, we wanna have the notion of “I know how much I’m betting, and I should have a capped downside and maybe an open-ended upside. At the most, I’m gonna bet six weeks on this thing and see what comes out of it…”, but I don’t wanna bet six weeks on it and then all of a sudden it turns out that I bet 2x, 3x, 4x that much time on it. That’s not a bet, you know what I mean? That’s like just getting screwed.

Todd McLeod: Yeah, I really believe in minimalism and stripping things down and keeping everything as light as possible. Every line of code, you should know what it’s doing and you should not have any extraneous code. If it’s there and you’re not using it, why is it there?

That’s actually one of the design practices, idioms of Go - you can’t have a variable if you’re not gonna use it; that’s an error, right? And I love that about Go. It’s similar for me in everything I approach, like in my house - I don’t wanna have stuff around if I’m not using it. Pass it on, donate it, let somebody else use it.

With HTML and CSS, I’m not a fan of frameworks and libraries. Yeah, they get a lot of things done for you quickly, and if you understand the fundamentals of HTML, CSS, JavaScript enough to be able to go in and know what they’re doing, and tweak them and adjust them when you need to do that - fantastic, use them all day, because it’s a tool and there’s a place where you can apply it. But if you just learn bootstrap, if you just learn jQuery, and if you do that at the expense of learning HTML and CSS, or at the expense of learning JavaScript, then I think you’ve really missed - I strongly, firmly believe you’ve really missed - a super important layer in your educational foundation to be competent and skilled developing applications on the web.

So if you look at metrics, and I’m really into building websites now, and we’re putting together a website ourselves, which I’m very excited about, and performance is so key, it’s such a key metric… When you look at studies on performance, like Amazon lowers their time by one hundred milliseconds and sales go up 1%, it’s like, “Wow, people are super sensitive.” Performance is really key, so the average size of a web page today - 2,200-2,400 kb, right? And the site we’re building right now is 400-600 kb each page. So we’re like one tenth, one-sixth the size of average sites. How do we do that? We don’t have any jQuery, we don’t have bootstrap, and when we want to put an image onto the page, we’re not using Font Awesome. Because if you download Font Awesome, that’s 100 kb. That’s one twenty-fourth of your size budget for that page if you’re trying to hit normal sizes today.
Instead, we extract SVGs, run them through Illustrator, get that SVG and then stick that code in. That costs us maybe 7 kb, or 5 kb on that page for those images.

[23:56] So just keeping things lean, and having complete control over your codebase, and a thorough understanding of your codebase - to me, that’s what it means to be a craftsperson. This is very much a craft; it’s the craft of building software. And just as somebody like a woodworker, they would know all the tools in the wood shop, and they would know the different characteristics of the different materials they work with. And when you first come into a craftsperson’s workspace - like a woodworker’s workspace - you might be like, “Holy cow, there’s a lot to know about this craft! There’s like a hundred different lays, or all these different little chisels and files, and where would you use each one and why, and which material would you choose for which situation…”, but after time you become that craftsperson and you have that intimate knowledge of all of your tools and all of your materials, and you can really start to build beautiful, well-built, performant applications.

So that’s my approach - just to do things well. When I looked at server-side languages, different languages approach things in different ways, but Go seemed to me to be the one that was the most zen - coming back to that word, zen - and the one that is the most fundamental. You’re working down at a low level and you understand everything that’s going on. You’re actually getting that request and parsing that text and you can build your own ServeMux if you want. A lot of that is abstracted away for you if you use the net/http package in the standard library. But you could grapple it, you could wrap your head around it and you could see what’s happening there.

I don’t know if that answered your question, Carlisia, but being a programmer’s language - and I don’t have the depth of experience with Ruby, I never did any Ruby, but from what I’ve heard of Ruby, it’s like there’s a lot of magic. Also, from what I saw at Python and Django, it’s like “Hey, put in a couple of lines of code, and KAPOW! in 15 minutes we have an entire CRUD application”, and it’s like, “Yeah, but what happened there?” How much ability do you have to customize it? Whereas if you learn web dev with Go, then you need to understand that the internet engineering task force are the people who put together the recommendations for how the internet and the web should be deployed and built, and they have these requests for comments, and there’s been a variety of requests for comments in this entire history of how the web’s been built, the 7230s up in there - those are the current ones - and then here are the specs, this is what happens… And then you could take that text that’s coming in and you could work with it, pull out the pieces, pull out the method, pull out the route, and you make decisions based upon that, do a little conditional logic… That’s what I mean about the programmer’s language.

Sometimes I say “Go suffers no fools!” Kind of like the Pirate Ride at Disneyland… [laughter] Deeeeead meeen teeell nooo taaaales! Goooo suuuuffers nooo fooools! Because I’ll hit these walls where I’m like, “I have no idea what is happening there”, and Go has humbled me so many times. It brought me to my knees with hours spent contemplating the simplest things. In retrospect, I look at them and I’m like “Okay, I get that now, I understand it. It’s simple now”, but at the time I’m like “What the heck is happening there with StripPrefix? How do I get StripPrefix to work and serve my files?” Hours spent on that…

So it doesn’t suffer fools. I mean, it’s got [unintelligible 00:27:42.12] it’s got a slice to bite. I mean, it gets down to all of your data structures and everything, and … I just gotta ramble a little bit more about Go - this is one of the things I love about Go: innovation. Technology is about innovation, and for me - that’s my perspective, but that’s the heart and cornerstone of what programming technology is about.

[28:10] A lot of people, as I’ve met people learning the language, they’re like “How come it doesn’t have generics? This isn’t OOP… Where is the inheritance?” [laughs] I’m like, “You’re a programmer! You create things that innovate. You need to innovate, too”, and I feel like that’s what they did with Go.

They looked at, “Hey, this is how programming’s evolved - spaghetti code, procedural, structured OOP”… They looked at research, like Tony Hoare’s CSP, and then they said, “Let’s integrate all that and do this new thing.” When I first started to learn Go, people were like “Go’s all about types. It’s all about types.” I’m like, “I don’t understand that yet.” But now I’m like, “Oh, it’s totally all about types!” You create your own types. I can create my own type “hotdog” if I want, and the underlying type can be an int, and then I gotta attach a method to that. And interfaces…

I feel like Go innovated and took programming and tried to bring it to the next level. Whatever name you might put on that, but from spaghetti to procedural structured OOP to the next evolution of coding. So it’s a programmer’s language and it suffers no fools, because you gotta know what’s going on, and to really use it you have to understand a lot of the most important concepts of programming.

David Heinemeier Hansson: I have. Curiously enough, there’s been occasion to ponder the model of open source recently… The whole WordPress drama, of course, was something I think brought a lot of folks to think about “What does the license actually mean? What about the norms?” This is one of the things we talk about in other realms of governance - how much is written in law and how much do we depend on a set of norms that we all just take for granted? And I think we had a bunch of norms that I took for granted about how open source collaboration works, that were - let’s just put it politely - challenged in the WordPress drama. Things like retaliatory actions when it comes to code repositories, and so forth. When I suddenly think “Oh man, Rails really depends on RubyGems being an open, neutral platform, that is not engaged in kind of trying to push its own advantage in these ways.” And the BDFL model I think too does hinge on a single individual. That is both its greatest strength and its greatest weakness. And we have to accept that those things come together as a package. There is no such thing as just a universally better governance model that does not have trade-offs.

The committee model has also been tried. You know what? I’ll hold the accomplishments of the BDFL model against the committee model any day of the week when it comes to open source progress. If you think of most of the major projects that we all used and love, they have run under the BDFL model. And I’ve seen what happens to projects that get [unintelligible 01:12:47.12] down in unclear vision, unclear direction, a thousand cooks all pulling in their own direction… It’s not great. We don’t get the kind of software we want to use and we enjoy to use. A lot of that is downstream from a Coherent, uniform vision. And that just happens to be the model you get with the BDFL setup. Again, you should look at that and go “Does that mean it’s always going to be flawless, or that we are never going to have WordPress scenarios?” No, of course it doesn’t. In fact, if anything, what’s remarkable about the WordPress drama is how novel it seems.

Why hasn’t this been happening every six months all over the place in projects left, right, and center? Why is this one of the first instances of a project of that size getting embroiled in these kinds of debates? That to me, more than anything, is kind of one of those exceptions that prove the rule. And the rule is the BDFL model is amazing. And it works, if not flawlessly, then exceptionally well the vast majority of the time.

[01:13:55.07] And we don’t have to speculate about this. We have literally three, four decades worth of prior art telling us exactly how well it works. And I think that’s actually the most mind boggling thing to me about all of this. It’s like, why? Why does it work? Why have we not had more spectacular failures? And it makes me appreciate that that was the way open source went. That this model that allowed open source to be competitive with closed source software to a large degree was downstream from exceptionally driven individuals.

Now, I think a lot of great open source is great now, because it’s more than an individual. Rails has near 7,000 contributors to the main codebase over the two decades. I mean, that’s a remarkable achievement. Although we also did just pull out an example that literally only has three contributors over 20 years. SQLite. So maybe that’s not the be all end all of it. But the initial inception of almost every great piece of software sprung from one person’s mind. And I think that’s just something I’m very willing to accept. I’m willing to accept that there’s no governance model that’s 100% effective.

Peter Cooper: Yeah. Sometimes it’s not even just that; it’s also that perhaps they just don’t like the way that you present stuff. They don’t like the layout of your newsletter; they just have a gut feeling about you, and they’re like “Oh, I don’t like this guy. I’m gonna unsubscribe.” But you have to be okay with that. If you do some tweets and you mention about politics and someone unfollows you and says “Oh, you should do developer tweets all the time.” Like, it’s actually not wrong for them to say that, but it’s also not wrong for them to leave, and it’s not wrong for you to be like “Well, I don’t care.”

I think we ought to be a bit more forgiving of each other and a little bit understanding that we can’t satisfy everyone. That’s definitely true of us. This is one of the points I put down in my notes - you can never fully saturate an audience. I can never reach 100% of all Go developers. That’s just not gonna happen.

[52:10] I encounter people that – say the JavaScript world; we’ve got 170k subscribers, and I run into people at conferences, and they’re like, “Oh, I’ve never heard of JavaScript Weekly.” And it’s like, I don’t sit there thinking “Well, that’s a bit weird… Surely, everyone knows about us by now”, because there’s millions of JavaScript developers. There’s just all these different pockets of different places. JavaScript developers don’t even know about each other half the time, let alone all the different resources are in the community.

So there are definitely Go developers who have never heard of this podcast, and it’s never even crossed their radar… And I kind of wonder, do we need to actually improve that or not? It’s always good to extend your reach, but is it something that’s worth worrying about? I don’t think it is.

For that reason, that’s why I also think it’s okay to have multiple versions of the same thing. So you’ve got Go Time, but then of course - and I’m not gonna name them, but there’s other Go podcasts out there; don’t listen to them, everyone… [laughter]

Paul Smith: …but they’re basically outboard brains with if/then/else statements that had their own lifecycle of change, and very, very complex… So yeah, it was just not a good environment in which to introduce something like “Hey, this emergency email queue.” So what we did was we just – we made the case that “Hey, we’re going to requisition these two servers over here, that have nothing to do with anything. They’re not part of the data center, but they’re within the same kind of like general security boundary as the rest of the thing… And we’re going to run our own code over here. And then at the CDN level, we’re going to route everything past healthcare.gov/emailqueue or whatever, we’re going to route that to those servers.” So the rest of the infrastructure stays the same, we don’t touch that. And if our thing blows up, you can just get rid of that route, and it’s fine.

So we decided to write this thing, so we had a couple of design decisions up at the top, which was this thing had to be dead simple… Because we were going to be the ones to develop it, and we were already sleep-deprived, and dealing with 100 different things… So we didn’t want to add any more complexity than we absolutely needed to. It needed to be dead simple to operate, for some of the same reasons… And we wanted something that was going to be easy to deploy, easy to operate, and then easy to kind of get people back to the site. So what we came up with was basically just a loop, a go thread, or goroutine that would pull off a JSON request from the website, a simple form in the website that we injected with JavaScript, that grabbed your email and a couple of other bits of metadata, and then we just wrote it to a file, and we did that atomically inside of a lock… And so literally, all this traffic is just flowing into these files, just text files that we’re just like appending rows to… Because we didn’t want to mess with the database, and like separate processes… We just want our process, our OS process that we had control over, that we could use like text processing tools on the backend to do the actual email send. So that’s what we did.

So we would just collect people’s emails all day, and then when we saw the load dip under the threshold that we thought it was safe, we would do these sends to invite them back, with a special code that let them sort of bypass the waiting room, if that was still a thing… And yeah, and we brought them back.

And I should mention that we have like this throttling mechanism that essentially dialed in whether you got the email waiting room, or if you could go straight in through the site. So it was this sort of like probabilistic thing that was like a function of the load on the site at the time.

Glauber Costa: Yes. 100%. And we will. And I want to clarify that now, because with LibSQL it became very clear to us in maybe a month that that was not going to happen. So with LibSQL – I joked about it today, we had this goal of replacing SQLite for 10 minutes, because it’s a dream etc. It’s obviously a big prize. It’s obviously this big prize etc. As I wrote in my blog post, were we bummed about it? No, because we built a pretty successful business on top of the code, on top of LibSQL, which is today our company, Turso. So for us, it was this exercise of – if you heard the expression “Shoot for the stars, because even if you miss, you’re going to hit the moon…” For us it was like “Yeah, it turns out – I don’t think it’s going to be possible for us to replace SQLite with the fork.” But that’s us hitting for the stars. We didn’t hit the stars, we hit the moon. The moon’s great. Again, we built a super-great project, we’re having fun. LibSQL brought something to SQLite that wasn’t there, which was like the serverless, and running this on serverless environments, on the cloud platform, syncing databases… One of the things that we do with Turso is you can sync your databases between different devices, and backups, and all of that. And this is going great. I mean, this is going fantastically for us, and so we were never sad or disappointed that this thing that crossed our mind – can you imagine how amazing it will be if people just start using our fork instead of SQLite? We were never disappointed at that.

But when we put the rewrite out, it became very clear - and I’ve been using this analogy from chemistry, that the reaction was right. LibSQL was something that the market was clamoring for. I think what it became very clear seeing the reaction of Limbo now is just that the activation energy is not enough to make the reaction happen.

[20:15] And after talking to a lot of people - and there’s a whole story of what happened after we announced this… I talked to a lot of the people who came and started contributing to the project… And what it became clear is that they all agreed with us, that we need a better version of SQLite. I mean, it will be fantastic if we have something like SQLite, but that does this, and this. All our original vision. But the fork just wasn’t differentiated enough. It just wasn’t ambitious enough. It just wasn’t bold enough to get people to come and contribute.

Furthermore, it is actually very hard to do, because one of the things you did mention - the SQLite test suit is fully proprietary. So once you start making changes that are very deep into the core, you start hitting a bunch of issues. And some of them – it’s just like yourself saying “I know that the best way to solve this problem is this, but I’m going to solve it in this other way, which is almost as good, but not great, but it’s less risky.” So you start having this on your mind. So the fork never really unshackles you fully to go pursue this, which is why when we announced the rewrite, what we saw, again, was something I’ve never seen before. We got 8,000 GitHub stars in a week, essentially. We had what I consider to be for the first time in the history of computer science a Hacker News comment section that was mostly positive, with just a couple, I think maybe one or two comments only that were trying to denigrate us. But the rest very supportive. We had this fantastic, inspiring story of this individual who is in prison at the moment, and he was one of the first people in the United States to be granted access to the internet in prison. He can access Discord, but not X, by the way; that maybe say something about it. But he is now the fourth top contributor to the Limbo project…

Anshul Ramachandran: Yeah, so the answer to any question of like “Does any IP leave Codeium for enterprises?”, the answer is always no. So in pretty much every part of this system, our guarantee is to actually be able to deploy this whole thing fully air-gapped. We’ve even deployed in places like AWS.gov cloud, which is entirely – it doesn’t even have connection to the internet kind of scenario. So nothing ever leaves there, to address some of the points you brought up there, Chris. Yeah, I mean, we’re not the only ones who are saying, “Oh, no, the data that a company has privately is super-important”, and is potentially even more important than the size of the model.

[25:46] I think a good example of this is actually Meta. Instead of using like a GitHub Copilot, or any generic system, they decided - I guess in classic Meta fashion - to train their own autocomplete model internally, using all of their code. And they actually published a paper, I think, a few weeks back. And their model was, in terms of size, I think like 1.3 billion parameters. Like, small in respect to the LLM world. And it just massively outperformed GitHub Copilot on pretty much every task. There’s now corroborating evidence to what we’re saying about fine-tuning, that doing this actually does lead to materially better performances for the user in question.

Now, is that Meta model going to be good for everyone else’s code? Probably not. But that’s also not the whole point. And in terms of being able to fine-tune locally - yeah, we’re able to do this completely local. And again, it comes down to scale of data. Our base model has been trained on trillions of tokens of code. That’s a lot. That’s why we need this multi-node GPU setup, to do all this training. But an actual company - if they have, say, even 10 million lines of code, that’s about 100 million or so tokens. There’s a huge order of magnitude difference still between this pre-training and the fine-tuning, which is why we can do this kind of locally, on - actually, surprisingly - whichever hardware they choose to provision for serving their developers.

So again, this comes to some of our infra background and all the stuff that we know how to do - we actually can do fine-tuning and inferences on that same piece of hardware. So we don’t actually ask companies to provision more hardware. And even more critically, we are able to do fine-tuning during any idle time with that GPU. So whenever that GPU is not being used performing inference, it’s actually doing backprop steps to continuously improve the model.

Fine-tuning is just one aspect of like a larger kind of personalization system… But we’ve instrumented all of this on hardware, using our inference, to actually create a system that is relatively easy to manage; it’s not like a crazy amount of overhead for any company to manage or use Codeium… But still, get the maximum possible wins from these AI tools.

David Yakobovitch: I think it could be a combination of both. I hope it’s more that we as developers reorient ourselves and our technology. Think about if you are a developer building a product - you don’t want your product to break; so whatever the user input is, you’re designing it so that if that’s a number, the number gets fed into the system. But if it’s text, the text gets converted to a number, so it still gets fed into the system.

I think just like data scientists and engineers today who are developing their APIs and their systems not to break based on inputs, the same thing should be that “Are we thinking about humans first in these systems?”

We’re starting to see AI guidelines. I know the European Union recently launched their own AI ethics standards; that came out a few months ago. There’s similar initiatives going on in the U.S. as well on ethics, and about integrating these systems for all. But it’s really about having the conversation and then making sure you take action.

Just like you mentioned about the languages, I had the opportunity to sit down on the panel, an advisory session with one of the leading candidates for the New York City mayoral election in 2021, and that exact topic you brought up, we brought up as well.

You mentioned 6,000-7,000+ languages, and in fact in New York City over 20% of the people do not speak English. 10% of that is about 800,000 of the eight million people in New York. So if you take some of those languages, just the top ten other languages that are not English or Spanish, it is about 25% of the population. That’s enough to have enough votes to win an election. Not that the goal should be just to win an election, but it should be serving all constituents. And when you’re in an accessible city like New York City, it should not just be translating services for eight languages, but how about at least 100 of them or more?

I think when we start thinking about processes, we need to do a lot more competitive intelligence, a lot of research on who our constituents are, and then to best serve them. And if that means rolling out a feature in tranches, that’s totally fine. But as long as you have the goals there and you’re thinking about accessibility on the onset, it’s paving you in the right direction.

On Freund: For me, it starts in 2014. We were three co-founders, each one of us has their own journey that brought them to where Wilco is today, and eventually Wilco is the synthesis of this vision of three different people… For me, it started in 2013. I just joined handy as a VP of engineering, and we raised a nice seed round, but we didn’t have a lot of money. I was rather new to New York, and I didn’t have a strong network just yet… And we were just starting to build out our engineering brand. So I realized I’d have to compromise on something as I’m building my team, and I said, “Alright, I’m going to hire a team of the best and brightest, from the best schools, but they’re going to be inexperienced… Because they can write amazing code, so what’s the problem?” And guess what - like I said many times today, writing code is not that important, so I had to mentor them very closely. And at some point, it became unscalable. Luckily, by then I was able to hire more people that would help me with the mentorship burden.

But I reached out to a few bootcamps and I said, “Let’s do this evening class where we expose developers to simulations of real-world events. I believe that that way, within months, they can gain the experience of years.” All the bootcamps said, “We’re focused on zero to one. We don’t really want to do 1 to 100.”

[01:02:05.13] And then at some point, I moved back to Israel, and I tried it with Israeli bootcamps, and I heard the same response. And then at the end of 2020, I left WeWork and I called up one of my former colleagues, and I told him about this school that I’ve been meaning to open for a while now. And I said, “You know what, I’m gonna go for it as like a side project. I have a little time off, I’m going to do this as a side project.” And I had already called up a few CTOs, they all said, “This is great. We’re gonna send students your way”, and that former colleague told me that I’m stupid. So I asked him why, and he said, “Well you’re gonna have six, maybe ten developers per class. You’re not really making a dent in the universe. Let’s figure out a scalable way to solve this. And by the way, a third former colleague who’s a good friend to both of us is also thinking about the same space, so let’s get three of us to brainstorm and figure out how to solve this.”

We looked at everything that existed, and then we realized that everything is so top-down, and developers hate it. And we realized that there’s nothing that actually simulates all the warts and all. Not just the perfect codebase with a simple function that you need to write. We need to actually simulate all of the bad things as well.

And then we realized that it’s not about taking juniors from 1 to 100, but you don’t really have a way to practice throughout your career. Going back to the flight simulator analogy, it doesn’t matter how many years of experience you have, or how many hours of flight you’re logging every week. You still have to go through the simulator. And the fact that we don’t do it as software engineers is pretty crazy to me. And apart from not letting us advance, it’s also very demotivating.

I have a lot of friends who are senior engineers. They love the company they work for, they love the team, and they’re telling me, “I think I might leave.” And I’m like, “Why? You love everything about this job.” “Well, I’m not being challenged anymore. I’m not seeing anything new.” So I ask them, “Are you going to switch to a different company that you’re most likely going to like less than the company you’re in right now, just because you’re going to gain new types of experiences?” and they say yes. And this is somewhat tragic, right? What if they could have access to new scenarios without having to leave the team they love so much? What if they could continue to practice and become better at the same pace that they could become better when they just started out? That really guided a lot of the thinking behind Wilco.

Cory Doctorow: Oh, there’s a ton of them. So I’ll talk a little about interop – and as I say, I’ve got this other book coming out, “The internet con”, from Verso, in September, that’s just about interop. But audiences and artists get locked into platforms because platforms use the law and technology to block interoperability. Systems are intrinsically interoperable. Turing-completeness just seems to be like a thing we can’t get away from. I go to DevCon, Hope, or CCC, and there’s inevitably some presentation from someone who’s like “Hey, guess what? It turns out PostScript is Turing-complete, and I wrote a printer virus”, right? “It turns out that the scripting language for MySpace that lets you do animated GIFs is Turing-complete, and I wrote that MySpace virus”, right? You just can’t get away from it.

[54:03] You can always make interoperable things. The problem is that the normal interoperable things that we do - reverse-engineering bots, scraping, whatever - have become increasingly prohibited behind a wall of copyright, patent, trademark, trade secrecy, contract law, Terms of Service, and so on. We’ve created what Jay Freeman from the Cydia Project, he calls it “felony contempt of business model”, right? Like, when Apple wants to solve the fact that everyone who uses Windows uses Office, and Office for the Mac sucks, and so people are just throwing away their Macs because they can’t talk to Windows computers - they don’t like beg Bill Gates to make a better Mac Office, right? What they do is they reverse-engineer those Office file formats, and they make Pages, Numbers and Keynote. And they’re just like “Yeah, now it just works. You can switch from the Mac to the PC, PC to the Mac, you can send files back and forth…” And they even ran this ad campaign, this Switch campaign, about how you can just switch from one to the other.

So if you were to make a runtime, like an iOS runtime that let you leave your iPhone behind, or your iPad, and go to another platform, Apple would say that you’re a pirate, right? When they did it, it was progress. When you do it, it’s theft. They would nuke you until the rubble glowed. They would come after you with Computer Fraud and Abuse claim for violating their terms of service, they would say that you’re engaged in tortuous interference with contract, they would say that you violated Section 12.1 of the Digital Millennium Copyright Act by reverse-engineering, whatever.

As a practical matter, engineers can figure out how to do this. Engineers can figure out how to add extra app stores to platforms. We’ve got legislation pending in the US now about to be reintroduced, that came up in the last session, to create what’s called the link tax, where we’re going to say “If you talk about the news on social media, the media company gets a piece of it”, which is crazy. Talking about the news is like not a copyright violation. If you’re not allowed to talk about the news, it’s not the news. It’s a secret. But there are a couple of ways that tech platforms seriously steal for media companies. For one thing, every app has a 30% commission on every sale, to process a transaction.

Do you want to increase the subscriber revenue of every media company in America? Write a law that lets people install alternative app stores. And then there will be a race to the bottom for payment commissions. It won’t go to zero. It might go to 2%. So 28% increase on revenue for every subscriber, with one law. That’s more than you would get from a link tax. And that’s an interoperability measure, right? Letting people choose other software.

If you want to go even further, follow through on this law that we’ve got pending that forces the platforms to disaggregate their ad tech stacks, right? Google, or Facebook, they both operate a marketplace, a demand side platform and a sell side platform. It’s like the NASDAQ also owning the companies and the brokerages. Right? It’s like the referee owning the team. And so it’s no surprise that their share of income from the platforms for ads went from like 7% to 50%.

You want to increase the amount that ad supported media gets? Break up and make interoperable and disaggregated these ad tech stacks, so that they have to compete, so that they can’t just command these ridiculous shares of every dollar brought in in advertising.

So those are remedies that actually are about distributional outcomes, right? They change the amount of money being made by different entities. Not only that, but where we’ve had link taxes, like in Australia, where we created a link tax, the Murdoch Press took the link tax, gave it to its shareholders, and fired its reporters. Meanwhile, the smaller papers didn’t get the share of the link tax that the Murdoch papers did.

[58:09] So if we get rid of the app tax and the ad tax, then people starting small, independent publications are going to be able to get 100% of the revenue that they’re entitled to. It wouldn’t just be a gift to large media companies that could bargain for these rights. It would be a gift to everyone who makes media, including small, local, crowdfund-supported media platforms that are doing the shoe leather work of going into school meetings, and the waterboard meetings, and whatever. It’s not just a way to transfer money to private equity companies that have bought and strip-mined newspapers up and down the country.

So these are ways that technologists building interoperable layers like abstraction layers on top of things, or shims that sit between two different platforms, can actually help entertainers and media brands make more money by building something that audiences like better, while taking money out of the pockets of the big tech platforms that once promised you “Hey, this is a place that you can come and work for three years until you do your own startup”, and then said, “Oh, you’re not really going to be able to do your own startup, but I’ll tell you what - come work for us and we’ll give you massages and kombucha”, and are now like “Hey, guess what?”

Nader Khalil: Oh, man, what excites me about this space… I don’t even know how to say it… There’s just so much to focus on in every realm, right? Within interactive and non-interactive compute - like, I’ve talked to Eric at Banana about just how both of our teams are about the same size, and we’re both 100% focused in our space, and it just feels like there’s an infinite amount, just looking down. So looking up, there’s even more.

You guys mentioned your last episode was on ChatGPT, and I think AI is really exciting, not so much in that it’s gonna replace us all, but it kind of lets us be more creative directors of our own lives. If you think about any creative process as having like some generative aspect, and then some like malleable aspect… So if someone’s making clay, they like throw a bunch of clay down - that’s the generative. And then you kind of like form it nicely into the bowl or cup that you want, and that’s the kind of like morphing it in. So there’s always those two aspects. And when AI is able to help us just kind of really push on that generative side, and we’re still in control of the output, we’re still the ones that are kind of morphing the final product, I view it just as an extremely empowering thing. So it’s been really exciting , with all the development in the space… Really bought into the idea that you make things a little bit easier and you can just dramatically increase the affordance for things to happen. So as much as possible, how do we get rid of machine problems, and that people who want to build really exciting things and build the next new affordances, and the new models, essentially be able to do that with as little friction as possible. And that’s not just within their fine-tuning and resource constraints that they might have, but also in terms of like moving it and shipping it and delivering it.

So on one hand, the things that are being built is very exciting. But on the other, the energy in the space is huge, right? I think everyone has been so inspired by what’s been done recently with ChatGPT and the recent AI models that are out, that it’s just galvanizing a lot of people to build a lot of really cool things. Everyone I know, especially here in San Francisco, founder or not – it’s funny seeing founders who have nothing to do with AI thinking about AI side projects, right? That’s galvanizing people. And so everyone is really excited about building this stuff right now, and I just hope we don’t lose that energy, and just make things as frictionless as possible as we do that. Even I, I’m guilty. I have a little Saturday project I’m throwing together with some generative AI stuff. There’s a lot of really cool stuff that’s happening.

Jon Kuperman: Yeah, I’d say there’s the big things, like to check out – if you start a new Pages app, or if you move your site onto Cloudflare, just playing around in the dashboard, which we’ve been working on design tweaks to, but… We basically have a lot of cool stuff about automatic image optimization, making your images smaller, or reading the device size and serving images… So that’s like Polish and Mirage.

[43:59] Then we have the Rocket Loader, which will turn off unnecessary [unintelligible 00:44:02.07] So we’ve got a lot of these options, and these are all one-click options for the most part. So if you start a new Cloudflare Pages app, or if you move your website over… Sort of looking at these things - automatic imagery sizing, automatic compression of JavaScript and CSS and HTML, serving images by the device width… Or with the automatic resizing, every time we have to do a resize, we save it, too. That’s like a paid service. And so then you just upload one image, and then as devices come in, we start resizing it based on their width, and then we save all those cached in the edge… So you end up with 10 or 20 copies of your image. Rocket Loader for speeding up JavaScript-heavy sites…

And then the other one that I wanted to talk about which I think is really cool is we have this tiered caching initiative which is now available in the dashboard… So the idea being let’s say you have 200 data centers, and you use our CDN. Or we have 200 data centers, you use our CDN. So your stuff gets cached in all 200 data centers, but then every time you do a cache purge, or a user requests something where there’s a miss, it has to always go back to your application.

So a hypothetical, I guess, would be like you have an app, you go buy a DigitalOcean VPS, you put it in your New York City, and you put your app on there… And it gets overwhelmed, so then you sign up for Cloudflare CDN, and poof - all of a sudden there’s all this caching stuff. But you can still have this problem where if somebody’s hammering on an end point, or just users are trying to request an end point and it’s a miss for the cache - all the caches will start going back to your one DigitalOcean VPS that you bought. It’s almost like a mini-DDOS, an innocent/accidental one…

But you see this – like, Facebook went down the other day… So all of these big sites, like Facebook, Twitter, Google - they take in more traffic than they could possibly serve cold; like, if all the computers went off and all went back on again - they’d just get hammered. So they rely on these caching mechanisms. So one of the problems for these companies when they have an outage and things go actually down - you’ll see this all the time, where they keep trying to go back up, and it’s like up and then down again. And then it’s like, they go up and they go down. So they have to come up with these clever ways of like “Okay, we’ll go up, but we’ll block any users whose names don’t start with A”, or they’ll come up with these ways of doing these slowly booting back on, so they don’t get just decimated again.

So the same thing can happen with just anybody’s application, where you have like this one VPS that can handle (whatever) 100 or 1,000 hits a second, and then something gets purged from the cache and all of your users are trying to hit it - boom, it takes it down again.

So we came up with the idea of tiered cache, where we take those 200 and we mark, let’s say, 25 of them as top-level cache CDNs. Like, cut top-level endpoints. So what happens is now the other 175 are no longer allowed to hit origin when they have a cache miss. What they have to do is they have to find their closest top-level cache system and ask it for the file.

So it’s like, “I need a JPEG, and I hit this one – I’m in Florida, so I hit the Miami one. Miami doesn’t have it. Miami is no longer allowed to just hit New York.” [unintelligible 00:47:10.19] it can’t do that anymore. It has to find its closest one, let’s say like in South Carolina, and be like “Do you have the JPEG?” And only if that doesn’t have the JPEG, South Carolina has permission to hit New York now.