Latest / Coffee with Developers / Project Chopin - Make Your Team and Agents Play Along - Krzysztof Cieślak

Transcript

Project Chopin - Make Your Team and Agents Play Along - Krzysztof Cieślak · Coffee with Developers

• 0:06
  SPEAKER_01: Hello and welcome to another We Are Developers Live.
• 0:09
  SPEAKER_01: It's Wednesday again, and we're gonna be talking to somebody here who I had uh lots of communication with before and who also got me access to free copilot uh when it when it came out.
• 0:19
  SPEAKER_01: So that was really exciting to see.
• 0:21
  SPEAKER_01: And uh another big big piece of news.
• 0:23
  SPEAKER_01: I'm back on X.
• 0:25
  SPEAKER_01: Like uh half a year ago, I had a I had a I had a phishing attack, and basically that attacker then set up a two-factor authentication with an uh with a uh a tool that actually validates him.
• 0:38
  SPEAKER_01: He didn't manage to do it either, or I guess it's a he.
• 0:41
  SPEAKER_01: Uh but uh yeah, and I didn't have it, so basically I was locked out and I emailed them like every two days or something, and nothing happened.
• 0:48
  SPEAKER_01: And then out of a sudden today, it was like, oh yeah, we turned off two-factor authentication, so you can turn on two-factor authentication in another way again, and I'm back on X.
• 0:56
  SPEAKER_01: So I got a lot of like DMs that I haven't answered in a while.
• 0:59
  SPEAKER_01: Um I I will not read the feed, uh, but I I think for DMs and stuff, it will be useful to have that again.
• 1:07
  SPEAKER_01: And yeah, I mean, 64,000 people follow me, follow me, so it's actually worthwhile having that as a platform to do things there.
• 1:14
  SPEAKER_01: I tried the boosting, but then I said like 100 pounds or something, and I'm like, nah, not gonna do that.
• 1:19
  SPEAKER_01: No idea what that is about.
• 1:21
  SPEAKER_01: Uh, but yeah, it's maybe somebody can sponsor me and boost my stuff.
• 1:25
  SPEAKER_01: I don't mind.
• 1:27
  SPEAKER_01: So one thing we have uh that is coming very up, very closely is down here.
• 1:32
  SPEAKER_01: You can see the September 23rd to 25th.
• 1:34
  SPEAKER_01: We're gonna be in America, we're gonna be in the Silicon Valley in San Jose, and we're gonna do our event there.
• 1:40
  SPEAKER_01: Uh again, 10,000 people, 600 speakers.
• 1:43
  SPEAKER_01: I'm just looking for moderators right now.
• 1:44
  SPEAKER_01: So if you want to moderate a stage, um, there's gonna be a tracker where you can actually sign yourself in.
• 1:49
  SPEAKER_01: Uh uh, you can email me at chris at vrdevelopers.com.
• 1:54
  SPEAKER_01: And yeah, it's gonna be an interesting conference.
• 1:56
  SPEAKER_01: It's gonna be a blast because for the first time we do it in America, I'm not freaking out at all about it, but like it's gonna be it's gonna be interesting because we've done it for 10 years in Berlin now.
• 2:06
  SPEAKER_01: Well, eight years in Berlin, two years in Vienna.
• 2:09
  SPEAKER_01: And um, yeah, it's gonna be an interesting audience.
• 2:12
  SPEAKER_01: I mean, we've been over there.
• 2:14
  SPEAKER_01: Um, are you ready for it?
• 2:15
  SPEAKER_01: Are you ready to to rock America?
• 2:18
  SPEAKER_06: Um I I I'm I'm not gonna jinx it and say yes, I'm completely ready, because you know, that would be famous last words.
• 2:22
  SPEAKER_06: But um, yeah, yeah, I'm really excited for it.
• 2:24
  SPEAKER_06: I think that it's um it's definitely a slightly different market, you know.
• 2:27
  SPEAKER_06: We we have kind of noticed a few differences, and I know we've gone over those in the past.
• 2:30
  SPEAKER_06: Um, and there's obviously a lot of things kind of in in the area, you know, when we go to San Francisco, go to San Jose, you realize quite how how how big the tech scene is that we call all hear so much about online.
• 2:41
  SPEAKER_06: Um and when you're there in amongst it, you know, the energy is is palpable and there's so much happening.
• 2:47
  SPEAKER_06: So it's been it's been really cool actually to kind of introduce we are developers to people that maybe didn't know about it as much, but also to hear about how many people, when we've gone over there from Europe, already knew about us.
• 2:56
  SPEAKER_06: And that's that's already all always really nice to hear.
• 2:58
  SPEAKER_06: Um that you know, we've we've got built up such a reputation in Europe and it's it's you know uh followed us over there.
• 3:04
  SPEAKER_06: So that's that's good to know.
• 3:05
  SPEAKER_01: And I mean the lineup is insane.
• 3:07
  SPEAKER_01: I mean, uh I love that we got John Romero again.
• 3:10
  SPEAKER_01: There's gonna be there's gonna be a Doom uh uh competition live against him, and nobody has a chance to beat him.
• 3:15
  SPEAKER_01: I think we should we we should do something that I'm good at, like you know, like play Tetris against each other or something like that.
• 3:21
  SPEAKER_01: Because whenever these games came out, like those old Doom games and stuff, my machine was too slow.
• 3:26
  SPEAKER_01: So basically I just tried to actually beat somebody and and well, I had like a four or 500 megahertz Pentium one back then, and basically like all of the games were shit.
• 3:36
  SPEAKER_01: Like it was just I had no chance of the lag basically killed me more than the other opponents, but yeah, it's uh it's interesting to see.
• 3:44
  SPEAKER_06: Yeah, yeah.
• 3:45
  SPEAKER_06: And you were you were programming on things like Commodore and things like this, right?
• 3:48
  SPEAKER_06: And um, whereas you know, I was very much like I think my brother had like an Amiga, was it even 5,000?
• 3:53
  SPEAKER_06: I don't even know what the number was now.
• 3:54
  SPEAKER_06: 500 500, there you go.
• 3:56
  SPEAKER_06: Yeah, and I remember sort of watching it, watching him kind of it looked all very complicated to me, but it sort of passed me by in that you know, my first console was very much like a PS1.
• 4:04
  SPEAKER_06: And so it was kind of the the days of being, yeah, really kind of locked up, you know, there was no real like hacking your own games and things like that.
• 4:11
  SPEAKER_06: I really missed out on that.
• 4:12
  SPEAKER_06: Um, but yeah, it sounds it sounds really cool when you talk about it.
• 4:15
  SPEAKER_01: Not even an Atari 2600 or a super NES or a GameCube.
• 4:19
  SPEAKER_06: I I once yeah, I did have that actually.
• 4:22
  SPEAKER_06: That's but again, my brothers.
• 4:23
  SPEAKER_06: My my first one was the PS1, but yeah, we did have the snares and the and the and the NES.
• 4:28
  SPEAKER_01: Okay, I mean the Wii uh I love the Wii because that was the first time my parents actually played with a computer because they're like, oh, I can move it like that, and that means it's it's doing that.
• 4:36
  SPEAKER_01: So so we got we got we got our guest waiting here, and he's uh chomping at the bit at the moment.
• 4:42
  SPEAKER_01: So uh without further ado, here's Christoph Chislack, uh, which is pronounced John Smith, interestingly enough, uh, to make that easier.
• 4:50
  SPEAKER_01: And uh he's as you can see from his name, there is a lot of like uh ROT 16 or base64 encoding happening, or he might be from Poland.
• 4:59
  SPEAKER_01: Christoph, how are you doing?
• 5:01
  SPEAKER_03: Yeah, uh I'm great.
• 5:02
  SPEAKER_03: Thank you for having me, and this is this is great.
• 5:06
  SPEAKER_01: Oh, and he's stuck.
• 5:08
  SPEAKER_01: Excellent.
• 5:11
  SPEAKER_01: Anyways, I mean Christoph is from GitHub Next, and he's one of the researchers there that actually helps GitHub go for the next bits and bops that they want to do in the nearer future.
• 5:20
  SPEAKER_01: And uh he was the one that got me into GitHub Copilot for free, and I was really excited about that one because I was working at Microsoft back then and even we didn't have access yet, and he got me into that, so that was really good.
• 5:32
  SPEAKER_06: And now he he is stuck.
• 5:34
  SPEAKER_06: I mean, the the the the sort of beautiful irony of this is that we've had about a 15-minute conversation before we've gone live, and it was uh completely seamless.
• 5:41
  SPEAKER_06: And then the moment we've gone live, uh Christoph's uh cameras decided to go kaput.
• 5:45
  SPEAKER_06: So, Christoph, it might be good for you to leave the uh leave the call and come back in, and we'll bring you back up on stage.
• 5:49
  SPEAKER_06: But um, I did while while we kind of do that, I I was always curious how you got access, Chris, because I know you're very well connected, let's say, amongst the um the tech scene.
• 5:58
  SPEAKER_06: Uh, but I was curious about how you got access to Copilot for free and so early.
• 6:02
  SPEAKER_06: Um because you were you you were always kind of aware of like, you know, it was very much like I was seeing articles about it, and there were a few people that had kind of gone in on it, but you you all seem to kind of have that um yeah, kind of next next look into what was coming and things like this.
• 6:14
  SPEAKER_06: So now we know where the mole is, is Christoph.
• 6:16
  SPEAKER_01: It is really tricky because I mean I have so many things, uh so many emails a day where people say, like, invite me to to put things on my blog to try out a new tool.
• 6:25
  SPEAKER_01: And I I I often like ask them, I don't care.
• 6:28
  SPEAKER_01: Like, but it's like some of them are actually really amazing, and then like half a year later they become a thing where people get excited.
• 6:34
  SPEAKER_01: You know what stopped happening was the whole like invite-only thing.
• 6:39
  SPEAKER_01: I uh companies used to do that.
• 6:41
  SPEAKER_01: Remember, I mean, I remember Twitter was invite-only, and I was one of the first 2,000 people on it and stuff, and you had to get like one invite that you can invite another person and they make it viral that way and make it more.
• 6:53
  SPEAKER_01: But nowadays people just release their software and don't go for that movement anymore.
• 6:58
  SPEAKER_01: So it feels really bizarre that that this trick is not being used anymore.
• 7:02
  SPEAKER_06: Yeah, you're right, because it was something I think they had for Blue Sky as well.
• 7:05
  SPEAKER_06: I remember that was the when I first heard about Blue Sky, it was actually through this kind of invite-only thing, and it was everybody saying, I've got one invite, you know, who who would like to get um invited to it.
• 7:14
  SPEAKER_06: And it it wasn't even that I knew what Blue Sky was, it was kind of the first day I'd seen these things.
• 7:18
  SPEAKER_06: But it it does, it does make you instantly think, okay, I want to be in on this.
• 7:21
  SPEAKER_06: It's the FOMO, right?
• 7:22
  SPEAKER_06: Um yeah, maybe that maybe that is a sort of let you know, maybe it's fallen out of favor as a technique, or or maybe it's just that there's so much competition for users that actually making it invite only has the negative effect because you get such a small uptake, maybe compared to what you would have if it was just free and open.
• 7:37
  SPEAKER_06: Um, who knows?
• 7:38
  SPEAKER_06: Yeah, I'm not sure of the reasons why.
• 7:40
  SPEAKER_01: It it's really interesting when the rollout is so bizarre.
• 7:43
  SPEAKER_01: Like, I mean, there was uh in the German news, there was a social platform called V or W or something, and that was basically like the German answer to uh the European answer to Twitter.
• 7:54
  SPEAKER_01: Basically, you have to validate with a European passport to be part of it.
• 7:58
  SPEAKER_01: It's real people, real conversations, uh uh like non-algorithmic uh news, and basically it said like we're gonna be the uh the adult version, or not adult content, but like grown-up version of uh of the mosh pit that is uh that is X by now.
• 8:12
  SPEAKER_01: And uh the the and on the day I basically I saw it in the news and I read it in a newspaper.
• 8:18
  SPEAKER_01: Yeah, I knew I read newspapers, and uh then I basically got the uh got the uh signed up for it, and I'm like, okay, cool, let's try it out.
• 8:26
  SPEAKER_01: And I was happy to promote it as well.
• 8:28
  SPEAKER_01: And then it took four months, and then the email came like, oh, you're now part of it, and now you can be into it.
• 8:34
  SPEAKER_01: Now you can you gotta sign in and scan your scan your ID.
• 8:39
  SPEAKER_01: And I tried to scan my ID like five times, nothing happened, and it it really didn't work either.
• 8:44
  SPEAKER_01: And I'm like, this is not how you kill an evil American company.
• 8:49
  SPEAKER_01: This is not like the plot has to be smooth, you know.
• 8:52
  SPEAKER_01: Like this should be much easier to do.
• 8:55
  SPEAKER_01: And I was just like, why is this completely gone by now?
• 8:58
  SPEAKER_01: So that's interesting to see.
• 9:00
  SPEAKER_06: Yeah, yeah.
• 9:00
  SPEAKER_06: I know we're still waiting for Christoph.
• 9:02
  SPEAKER_06: Um, and hopefully he'll be back in a moment.
• 9:04
  SPEAKER_06: But but I know what you mean.
• 9:05
  SPEAKER_06: I think there's so much to that actually around that UX around onboarding and that making something frictionless, you know, there's such an art to it.
• 9:11
  SPEAKER_06: And we talk we talk now about kind of you know, vibe coding and building things agentically and things like this, and how um so much of actually UX is about being a real human being.
• 9:19
  SPEAKER_06: That's what you said, like real people, real conversations.
• 9:21
  SPEAKER_06: Um, and so the irony is then that they maybe didn't kind of empathize with actually that if you they want users to use the tool, they need to make that thing frictionless, you know, that onboarding process.
• 9:30
  SPEAKER_06: Um, because there is such such an art form to it.
• 9:32
  SPEAKER_06: And and I mean the fact is in that four months, you probably forgot that you'd signed up, right?
• 9:37
  SPEAKER_01: Yeah, and the other thing is of course, like it says like NFC, like it's NFC enabled, and of course, German past uh ID or European IDs are NFC enabled, so you can just touch your phone and you would be logged in with that thing, but then it says like, oh, log in with NFC, and I'm like, excellent, cool.
• 9:53
  SPEAKER_01: And then it says, like, and now take a picture of your ID, and I'm like, how is that NFC?
• 9:58
  SPEAKER_01: Like, you know, yeah, this is like oh I'm in Holmes.
• 10:04
  SPEAKER_06: Let's say oh let's see if we've got him on camera.
• 10:07
  SPEAKER_06: There we are, Christoph.
• 10:08
  SPEAKER_06: How are you?
• 10:09
  SPEAKER_03: Yeah, sorry, sorry for that.
• 10:11
  SPEAKER_03: Uh I think my I've lost power for a moment in Hall Flat, which was weird, and then my router had to restart, and that takes a moment.
• 10:19
  SPEAKER_03: Wow, so yeah, that was power in you.
• 10:24
  SPEAKER_01: Okay, you only hear that about American data centers, not about how houses in Poland, so that's really weird.
• 10:30
  SPEAKER_03: Yeah, that was really weird.
• 10:31
  SPEAKER_03: I don't know what happened.
• 10:32
  SPEAKER_03: Uh but it was like, you know, just five seconds, that was just like five seconds, but then you know, everything turned off in my house.
• 10:38
  SPEAKER_03: Like every every single device was like, okay, I'm off.
• 10:42
  SPEAKER_03: Uh, which means like router and uh look kind of like internet thingy modem.
• 10:47
  SPEAKER_03: Uh it's like, oh, I need to restart this now, and that takes time.
• 10:51
  SPEAKER_01: It's like when you update your operating system, and then like every single app also updates afterwards, and like you know, you get stuck in that loop as well, and you're like, oh god, stop doing that.
• 11:01
  SPEAKER_01: Although I have to say, uh, especially uh, I mean, I'm I'm not sure about Windows, but Mac, I'm very excited that like nowadays you can't lose anything anymore.
• 11:10
  SPEAKER_01: Like, even if your machine is like uh running out of memory or something and needs to reboot, everything starts and is the way it was before.
• 11:18
  SPEAKER_01: Like the whole idea of like having lost any work because you didn't save it in time is kind of gone.
• 11:25
  SPEAKER_01: VS Code has no problem with that, Edge has no problem with that.
• 11:28
  SPEAKER_01: Like, I find that really, really satisfying.
• 11:31
  SPEAKER_01: That uh I mean, back in the days it was much more annoying than basically like all right when you restart Edge or Chrome and you do Command Shift T and it opens all the tabs that you had before, and not like, oh god, how many of those did I did I open?
• 11:44
  SPEAKER_01: And I'm horrible at that.
• 11:45
  SPEAKER_01: I got about 300 tabs open at uh at the time.
• 11:48
  SPEAKER_01: Uh and like, yeah, I'm gonna read this.
• 11:51
  SPEAKER_01: Um, like it would be I saw extensions in the past that basically closed the things and made it bookmarks when you didn't read it for like seven days or something.
• 12:00
  SPEAKER_01: So I found that a really good idea.
• 12:01
  SPEAKER_01: When we did Firefox OS, we thought about making the app icons that you haven't used in a while start to rust and crumble.
• 12:09
  unknown: Yes.
• 12:09
  SPEAKER_01: Well, basically, you feel bad for them when you don't use them.
• 12:13
  SPEAKER_03: One one of my one of my colleagues from the team has created uh he has created the extension that uses AI to manage your open tabs.
• 12:22
  SPEAKER_03: So it groups them it by topic and then it puts them uh into archive or like whatever bookmarks, uh, if if they're open too long and stuff like that.
• 12:32
  SPEAKER_01: So obviously AI isn't that what Arc set uh is gonna do for you?
• 12:37
  SPEAKER_01: Just think about it.
• 12:38
  SPEAKER_03: Yeah, ARC, but you know, kind of like I don't know what is the status of ARC nowadays.
• 12:44
  SPEAKER_03: Uh I don't think it's developed anymore.
• 12:46
  SPEAKER_06: Uh I think I'm the only user still, and it's uh purely because of you know, you know what it's like when you've just got a few things that kind of work quite well and the kind of split screen that I know they've got in edge and things like this.
• 12:56
  SPEAKER_06: But I've been saying to you for ages, Chris, have an eye back moving away from it because I I really only use it for for for kind of viewing these articles and things on the show because it's an easy setup.
• 13:05
  SPEAKER_06: But yeah, I think actually the the brand the um the tab kind of cleanup thing that they've got just closes the tabs.
• 13:11
  SPEAKER_06: Um it doesn't really do much, I think, in the way of archiving.
• 13:14
  SPEAKER_03: And so that's kind of annoying because if you're they do have they do have this concept of archive.
• 13:18
  SPEAKER_03: I don't so I've I've also been user of ARC for a while.
• 13:21
  SPEAKER_03: Uh but kind of like it's it's again like as just as you said, like I'm using probably like just 10% of the capabilities of the of the product.
• 13:30
  SPEAKER_03: It's like oh, it looks nice, and then you can do split screen nice.
• 13:34
  SPEAKER_03: Uh but now I move to Helium, which is this like privacy focused, it's another Chromium fork because like everything is based on Chromium nowadays, right?
• 13:44
  SPEAKER_03: Uh but it's kind of like it also have a bunch of features I liked from from Arc, which is the split screen and and so on, but it's very kind of like privacy focused.
• 13:53
  SPEAKER_03: And uh I'm trying this out.
• 13:55
  SPEAKER_01: I'm using I'm using LibreVolve for that, like basically it's a fork of Firefox that has all the trackers enabled and stuff, and especially now with like Chrome not supporting uh u-block origin anymore because all the uh the the manifest two extensions are basically being discontinued.
• 14:14
  SPEAKER_01: It's an interesting thing how that's gonna be a bit, but then to me, uh to me, like app blocking has always been part of the or tracker blocking has always to me been part of the operating system and not the browser.
• 14:25
  SPEAKER_01: I think if you do it directly on a router level, or do you do it like directly on a uh on a on a network level, that makes so much more sense than having to slow down your computer by having blocking things before they're being loaded rather than just like doing it directly on your internet connection.
• 14:42
  SPEAKER_03: The issue is that this is kind of like developer statement, right?
• 14:45
  SPEAKER_03: Like for for normie, uh like normal people, like my mom will not set uh router level blocking for ads of their kind of like local internet.
• 14:56
  SPEAKER_01: Uh that's yeah, but then they don't set up their router either.
• 14:59
  SPEAKER_01: They always ask one of us nerds to do it for them, right?
• 15:03
  SPEAKER_01: Yeah.
• 15:05
  SPEAKER_01: Fair enough.
• 15:06
  SPEAKER_01: That's the other bit there.
• 15:08
  SPEAKER_01: Cool.
• 15:08
  SPEAKER_01: So let's took a let's take a look at some resources that I collected that we should that we can give a have a chat about.
• 15:14
  SPEAKER_01: So uh the first thing I found is a research that actually looked at uh what cursor and cloud code and codecs are using, what kind of tools.
• 15:24
  SPEAKER_01: And it basically is this it has this big um dashboard here where it basically the agent sandboxes, the observability, payment systems, um deployment, author authentication, email providers.
• 15:40
  SPEAKER_01: Uh I didn't know that resent was that big, for example.
• 15:42
  SPEAKER_01: That was interesting to see.
• 15:44
  SPEAKER_01: Uh, product analytics and so on and so forth.
• 15:47
  SPEAKER_01: So these are the things that basically uh when code is generated, those are the services that these things use.
• 15:54
  SPEAKER_01: And uh I saw a talk at the um at Bethrelcon in New York, where somebody said, like, how do you get your tool to be used by services and by agents?
• 16:05
  SPEAKER_01: Like, that's the kind of new SEO for tool makers.
• 16:09
  SPEAKER_01: And uh, this will be interesting to see for people like that to see how they can beat those.
• 16:14
  SPEAKER_01: And it would be also interesting to know why they actually chose those.
• 16:18
  SPEAKER_01: I mean, Amazon S3 as the main storage thing is not a surprise because I think they were the first one in that uh in that uh that play.
• 16:26
  SPEAKER_01: But yeah, some of them I have never heard of, and I have no idea.
• 16:30
  SPEAKER_03: Neon in in database in database category, neon was super super was very surprising to me.
• 16:36
  SPEAKER_03: Uh yeah, not because I haven't ever heard about neon, I I obviously have heard about, but the uh the percentage amounts are crazy, like the the lead it has over the second one seems like uh very surprising to me.
• 16:50
  SPEAKER_00: Uh maybe tier massive three tier, maybe.
• 16:54
  SPEAKER_00: Yeah.
• 16:55
  SPEAKER_03: Also, you need to remember that what it goes back to what those AI systems, those LLM systems are, right?
• 17:02
  SPEAKER_03: So they have been trained on the whole internet and they kind of like show you the most likely answer in principle.
• 17:11
  SPEAKER_03: So maybe Neon has a very good DevRel story, maybe they are very popular on social media, maybe they are uh mentioned often in the blog post, and that's all those can so all this traditional DevRel things that that that that you all have been doing can uh impact how those those agents answer and like what what what uh what what are the main tools that they are choosing.
• 17:40
  SPEAKER_06: I think also what I think um just thinking then about something to do with the the reinforcement of it in that if you use something once, you know, it's then recalling memories.
• 17:48
  SPEAKER_06: And so if you say I'd like to now build something else, it's gonna kind of pick the same stack.
• 17:51
  SPEAKER_06: And so um there's almost something to be said for if you can be that that person or that company rather that kind of gets uh selected, you know, as the just give me a database option and it chooses that, then you're probably likely to kind of be reinforced um time and time again.
• 18:04
  SPEAKER_06: So yeah, but that jumped out to me as well, Christoph, Neon.
• 18:07
  SPEAKER_06: Surprising.
• 18:08
  SPEAKER_06: 66%.
• 18:09
  SPEAKER_01: Yep.
• 18:10
  SPEAKER_01: Yeah, Stripe, Stripe also, uh, because I mean uh it's an American-only thing, I think.
• 18:17
  SPEAKER_01: Do you don't you need to use PayPal for European stuff?
• 18:20
  SPEAKER_06: I believe now they've uh folded in things like Lemon Squeezy, though, as well.
• 18:24
  SPEAKER_06: And I know Lemon Squeezy, I think, was trying to kind of uh make um make it sort of less less important if you're in the US or not.
• 18:30
  SPEAKER_06: So yeah, I'm not quite sure, you know.
• 18:32
  SPEAKER_06: But um yeah.
• 18:34
  SPEAKER_06: Still not much interesting to see Strap up there.
• 18:37
  SPEAKER_01: It's interesting to see that all these things are in there.
• 18:39
  SPEAKER_01: So if you're looking for companies to join, maybe that's the ones that actually were interesting to look at.
• 18:45
  SPEAKER_01: Uh another thing, uh, another research is like the uh where do actually AI and perplexity and other uh where does perplexity get information from?
• 18:55
  SPEAKER_01: And I found that fascinating because it's like I mean, Reddit I knew, but g2.com I never heard of.
• 19:03
  SPEAKER_01: Yeah, guideflow.com I never heard of either.
• 19:05
  SPEAKER_01: Wi-Fi talent sounds like something somebody bought a domain and doesn't know what to do with it.
• 19:10
  SPEAKER_01: Like it's just it's fascinating.
• 19:12
  SPEAKER_01: And why christianheiman.com is not in there is really fascinating as well.
• 19:18
  SPEAKER_06: It's so true.
• 19:20
  SPEAKER_06: Yeah.
• 19:21
  SPEAKER_01: But yeah, guideflow.
• 19:28
  SPEAKER_01: Yep.
• 19:29
  SPEAKER_03: And I think that that there was interesting somewhere in this in this blog post, there was this interesting uh thing that uh there was a lot of citations from the kind of like just three web pages that were kind of like probably published by the same person or by the same company, and they were just kind of like marketing each other.
• 19:48
  SPEAKER_01: Uh also like like like link farming, SEO, yeah, yeah, basically, basically basically link farming.
• 19:54
  SPEAKER_03: And I think that was that that's kind of like uh what was interesting to me.
• 19:59
  SPEAKER_03: Uh and it's I think the main main thing is that the the AI is amplifying exactly the same problems that we had in the internet before, right?
• 20:09
  SPEAKER_03: So kind of like just the fact that something is in Google uh as a as a top result, it doesn't really mean that it's the actual the quality or it's true, like people were manipulating uh uh those results all the time.
• 20:24
  SPEAKER_03: Uh and and with AI it's even it's even a bigger problem because we have less control, right?
• 20:30
  SPEAKER_03: Uh just the fact that AI is putting a footnote to the source, it doesn't mean that it's actually fact-checked information, right?
• 20:38
  SPEAKER_03: And it's actually true, like it's just a link to some source.
• 20:41
  SPEAKER_03: You have no idea what is the quality of the source.
• 20:45
  SPEAKER_03: And as as as this research shows, like AI is is is choosing those links to various places, even Reddit, like sure, like we all probably like Reddit to a degree, but uh the level of quality of information on Reddit varies a lot depending on topic and and uh you know uh who is answering and so on.
• 21:06
  SPEAKER_03: So yeah.
• 21:11
  SPEAKER_06: It's anonymous on Reddit.
• 21:12
  SPEAKER_06: You know, it's worth noting that it's anonymous, so by definition, you can never really tell who the who the source actually is.
• 21:17
  SPEAKER_06: And I think that's the kind of major problem that you know if you put that on on any news site, and if you imagine there was an article and they just say we can't tell you who wrote this, but trust us, um worrying, you know.
• 21:28
  SPEAKER_01: Yeah, but I mean that's why that's why they index it as well, because there's no worry there.
• 21:32
  SPEAKER_01: I mean, Anthropic has this issue with like book publishers right now that basically say, like, okay, my publisher sold my content to Anthropic without telling me.
• 21:40
  SPEAKER_01: Uh so I mean, I don't mind my books are fifteen years old, but like it's uh it's still fascinating to see how many publishers are uh how many book authors are super annoyed about that right now.
• 21:51
  SPEAKER_01: And I mean the the other thing is that like the books being scanned by Amazon and then destroyed, that's one of the weirdest stories that I've seen lately.
• 21:59
  SPEAKER_01: And that's just sounds absolute absolutely terrible uh i think the the problem with so at least when you when when when ai like perplexity cites the web page at at least there is source the problem with like scanning books is that the author gets nothing out of it there is no usually there is no like kind of like footnote that shows his book that you can buy or something like that it's just information that is in the in the training set uh and that's that kind of like creates interesting interesting uh situation interesting yeah I found that also fascinating the idea of work slop that basically is like people your work colleagues using ai to generate lots of stuff and sending it to you as action items and what to do against it so I found that fascinating as well because there is there is there is the idea of like AI makes me so much more efficient that I'm now sending out work uh requirements to all my colleagues that are basically wondering where that came from.
• 23:03
  SPEAKER_03: Sean is in in general like he he he he has great opinions he writes really good blog posts so his whole blog is worth reading not only just this post.
• 23:12
  SPEAKER_01: He works hook doesn't he think so yeah yeah he works hook it up I think as well yeah we should get him on the show will be interesting to have him here but yeah he he has good opinions and like very opinionated opinions as well but not without being without being a terrible person like other people with very opinionated blocks.
• 23:32
  SPEAKER_01: So that's good to see naming names but yeah uh this was also interesting that like um 42% of CEOs report that nothing changed because of when they started embracing AI in the company costs flat revenue flat 12% got both benefits at once and 13% came out worse.
• 23:51
  SPEAKER_01: And I'm not surprised I mean the idea of like saying like we give people a tool and then everybody is like super efficient and like an expensive tool then you shouldn't be surprised if less money comes in.
• 24:04
  SPEAKER_03: Also we we see very good results with AI in software development but just making software developers slightly more productive or like even more productive doesn't necessarily mean that company revenue will grow.
• 24:18
  SPEAKER_03: Like there is a huge huge space between actually just writing software and writing code which is what what those AI agents are really good at to you know selling software to to to building customer relationship to to doing whatever else is required to to grow uh your company and grow revenue so it's not just one-to-one direct direct relationship of let's just make developers more productive ergo we will earn more money like it doesn't work like that it is fascinating to see though like when when chen ai came out everybody thought it's gonna make it be like for filmmaking it's gonna make writing a thing it's gonna make uh the creative process is gonna be dead and like now it's basically we optimize development with it because it was uh it was an algorithmic thing that can be done by a machine easily compared to a creative thing that can't be done that easily and uh being a good writer is still one of the best job uh uh safety measures that we have right now rather than just like being somebody who knows the syntax of uh a coding language and like uh and these kind of things so it's interesting to see that this yeah yeah the very important thing about software development and AI in in in this context is that we have a ways to verify solutions very easily right like we see whether the the software is compiling we we can run unit tests we can run integration tests yes someone needs to write them and make sure that they're right but there is this feedback loop that ai can optimize towards and and this feedback loop where ai see the like results and can kind of like if you if you give ai tasks that let's make code more performant and then you you give ai tool to measure performance of of your software it will go crazily to to just do some iteration after iteration after iteration after iteration to to improve the the thing this is not true in any creative art right like you cannot measure creativeness you cannot measure how beautiful some picture is you cannot measure how good your writing is right so that means that AI has no there is no this this built-in feedback loop into into AI tools which is true in software development and that what makes AI very software develop it makes software development very unique target for AI and that's why I think the adaption of AI in software development is going so well yeah and the only companies that actually make money with it a lot are actually software companies making software products for developers so that's the weird thing as well like I mean this is really odd because when I worked on uh in uh on editors and stuff like that nobody wanted to spend money on them but like with a subscription to an AI service people are happy to do that probably not because the developers are pay are paying for it because but because the CEOs are paying for it for the developers.
• 27:10
  SPEAKER_03: Also it it has been a cycle right like I I I'm slightly younger than you but I do remember times when like Visual Studio was a very expensive bit of software and I do also remember times in the 90s when you were like paying for compilers so so the fact that in like 2000s and 2010s we went into this like everything is free and everything is open source uh that has been rather a unique moment in the history of software development.
• 27:39
  SPEAKER_01: Yeah and one that we can't let die that one that is very important to defend as well and it's like open rate models and these kind of things.
• 27:47
  SPEAKER_03: I mean you know you know me like I do have background in open source so uh yeah I'm I'm always happy to do open source.
• 27:55
  SPEAKER_01: This was also interesting that basically uh uh codex and some other Hermes and Claude actually installed unowned code inside corporate networks so basically uh all the things that we as developers had to suffer that basically we had audits for everything that we did.
• 28:11
  SPEAKER_01: People don't do that with their coding coding agents.
• 28:14
  SPEAKER_01: So um yeah you can't trust them to actually only install things that is supposed to be on your uh on your network.
• 28:21
  SPEAKER_01: So if you're in enterprise probably a good idea to actually think about twice before you actually allow full access to your machine.
• 28:28
  SPEAKER_03: The attack vector in this one was very very interesting because uh researchers found that there were some like documentation files that were pointing at uh packages that were not existing yet and they've claimed claimed those names of non-existing packages that were mentioned in the documentation and then in the next cycle uh agents actually installed those packages that were previously not not existing they were just documented for unknown reason uh and yeah that's that's very that's very interesting very interesting angle of attack slop squatting it's called that's actually the the attack name of it uh much like much like typo squatting we had before when basically you just had an npm module that's basically an npm package that's just one character off or something like that.
• 29:17
  SPEAKER_03: Another big thing go ahead yeah no go ahead no go ahead I feel like more and more companies will try uh to to to be more careful about dependencies like inside Microsoft we have this whole thing that we are not actually using the public version of npm slash nuget slash whatever is your package management uh software it's like we are pointing we we do have like internal versions of of the of the feeds of the package feeds and then we use we can use only only dependencies from those and then adding adding stuff to this internal version means that he that it needs to get reviewed and so on and then uh new versions are published only after like seven days since getting published on the public feed so there is this this period of time uh which gives some space for people to to to review stuff and make sure that uh everything is safe.
• 30:18
  SPEAKER_01: Then you're humming oh sorry that was probably where I touched my microphone apologies.
• 30:25
  SPEAKER_01: ASCII smuggling was another attack vector that was originally just for AI prompt injection but now there is a phishing uh problem with that one as well and that's some Microsoft research that is really interesting about it explaining what ASCII smuggling is like putting random characters in there that are basically invisible and that way giving another message inside a text uh people use that to do uh to do like um yeah injection attacks and now it's actually phishing to basically filter out blocking email clients kind of thing I mean the best way to actually the the best way for a malware link is always the unsubscribe link that's basically the best way of doing it but yeah the the only the only thing I have the only comment I have about this one it's called ASCII smuggling but in in practice it's like about Unicode characters so the name is UTF 8 smuggling UTF 16 smuggling actually in most cases yeah yes but yeah that's I mean I I I shocked Dan with that one in the past when I showed him like UTF 8 and like Unicode my favorite is when you do split javascript split on uh on emoji and then you turn for example the family into four emoji and these kind of things it's just yeah it's it's a mixture of text and binary that is very confusing for people.
• 31:43
  SPEAKER_01: And it's something you just never really think about is it you know when you show we first showed me that it was like it seems so obvious but it's just something you you know you don't really think about too much yeah this is really interesting because the United Nations now finally have uh gotten rid of the uh of the world map that we used to have like in a long time ago I mean at one time it basically the first maps were drawn in England so England was the center of the world and that means that like the projection of the earth looked like Europe is much bigger than it is America is much bigger than it is and Africa is actually much smaller than it is and now the United Nations have said like no new world maps coming out will have to have a different turn so basically it shows the true size of Africa true size of other things every uh uh well almost every country has said okay to this there was one that said no I I can't remember what it was right now but like yeah it was interesting to see but this is actually this is actually the true size of is a really fascinating slide that I used to play along with as well like this is basically America inside Africa this is India inside Africa and this is China inside Africa.
• 32:50
  SPEAKER_01: So basically you can drag one country on top of another to see the difference between the sizes of those and uh it's just fascinating to see because people don't realize just that yeah it's it's basically uh the world is a ball well it's not it's a spheroid it's not really a ball but like uh it's it's fascinating to see how these things change according to where you go on the planet like how big can you can can you drag and drop Greenland I think you can but I can't remember how it's done right now.
• 33:20
  SPEAKER_01: Oh never mind there you go there you go Greenland popular search yeah let's drag it into Africa yeah there we go there we go yeah uh also it's it is important that that markator which is this like standard projection it's not actually bad map in principle like it is useful like the main reason it it got popular uh is because it's great for local navigation because it preserves angles and it's great for yeah uh so it it has some characteristic that makes it very useful for uh especially for local navigation especially in uh uh in in ship navigation uh but obviously kind of like uh it has problems every projection have problems there because because because every single flat projection of spheroid will create some kind of inconsistency uh but it's it is what it is but you know I'm actually happy about this change like it it doesn't yes we have lesser pollen than greater poland uh because we need multiple polands uh yeah and I mean uh well yeah yeah I mean that's that's the that's the interesting thing about the uh the German national anthem we actually only sing the third verse of it because the first one is like Germany is better than everybody else and we agreed after the second world we shouldn't say that anymore and the second one actually uh describes the borders of Germany at the time with the where the anthem was written so singing that one would mean that half of the Netherlands and half of uh poland would still belong to Germany so that's why we can't sing that anymore so that means like if you write an anthem for a country don't fix it with like geographical landmarks because they might change because yeah I mean borders are just random numbers on the map like it's they can change all the time but I found that fascinating in terms of like when people get like show how super exciting their country is and how big the other ones are so it's fascinating to see it's a wonderful little website to play with.
• 35:27
  SPEAKER_01: Yeah this is also really cool uh the top 10 web design styles of 1993 it's basically when the internet started and it was just like a document sharing thing and it was basically the home pages of all these different professors that were in 1993 already on the internet and what they looked like and he analyzed the the styles of what they've been doing like what kind of like uh how the URL was defined with tilders like um priorities the coloring what people chose like what what colors they use in Netscape back then and also the backgrounds they changed and like some of them even had like a tiled background like green white paper.gif and these kind of things and yeah it's uh I mean I my I did my first website in 96 I think and before that I I read email on a Commodore 64 with uh with a modem and uh yeah and it was just like I mean these were so common those wonderful animated things.
• 36:26
  SPEAKER_06: The mailbox.gif I think I used as well uh and the rotating dot gif I've seen before as well but it's so cool to see that kind of like styling and also the layout the framesets that they've used so yeah people just forget how much how old that stuff is I did have one point on that Chris the the rainbow hrs I I I remember them I remember all of this really from from I mean my my first website was 2002 so it was a few years after this but this would have been what inspired me to make them right in and kind of seeing all this stuff but but the rainbow HRs I I was that um was that a trend or was that just like uh the the kind of baseline style because it says in there like HRs got transformed into rainbow and I was thinking what what what when they say got transformed it was it was it was it a trend or was it just that that was the I would say it's a it was a trend like yeah basically this is still one of my most liked uh CSS things I've ever done.
• 37:18
  SPEAKER_01: I did the CSS3 rainbow dividers when CSS got animations and uh uh the Leia Baru worked with me on that one and uh just for shits and giggles I've done that and yeah it was it was a big GitHub success back for me back then so I've done great things to the world I gave I gave CSS V rainbow dividers so you don't have to load a GIF.
• 37:39
  SPEAKER_01: Well there you go and this is why we say I'm getting indexed by AI.
• 37:44
  SPEAKER_01: Nobody ever thanked me that's really really annoying.
• 37:47
  SPEAKER_01: There's also gallery of web design history that basically showed you how websites looked like when they came out like Google had this thing.
• 37:55
  SPEAKER_01: Do you know Christoph why Google has these colors no idea it was because their computer the first computer that the desktop machine that the search engine was running on had a broken uh uh uh broken frame basically the outside was broken so they built it with Lego so they put the computer inside a Lego uh uh bricks uh controller and that's that those are the colors of the Lego bricks fascinating it's I I've never heard this story yeah it's in the basement in Stanford University I took a picture of it at one time and this note don't show it to anybody and uh I it got into a book about Lego called the Art of the Brick and my photo because I put it up on Flickr with Creative Commons was actually part of that book and they sent me a free copy of the book because of it as well.
• 38:42
  SPEAKER_01: Wonderful so that's amazing yeah that's why that's why the colours of Google are that because they're the colours of Lego bricks.
• 38:48
  SPEAKER_06: I mean the Apple website was today today I learned today I learned yeah I mean what I like about these is everybody remembers the Space Jam website but uh the Amazon website is the one that I think kind of people often forget quite how rudimentary that was and to kind of see that you know what it's grown into obviously I know we talked to um uh Werner Vogels before and things like this about kind of yeah where what it's developed into but I mean these these are just amazing I love this so much and I think Microsoft one is done more bonkers as well it's like it's like half a cd you know yeah the X Files one weirdly I remember that yeah I I remember that X Files and what was the other one that was famous uh the Jurassic the Jurassic Park the one that got hacked uh right into when they've changed the the the T-Rex into dark or something like that that was like the famous story uh back in the those early internet days I did I did mcdonald's co.uk from like 2001 to 2003 I was the main uh main developer of that one and it still had that yellow text on red background so by working on that the whole day my head was hurting it was just absolutely atrocious to do that I mean I love that the you can see the background gray that basically they didn't set a background color that's why that's what basically Netscape had as a background when you didn't set a background color on your thing.
• 40:09
  SPEAKER_01: So that's why they all look like that wired what else do we have here?
• 40:14
  SPEAKER_01: Sandgarden of course I I had two submissions in there as well a list of heart I was an author of as well uh that was quite good and uh I'm I'm talking to Sedman on uh on Blue Sky a lot but he doesn't want to speak at events anymore he said like I can't be outbothered anymore sadly enough it's a shame because he he's a wonderful wonderful person.
• 40:34
  SPEAKER_01: Metzo blue I mean that was he then did the design for blogger.com and I think that was the first talk where they actually showed uh the power of CSS compared to power of inline styles or in like background color and fonts and these kind of things and he's he pitched it by showing them how many CDs of data he's saving by actually using CSS instead of inline phones.
• 40:56
  SPEAKER_01: And there was a really good talk in 2006 I think or something like that.
• 41:01
  SPEAKER_01: Newgrounds conspiracy games there's so many we could go whole day just looking at those yeah but that's also the internet internet history roulette which I found really cute where basically it just gives you different websites like what they look like and then it has these like pop-up facts like uh mt used to have like pop-up video kind of thing so you can see like different websites oh it makes this dog pile with a search engine as well meta search engine I remember yeah that was a pain I got listed on I love that when you press the button you can't hear it it actually gives a dial-up sound to connect to things yeah so that's that's a really really cute website just to see the history of like brush report no idea what that was wow I love this really nice as well yeah and it gives you we it gives especially like younger people window into how the how it all started right because you know and why we're so defensive about it you know like that's the other thing like Craigslist is there any point actually putting that as the way back machine because I'm pretty sure the UI is still the same now right it is the same yes yeah just go there go to Craigslist yeah it changed the content like there's there's several uh things that are not available in Craigslist that used to be available there was a whole dating part of it that's not available anymore but yeah it's interesting oh fark.com I remember that one holy crap okay uh this is also quite a cute it's an indie uh developer simulator where you basically you go in and you you basically all the things that you have to do to start developing and what you want to build I want to do a Pomodoro timer and then it actually tells you to be like la whatever wait a minute that install was way too quick that um that um Xcode install because I think we talked before about isn't it something like an 11 gigabyte install and it takes forever um so you know they they should probably make that a little bit more realistic just make it sit there for for two days.
• 43:03
  SPEAKER_01: Oh that's good.
• 43:04
  SPEAKER_01: Jesus Christ so it basically it it gives you like it already makes you stuck at like getting a name that isn't copyrighted and stuff.
• 43:11
  SPEAKER_01: So it's a really frustrating experience how to be an indie developer as a simulator.
• 43:16
  SPEAKER_01: And I found that really really cute as well because I mean Dan you you're an indie developer that has a few projects running.
• 43:22
  SPEAKER_01: Did you feel the same way?
• 43:24
  SPEAKER_06: Yeah I even looked into kind of doing some iOS and um and for the for these reasons mostly decided against it um because you know it's obviously the fact that you're kind of mostly tied into the App Store.
• 43:33
  SPEAKER_06: You've got all of the uh bureaucracy let's call it that that goes with it and getting on there um and you've got your nice uh$100 uh subscription to it on top of that so you know you've you're kind of when you've got so much available that's out there for for free or for kind of um you know less than that let's put it like that and you're not as tied in uh yeah I can I think this does a really good ex uh good good kind of case for why you you you might not lean into that and I just love building for the web you know that's that's
• 44:00
  SPEAKER_01: beauty of the web really it's free and open to everybody so yeah this is really really hilarious because it gives you all the things that you that you might not think about when it comes to releasing software and yeah it's a it's a really frustrating experience but that is kind of like when the development has become in a world where you have to pay to play kind of thing.
• 44:19
  SPEAKER_03: But also anyways uh also it shows uh the the thing that we've mentioned before the that kind of like just building software itself is just the first step and then there is this whole process of getting into the app store marketing it getting it to actually to users like that there is this whole like the the building software is is just the first step if you want to be indie developer doing doing the things on your own yeah I mean the I mean the main thing like I mean seeing like that the name is part of another company and kind of things that's kind of like overshooting the mark because you will not get sued for that I guess unless you call it google.com or something like then it's a different story but like I think the the it's interesting to see where that goes.
• 45:04
  SPEAKER_01: Now it's time to play it's time to play a game that we call uh fake on news where I show you headlines and I read them out to you and you have to tell me if these headlines are true or not.
• 45:15
  SPEAKER_01: We're gonna play that live at the Congress as well in America.
• 45:18
  SPEAKER_01: So uh it's gonna be interesting to see if Dan's like randomly uh white coded app will hold up to like 100 people doing it at the same time.
• 45:27
  SPEAKER_01: But for now let's do Christoph on his own to actually play fake or news with us researchers discover prison gang like structures in agent swarms where agents protect each other and attack others I will go with news on this one.
• 45:49
  SPEAKER_06: Are you confident?
• 45:51
  SPEAKER_03: Yes no I'm not but it's it sounds plausible to me.
• 45:55
  SPEAKER_03: As someone dealing with agents a lot that seems plausible it does.
• 46:00
  SPEAKER_01: So we'll go for news and the correct answer is oh fake I'm surprised number two can help reduce those white contrails behind airplanes that has to be fake surely what do you think?
• 46:26
  SPEAKER_03: Yeah it must be fake yeah let's go fake I've got a horrible feeling every time we say it must be fake it's used so yeah all the uh all the all the tinfoil people are now fans of Google and AI life forms can survive on significant regions of the moon study finds well it depends on what totally depends on what it means significant by significant and life and life form I guess yeah but yeah what do we think no news news news news news last minute and it was news well done number four Tesla RoboTexis will offer cheaper rides for developers who sign on with their GitHub account surely not Tesla RoboTax surely not why why would they it's data minus it's data it's expecting yeah it's news oh okay I like it sounds like let's have a look probably I'm I'm really bad at it at this game as we said before we're not sure if being bad is actually a good thing right now because these headlines are so bonkers uh so question number five antropic to release clawed cryptocurrency to ease token payments no new i i think news uh no i think fake sorry fake fake fake fake fake fake even you're getting confused now aren't we yeah are you sure fake yes fake I think anthropic is too smart to get into like given how bad a reputation crypto has nowadays I I think that they don't want to deal with stuff like that you're probably right number six Dyson debut a 499 dollar ai powered toothbrush with a built-in camera news yeah I've seen this but was it 499 dollars there's always something Chris will catch us at with the tiniest details so well I it's yes I I have no idea about the price but uh the AI powered toothbrush with built-in camera that sounds plausible so I just assume it's news and it is news number seven there are carrier pigeon apps now because pigeon technology is actually superior what uh that screams over Chris headline but then every time I say that I'm wrong so remember there was an HTTP HTTP standard uh uh over carrier pigeon as well that's that was an April fold but like yeah five seconds left fake fake it's actually news Jesus Christ that's crazy you can call me Chris you're doing well you've got a streak of two as well so you're on the leaderboard at least bytedance reportedly testing hyperlocal commerce ai for small scale delivery robots in Southeast Asia that sounds that sounds plausible yeah news yeah byte dance are an Asian company as well I believe so that's yeah obviously they're they're the Chinese I think or like something like that and they're into AI a lot so it's like he got us he got us he got us yes it's too good too good at coming up with crazy headlines tongue control device offers an alternative to mouse or keyboard news yeah yeah that does feel like it absolutely is uh an alternative and it's true so for the first time on lick is not a typo Microsoft cracks down on employee AI use after one worker spent$28,000 in 28 days so I know that one worker uh spent indeed$28,000 in 28 days but I don't know if Microsoft cracks down because of that I haven't heard anything about being cracked down on my AI usage.
• 51:07
  SPEAKER_01: Uh oh three seconds it's fake it news they did crackdown probably not on GitHub next people because they they have to do research and stuff.
• 51:20
  SPEAKER_03: Also so so the article that this was mentioned like there was another interesting thing is that the med because it was kind of like some test on some selected group of developers it was not across whole company and the median uh spent was something like$300 per month which seems which to me as as someone that's kind of like deep into the latest greatest state of the art seems like very low amount of of tokens spent especially for people that has access to basically unlimited tokens yeah which I can like I can more relate to the 28000 figure than to 300 in 28 days.
• 52:01
  SPEAKER_06: Yeah I even wondered if the number was actually higher than that you know that was my kind of thinking was that the the number I'd heard was more than 28000 but who knows.
• 52:08
  SPEAKER_01: Again this was on on some selected group of employees uh I I know people that are probably spending more uh on tokens than that uh because they are running like whole swarms of agents uh yeah oh number 12 then openAI has started letting some customers pay only when the AI works fake no it has to be it has to be fake no it's not but how do they define that AI was uh oh my god yeah that's a good question is like people not complaining or not restarting the session or finishing the session off I mean that's always the question like when you don't disagree with the machine anymore was it correct or are you just so annoyed that you don't want to talk to it anymore yeah that's the thing there well that was fake on news you've done well uh I guess you are on third place yeah that that was absolutely absolutely super difficult like having those having those in 2026 yeah it's it's it's crazy it's very tricky so yep cool so uh let's talk about what you're working on uh so basically we talked about stop doing the music thing it's yeah I don't know what what's going on there so you'll be working on a product called Chopin which is basically is the airport in Warsaw right yes and and some kind of old guy that did some music like I don't know anything about that yeah it's just random name uh yeah so so indeed Chopin uh I think that we we are at the point in time where we can agree that just writing the code is a relatively solved problem in software development like we all use agents those agents are like writing code faster than we can they are writing more code than we can uh and in general like we are not typing code anymore in our IDE manually right like I last time I've typed code manually into my IDE was I don't know eight months ago uh however we still want to control exactly what's going on in our with our agents and we often do this workflows of using for example plan mode uh in in our agents where you kind of like figure out plan and you you you create every detail and you discuss with agent like what's what's the problem that you want to solve and what how you will solve it.
• 54:49
  SPEAKER_03: And then when plan is ready then the implementation part of the thing is kind of like okay you just give it to a to to coding agent uh and yes obviously reviewing the code is is is is very difficult.
• 55:02
  SPEAKER_03: So existing tools for the for planning have limitations all our existing tools that we use in in in ai all those agents are all single player so it's just single person sitting in front of their terminal talking uh they're non-interactive uh which I mean you get wall of text right in your terminal or in your you know graphical user interface uh so we want to to have something yeah and there's a couple of examples here how it looks like uh you can use skills like grill bean right like this is one of the most famous skills probably right now uh where the agent will keep asking questions about the plan and and and do those like question after question after question after question uh however all this is getting lost like this this this content is done somewhere in your terminal and then it kind of like disappears you you do implementation you open pull requests and then you merge pull requests without reviewing because who would review hundreds of thousands of lines of code that agents produce nowadays and you just merge stuff but there is no log of decisions there is no there is no interactivity in in the process there is no multiplayer in the process and this this has been topics that have been on my mind for a while now so especially collaboration software development in principle is just team sport right it is collaborative problem solving when we remove all the lingo the magical nerdy programming languages tools technologies software development is about getting group of people together into the room or kind of like virtual room uh and then telling them and those people are probably from different backgrounds they can be engineers they can be user experience designers there can be project managers product owners domain specialists and we tell all those this group of people to to to to think together and solve some problem that is what software development in principle is however the current tools are changing this very much because they are moving to this like very individual focused workflows so I decided that we want we we want tools that are actually allowing us to still collaborate in the context of AI tools that are interactive that have real-time multiplayer collaboration tools that are working on this higher level of abstraction which is managing plans and documents because that's how we work nowadays with AI like Markdown is new programming language of the 2026 it's probably the most popular language of 2026 uh so this is what Japan is it's basically imagine Google Docs we've built in we've built in ai uh with decision slogs so every decision that you make during chat with AI or every comment thread that you that you resolve is actually saved as a decision and there is log of decisions and then parts of the document or the plan can be linked to the decision so you know how to how decision relates to the plan.
• 58:12
  SPEAKER_03: Uh put together into nice and it's all put together in nice web app that you can host on your own in it's open sourced.
• 58:20
  SPEAKER_03: That's basically the the the the short version the the the the short description so people can take it for a spin now and install it on their own machine and get started?
• 58:32
  SPEAKER_03: Yes absolutely uh so uh I'm not sure if it's secret but kind of like I will share it anyway.
• 58:38
  SPEAKER_03: Uh so github at GitHub next we we are now in this period of time where we got permission to do like more open source stuff and more experiments in the open rather than trying to be super and then trying to have those like big announcements at the universe or like whatever Microsoft build instead of this old school strategy we will be doing a lot of open source and shop on is probably the first of our projects that can like it was actually getting ready while the decision to make more open source stuff uh has been made so we've been ready to go uh we've published it there is a link to the repository at the bottom of the of this blog post this wonderful blog post that Maggie uh my coworker has written uh and she also did videos of it like with voiceover so you can watch it there and see what it does before you install it but that's really nice yes absolutely and uh it is web application it it has backend that you can host on any Docker hosting thing it it's just Docker compose app and you have think running on your back end it's it's a web application with Postgres SQL uh uh database uh there are instructions to run locally you need to create like GitHub app to to get GitHub authentication uh it is all about like Google Docs for software development so it's about managing managing documents in the context of your repository uh it's about again as I've mentioned a lot about a lot of effort is put on collaboration because I think we need to get back to collaborating when building software which is something that we've lost over the last couple of years with with Rise of AI because no single developer sitting in front of their terminal will know full context of every problem that they are trying to solve most problems are too big to for that uh so yeah uh and then it's also what is important is that this is like just purely research prototype like there is no path for this particular thing to become like real GitHub product uh it's it is space for me and Maggie and the rest of people in in GitHub next to to experiment with the ideas.
• 1:00:58
  SPEAKER_03: I think that it's very clear for everyone that next six to twelve months will be about product like this.
• 1:01:06
  SPEAKER_03: So this multiplayer markdown editors there is a couple of startups doing that uh again this probably is a secret but I will share there is a team inside GitHub that is also working on on multiplayer markdown editor uh product to become just part of the GitHub as a platform uh github user interface as a as a platform uh and that's all I can share about it right now I guess uh that's what it is you can always find a new job that's yeah I know I know well in in the worst case I will just come to you Chris and be like Chris you got me fired now now you need to hire me at we are developers I can do public speaking or something.
• 1:01:50
  SPEAKER_03: So this is for us is it's like a research space uh we we are going to next GitHub next in in in general is a group that is trying to figure out not what not what is happening in the next six to twelve months because that's usually fairly obvious to everyone but we try to figure out what is happening further in the future we had this this term before horizon free problems uh where horizon one is the kind of like immediate problems horizon two is this like six to twelve months we need to we need to build new product type of things and horizon three is the crazy ideas in the future uh so again every everyone knows that there will be markdown multiplayer with ai things showing on the market in the next six months uh and we need to figure out what will be the future ways of collaborating and future ways of managing your documents uh we believe a lot into into visualizations uh because when you have too much text uh it's getting difficult to read uh Maggie especially has background in in like visual stuff uh so so she she she's doing she's doing a lot of of visualization explorations uh yeah and and it's it's it's it's it's a cool project especially that I love that I can do open source again uh as a priest i i've been kind of like i i did i did a lot of open source back before uh joining joining evil evil corporations uh and i'm now happy that i can i can do do that again uh so i hope that people will vibe with it uh yeah the blog post the blog post is great i i do recommend to everyone to to read through it uh it's there is fun meme at the beginning there is there is nice yeah this this about that's it though that is that sums it up yeah it's very there is uh i think the explor explanations in the blog post are probably way better than what i can do uh and yeah uh well give give it a read and it will be interesting to get maggie on the show or some uh in the in the near future as well or maybe if you if you got more feedback from people what do you want from people to actually take part in it give it for a spin uh can people contribute to it are you interested in contributions as well so we haven't decided yet about contributions uh in aw agentic workflows other other open source projects from github next we have very uh specific style of of getting contributions we which is we only accept uh open issues uh and we want those issues to be well researched and kind of like not just something is broken fix it but kind of like well researched issues and then if we get well researched issues we can easily throw agents at the implementation uh implementation is the code contribution is not necessary nowadays like I think the the biggest value of the open source is changing or maybe it always was uh about having good ideas and having kind of the taste of like what is supposed to do what the product is supposed to do and what what it should uh how it should work uh the code itself is now fixed so maybe we will go with similar contribution style that we will expect from people to just you know tell us what we should do and then we can throw agents at that that's not a problem.
• 1:05:28
  SPEAKER_03: But we'll see like we haven't decided but yes please take it for the spin uh tell us what you think about that uh we are open to feedback the main reason why we are open sourcing that is because we want collaboration and we want people feedback.
• 1:05:44
  SPEAKER_03: Uh it's as I've mentioned it is not a product it's not designed to compete with whatever startup is on the market nowadays doing similar things like we want to have knowledge share we want to you know make software development world slightly better Better place uh than it is right now.
• 1:06:02
  SPEAKER_03: And that's the goal.
• 1:06:03
  SPEAKER_01: Uh so and I I love the idea that you can install it yourself, that you don't actually need to rely on Twit on GitHub to actually use it.
• 1:06:11
  SPEAKER_01: So I think that's a that's a really good thing as well.
• 1:06:14
  SPEAKER_03: Yeah, it was it was it was a very important point for me uh to to make sure that that kind of like because just showing the code is useless, right?
• 1:06:22
  SPEAKER_03: Like I actually we need to give people ability to use it on their own.
• 1:06:27
  SPEAKER_03: Uh we don't want to because it's just research prototype, we don't want to go into hosting the instance for like everyone on the internet to use uh on our own, because that's obviously creates infrastructure problems and so on that we don't want to be solving.
• 1:06:43
  SPEAKER_03: Like we want to be researching the user experience part of the story rather than deal with the infrastructure of scaling this to whatever amount of users it would have.
• 1:06:53
  SPEAKER_03: Uh so yeah, so so internally we're also hosting it on using Coolify, which is like you know, standard uh standard uh open source thing.
• 1:07:03
  SPEAKER_03: Uh it's just Docker Compose app and it's working.
• 1:07:07
  SPEAKER_03: Uh and you need to set up some configuration, obviously, with like allow you can you need to create GitHub app to have out GitHub authentication uh working, and then you set up like allow list so your instance is uh is accessible only by uh by uh some people with given GitHub usernames or belonging to given GitHub organization and so on.
• 1:07:30
  SPEAKER_03: So there is some configuration to be done, but uh hosting it is very simple, and that was like one of my goals.
• 1:07:36
  SPEAKER_03: Uh and yeah, if you if you want to use it nowadays in your company, like it's MIT license, like you know, you can just take it and and do whatever you want with it.
• 1:07:48
  SPEAKER_03: Uh at the same time, yeah.
• 1:07:51
  SPEAKER_03: Uh at the same time, it is important to remember that it is MIT license, so it's just it is you know as is.
• 1:07:57
  SPEAKER_03: Like we are not really providing it's not official GitHub product, it's not handled by GitHub support or anything like that.
• 1:08:06
  SPEAKER_03: It's just you know, two of us doing experiment and sharing it with people.
• 1:08:11
  SPEAKER_03: So there are you know question marks and stars and kind of like you know those disclaimers everywhere.
• 1:08:18
  SPEAKER_03: But in we are we are using it internally in our team and it's working, so you know take it for it.
• 1:08:26
  SPEAKER_01: But that would be the insight to share as well, then like how how much more effective did it make you?
• 1:08:30
  SPEAKER_01: Like, what what what do people think?
• 1:08:32
  SPEAKER_01: Kind of thing.
• 1:08:33
  SPEAKER_01: That's hard to measure, but I think it's interesting to to uh to learn these kind of backstories as well.
• 1:08:38
  SPEAKER_03: Then it is it is it is useful.
• 1:08:41
  SPEAKER_03: Uh I think it kind of like moves some some of the communication that we used to have in random, you know, random threads on Slack or places like that on Teams, to the centralized place where actually your conversation is grounded in in the existing plan, documents, uh RFC about PRD, about the feature product, like whatever, whatever we are building.
• 1:09:05
  SPEAKER_03: So I think that there is value that the the information is more centralized.
• 1:09:11
  SPEAKER_03: Uh at the same time, you we are a weird team.
• 1:09:15
  SPEAKER_03: We we we are team well while I was saying that software development is very often a team sport, and kind of like in principle it's like collaborative problem solving.
• 1:09:25
  SPEAKER_03: Uh we are a group of crazy researchers, so we often do this individual kind of like go in on individual adventure before collaborating with people.
• 1:09:34
  SPEAKER_03: Uh, but kind of like, for example, for the project like Shoppen, which is slightly more established, and we are actually we know what we have in our minds, and uh we do use it.
• 1:09:44
  SPEAKER_03: We do use Shopen to build Shopen.
• 1:09:47
  SPEAKER_03: Uh so for example, right now we uh one of the features that we've been discussing, I think it's also shown on the videos that Maggie did, was is adding the drawing canvas to the Shopen.
• 1:10:00
  SPEAKER_03: So your documents are not only just text documents, but there will be also like you can actually draw diagrams or something, like sketch diagrams.
• 1:10:08
  SPEAKER_03: And there is like we have like this like long document discussing exactly how it is supposed to work, and and especially in the context of real-time multiplayer with like CRDT.
• 1:10:18
  SPEAKER_03: So we are going into fairly detailed, uh detailed decisions there, and and then having document, having the decision log, which is again this kind of like every question that agent asks you is and every your your every answer is is recorded as a separate thing in the UI, which is called decision.
• 1:10:39
  SPEAKER_03: Uh having this is very useful uh because we kind of like easily just you know go to the decision tab and kind of like know what happened, it gives us the ability to reason about documents.
• 1:10:52
  SPEAKER_03: I've always in my old school days when I was still doing like programming languages, I always talked a lot about like reasoning about code, uh, and how how you know some languages are better than at reasoning about code, and some are worse, and so on, and how some programming paradigms are better than others for reasoning.
• 1:11:10
  SPEAKER_03: And here it's it's also bringing a lot of of my background of this background that I had to to just markdown document.
• 1:11:17
  SPEAKER_03: It's all about reasoning about like your markdown documents, reasoning about like what were decisions that you've made, reasoning about how those decisions are actually uh presented in the final document or the plan.
• 1:11:29
  SPEAKER_03: Uh and then we also have some ideas on another part, which is like what happens between the the the imagine that file is ready, how it is handoffed to the coding agent.
• 1:11:42
  SPEAKER_03: Because it's not enough to just go and give single agent your document and be like, do whatever you want with it.
• 1:11:51
  SPEAKER_03: Such document can can be fairly detailed, but also you know, it basically can describe a task list or or actually the task graph.
• 1:12:00
  SPEAKER_03: You know, some stuff can be parallelized, some stuff that is described in the plan can be parallelized, some stuff needs to be done in sequential matter manner.
• 1:12:10
  SPEAKER_03: So we we are also thinking about that part, that kind of like how do we create a task graph out of the document and so on.
• 1:12:18
  SPEAKER_03: So then you can hand it off not only to single agent, but can like to multiple agents at the same time to to to implement uh things quicker and faster and have better feedback loops.
• 1:12:29
  SPEAKER_03: So there is a lot of interesting, interesting things that we are exploring here.
• 1:12:33
  SPEAKER_03: Uh it's very early days, like we and it's on purpose.
• 1:12:38
  SPEAKER_03: We've we've released it very early, so it's like expect that things are half broken.
• 1:12:43
  SPEAKER_03: Some features are like, you know, some of the UI is still kind of terrible.
• 1:12:48
  SPEAKER_03: Uh but uh it's all about just sharing with people ideas, basically.
• 1:12:53
  SPEAKER_03: Uh we want we want we want to show our ideas to to to greater community and to hear from community and then see what what is what's what's what's what comes out of it.
• 1:13:04
  SPEAKER_03: Uh collaboration is collaboration with people is like my thing.
• 1:13:08
  SPEAKER_03: It's always been my thing, so uh so I'm happy to do that.
• 1:13:14
  SPEAKER_01: Excellent.
• 1:13:14
  SPEAKER_01: So give Chopin a go.
• 1:13:16
  SPEAKER_01: Uh it basically you you can install it yourself, you can try it out.
• 1:13:19
  SPEAKER_01: I think the idea is something that is coming in the nearer future.
• 1:13:23
  SPEAKER_01: I mean, Google just announced that they're doing much more with AI in docs itself.
• 1:13:26
  SPEAKER_01: I don't think it's gonna be in that direction, but uh I wouldn't be surprised if like uh if like Claude or things like that will do something similar as well.
• 1:13:35
  SPEAKER_01: So this is a good research thing.
• 1:13:37
  SPEAKER_03: Yeah, I as I as I as I mentioned, like I think it's it's obvious that this is like what's what's happening in the next six months.
• 1:13:44
  SPEAKER_03: Uh in the uh that there will be there is a bunch of startups working on on very similar products.
• 1:13:50
  SPEAKER_03: Uh I wouldn't be surprised if if both Anthropic and OpenAI added similar features to their uh hardnesses.
• 1:13:59
  SPEAKER_03: Uh and Google obviously is in a very good position to build similar things because they already have Google Docs, uh which is like they have Head Start on everyone else.
• 1:14:11
  SPEAKER_03: Uh but you know, we are scoping ourselves just to software development, uh, which gives some advantages.
• 1:14:18
  SPEAKER_03: Uh we can be uh we can we can be a bit more specific about like user experiences that we want to build rather than Google that needs to support every possible imaginable scenario with documents.
• 1:14:31
  SPEAKER_03: Uh so yeah, uh I think it's it's uh it's great space, it's great, it's great area that kind of excites me.
• 1:14:38
  SPEAKER_03: And I think it will be exciting, uh exciting times for developers uh as we move towards this coding with markdown more, which is I mean, whatever it is, what it is.
• 1:14:52
  SPEAKER_01: Well, I mean, there will be people from uh from GitHub at GitHub next as well at the Congress, so it's gonna be interesting to see what they have there.
• 1:15:00
  SPEAKER_01: We're we had a lot of technical issues, so we're quite over time, but it doesn't really matter because we still have one tool that we want to show with you that we actually had in the earlier part of the show.
• 1:15:10
  SPEAKER_01: So here's a quick tool that I've been using every day that's super old school but actually really, really useful.
• 1:15:16
  SPEAKER_01: So if you want to take a look at that, it's um it's a quick two-minute thing that might actually help you clean up your hard drive in ways that you haven't known before.
• 1:15:31
  SPEAKER_01: When it comes to organizing a lot of files, the file tools of the operating system, Finder and Explorer, soon become annoying to use, especially for a dedicated keyboard user.
• 1:15:41
  SPEAKER_01: Traditionally, the tools to make moving tons of files and dealing with complex file structures were two-pane file managers, like Notten Commander, WinCommander, Total Commander, and the like.
• 1:15:51
  SPEAKER_01: Now we have Martup, which is a free native macOS two-pane file manager written in Swift by Jan Solanov.
• 1:15:59
  SPEAKER_01: It allows you to easily pick and discard files with your keyboard and move and copy them around.
• 1:16:05
  SPEAKER_01: Another excellent feature is that you can open a terminal in the current folder without having to switch to another tool in case you need even more power.
• 1:16:13
  SPEAKER_01: Mata is macOS native and can just be installed.
• 1:16:17
  SPEAKER_01: It is fully keyboard accessible and doesn't collect any information.
• 1:16:21
  SPEAKER_01: It treats archives like zip and RAR like folders, so you don't need to unpack everything to see just one of the files.
• 1:16:28
  SPEAKER_01: It is fully customizable and comes with several themes.
• 1:16:32
  SPEAKER_01: There's also a plugin API where you can extend the tool with your own Lua plugins.
• 1:16:37
  SPEAKER_01: You can access all its features with a shortcut action panel.
• 1:16:41
  SPEAKER_01: And there is a quick search to find files globally and open their parent folder.
• 1:16:46
  SPEAKER_01: Another extant feature is disk user, which gives you full insight into how big different folders are and why things are slow.
• 1:16:53
  SPEAKER_01: One feature I missed that older two pane file managers had it's comparing folders and automatically highlighting the differences.
• 1:17:00
  SPEAKER_01: Good thing that Martha is available on GitHub where you can request new features.
• 1:17:04
  SPEAKER_01: So I do that right now.
• 1:17:08
  SPEAKER_01: And that's Marta.
• 1:17:09
  SPEAKER_01: And I use it every day.
• 1:17:11
  SPEAKER_01: It's actually really incredibly useful and feels so old school to do it, but it's still the most effective way to get to plow through hundreds of files that have been named randomly.
• 1:17:20
  SPEAKER_01: And uh I uh we're all accused of that basically called things final.
• 1:17:25
  SPEAKER_01: Now really final.
• 1:17:26
  SPEAKER_01: Now it's the final one, and really, really use that one instead, instead of using version controlling for some reason.
• 1:17:32
  SPEAKER_01: But uh what I also found interesting is that there's another engineer in Japan because we the tool we had last week uh was also NeoDisc, was also by somebody in Japan.
• 1:17:40
  SPEAKER_01: He works that engineer Jan Sulanov works for uh JetBrains.
• 1:17:45
  SPEAKER_01: Uh, it's interesting to see that there are so many people there that obviously are not Japanese named, but actually work there.
• 1:17:50
  SPEAKER_01: Must be an interesting space to work.
• 1:17:53
  SPEAKER_01: And uh I found it really cool to I I don't know Lua, that was my only problem.
• 1:17:57
  SPEAKER_01: I would have loved to extend it in another language, but seems like Lua and Swift seem to have a connection.
• 1:18:03
  SPEAKER_01: Oh, I don't know why they did that.
• 1:18:05
  SPEAKER_01: But yeah, just throw just throw agents at it.
• 1:18:10
  SPEAKER_01: Fair enough.
• 1:18:10
  SPEAKER_01: Yeah, that would be one way of doing it.
• 1:18:12
  SPEAKER_01: But it really helped me to clean up a lot of like files that uh that are be basically a mess, like old hard drives and these kind of things, which in Finder with just a trackpad would be a total pain to actually do.
• 1:18:24
  SPEAKER_03: No, my my first thought was that it's it's got like Norton Commander, but for Mac for like native Mac instead of just without.
• 1:18:33
  SPEAKER_03: It seems fun.
• 1:18:35
  SPEAKER_01: Yeah, I mean there were there were other ones before that I had on Mac, even on Mac OS 9, that actually were like Norton Commander.
• 1:18:42
  SPEAKER_01: I use Total Commander on Android, weirdly enough as well, because it's available there.
• 1:18:46
  SPEAKER_01: And you can even run it as an FTP client to actually send files from your hard drive to your uh to your Android that way.
• 1:18:52
  SPEAKER_01: So uh lots of old school stuff going on there.
• 1:18:55
  SPEAKER_01: So um it's it's a cool tool to play with, and I think it was fun to actually just do that little tool.
• 1:19:01
  SPEAKER_01: There's also an article on the magazine if you want to go and see and uh try it out yourself.
• 1:19:07
  SPEAKER_01: And yeah, it's free.
• 1:19:07
  SPEAKER_01: It's always a good thing to have on your machine.
• 1:19:10
  SPEAKER_01: So yeah, we're uh we're at the end of the live show.
• 1:19:12
  SPEAKER_01: We're we're getting ready for America, we're getting ready for the show.
• 1:19:16
  SPEAKER_01: In case you actually want to have a free ticket, uh you can actually use that QR code on the post uh developers welcome underscore welcome, and that would give you a free ticket to come to the Congress.
• 1:19:28
  SPEAKER_01: Not a full-length ticket for all the VIP sessions and stuff like that.
• 1:19:31
  SPEAKER_01: If you also want to access the speaker booth and you want to access the uh uh the VIP lounge, become a moderator because we really need you as well.
• 1:19:39
  SPEAKER_01: But this is a free uh way to get a free ticket for the American event in two weeks' time, not even two weeks' time.
• 1:19:46
  SPEAKER_01: Holy cow! Yeah, yeah.
• 1:19:48
  SPEAKER_01: So that's what we're gonna plan for next.
• 1:19:50
  SPEAKER_01: So this will now actually get thousands of people to come more to the event.
• 1:19:55
  SPEAKER_01: But we have space, so that's good, and the more people, the merrier.
• 1:19:59
  SPEAKER_01: So thanks very much, Christoph.
• 1:20:00
  SPEAKER_01: This was great.
• 1:20:01
  SPEAKER_01: Uh I hope you had fun and did not uh try uh scarred you for life with our with our fake or news.
• 1:20:08
  SPEAKER_01: But that was terrifying.
• 1:20:09
  SPEAKER_03: That was that was terrifying, but yeah, uh I will deal with it.
• 1:20:14
  SPEAKER_03: I will process my trauma.
• 1:20:16
  SPEAKER_01: I will process my trauma.
• 1:20:18
  SPEAKER_01: Exactly.
• 1:20:19
  SPEAKER_01: Real artists don't use AI, real artists use childhood trauma to actually be creative.
• 1:20:26
  SPEAKER_03: True.
• 1:20:27
  SPEAKER_03: Uh thank you for having me.
• 1:20:28
  SPEAKER_03: It it was a pleasure, and everyone should go to the World Congress.
• 1:20:32
  SPEAKER_03: I I've been a couple of times in Berlin and it's always been amazing.
• 1:20:36
  SPEAKER_03: So I imagine that San Jose will be also fun.
• 1:20:40
  SPEAKER_06: Cool.
• 1:20:41
  SPEAKER_06: Well, Daniel, any last words from you?
• 1:20:44
  SPEAKER_06: Um just to just go back to what you said, Christoph, you know, that like um at the at the heart of it, rather, you know, uh software development is a collaborative uh what was the phrase you use, collaborative problem solving.
• 1:20:54
  SPEAKER_06: And I really like that.
• 1:20:54
  SPEAKER_06: So yeah, anything that gives us more collaboration, so we're not just building on our own.
• 1:20:59
  SPEAKER_06: Um, I'm all for.
• 1:21:00
  SPEAKER_06: So yeah, thanks for sharing that with us.
• 1:21:01
  SPEAKER_06: I appreciate that.
• 1:21:02
  SPEAKER_06: Yep.
• 1:21:03
  SPEAKER_01: Cool.
• 1:21:04
  SPEAKER_01: Well, we see you next Wednesday, and then we see you in San Jose.
• 1:21:07
  SPEAKER_01: That was it for today.
• 1:21:08
  SPEAKER_01: Bye bye.