Francesc Campoy: It went really well, it was full all the way. That was very interesting, seeing so many people interested in talking about Diversity, and not only people that are members of minorities represented at GopherCon, black people and women, and LGBTQ etc., but there were also a lot of allies that were there to represent and basically to give support. That was very interesting.

We talked about many things. One of the things that I tried to propose and I hope is gonna happen - the same way there was this talk about mining the Go community, and that understands how Gophers do technical things, I was wondering if we could do something similar about who are those Gophers, where are they? Do they live somewhere in the States, in cities where they have meetups, or are they far apart? Do they have kids? Is it easy for them to attend conferences like GopherCon, or do they prefer things like the remote meetups organized by GoBridge? Is that useful for them? That was pretty interesting.

Then we moved into talking about how to make the Go community more approachable. There was a lot of discussion on basically improving the newcomers’ documents that we have right now. I could say they’re not bad, but we could definitely do much better. There were a lot of people talking about how to make those better, how to learn a little bit from the Ruby community. The Ruby community has a lot of funny examples; it is not only about technology, it is also about having fun while learning and things like that, and trying to apply that to the Go community. I think it’s very interesting.

There was also - I don’t wanna say his name wrong, Johnny Boursiquot. He pointed out that it is good to have those materials, but once we have them we need to go out to other communities and teach. He was talking how he’s going to schools that don’t have any technological plan, they don’t have a technical curriculum. And he goes there and volunteers to teach kids that otherwise cannot have access to programming classes, and he teaches that in Go. I think that is amazing, I’d like to see more of that.

I think that both things - having better materials for newcomers and then the Go community ourselves going out to reach into these communities that otherwise don’t have access to it could make our community much better, much more diverse and also much larger and welcoming.

Adam Stacoviak: Well, it’s line by lining us here… Let’s go back to what you said before, which was private AI is watching. I’m just kidding, it’s not watching you. It’s actually not watching you. So private AI versus this tractor beam, this gravity, this…

Jerod Santo: So you picked Stripe. These were the good days, when you could just line up offers, and you were lining them up, and good for you. Stripe was obviously an attractive – especially from the outside, it’s an attractive place to work. I’m sure the compensation was great, the tech problems are interesting…

Jerod Santo: It’s kind of the cleansing of – I mean, we’re talking about a correction. There’s a reason why that’s the word that’s used. Because something was wrong, you know what I’m saying? And it was wrong for a long time, and it just bubbles, and it percolates… And then eventually it corrects, and things get back to, as Sean calls it, the real world. And that’s his silver lining as well, is like actually now the companies who are focused and deliver on quality, and can actually ship a thing that’s reliable, that people want, and will pay for - they make money and they grow. And the ones who are just really good at convincing people with lots of money that they have the next big idea, and all they need is another thousand employees and that idea will somehow just start printing money… Like, that’s not always real. In fact, 99 times out of a hundred, or maybe more, it’s not real, which is why VCs are playing essentially an odds game. VCs play a numbers game. They invest in enough times that eventually they’ll hit the Uber, and that will make up for all their failures. And it seems like their money is still floating around though, it’s just floating around specifically AI startups.

Jerod Santo: Yeah, exactly. You try to get like silver linings out of it, you know…

Ashley Jeffs: Made it open source. It didn’t really do much. For about two years I was working on it and it didn’t really – it read from Kafka or ZeroMQ, and vice versa. I had a few brokering patterns for mixing things up… And then you could have like a disk buffer in there as well if you wanted it… Which I didn’t like. In my opinion –

Rachel Plotnick: Yeah, there’s a good section of the book that looks at torture testing and all of the kind of bizarre trials we put our devices to see what will they stand up to. And in fact, I look at programmers and think about all night coding, and kind of the desire to have food and drinks near the device, and how people have had to navigate that… Now I’m doing something on energy drinks and gamers… So I think the mixing of food and drinks and snacks with computers is super-interesting.

Jeff Smith: Well, I think music is very personal, and how you approach Gregorian chants is different than how I do. And what we designed was a system that’s looking for how you approach music and creates a baseline. You’re looking for deviation… And I can kind of go into all that. But it gets to know you over time. Creating that persona, is that - okay, you’re making certain choices in your music selections that are unique to you.

So you’ve got all this incredible data coming through. It’s danceability, it’s lyrics, chord progression, balance, happiness, beats per minute… I mean, there’s all this raw data that’s coming in. And if in a vacuum you’re looking at that saying “Okay, well, this is my mirror neurons lining up. I get it. This is how I’m feeling. So I’m choosing that.” Then you look at certain choices you’re making. “I’m fast-forwarding through this song. I’m skipping this. I’m adding this to the playlist. I’m playing this again. My beats are going up in the afternoon, because I’m working out…” There’s all of this usage data that then you’re looking at in comparison to that, and then it looks at it over time. So what you’re looking for is “Great. Here’s your baseline, and you’re trending, now, all of a sudden you’re deviating in a certain direction… So what does that mean?”

And so I’d say that one of the surprises - I think you asked that - in our discovery was, although music is incredibly personal and individualized, a majority of people still get hit the same way, with the same song. 60% of people listening to the song “Happy” will feel happy. I mean, it’s just something like that. And then you’ll get that extra 20%, 30% accuracy because of your individualizing it. But I thought that was really cool, that we could screen majority of population against a certain album, genre, songs… So that was one surprise.

And the other was the flip it around and how the music industry starts supporting us saying “Ooh, this is interesting. Can we use your algorithm to write songs, to achieve the emotion that the brands want to accomplish in these commercials or these movies?” I thought “Okay, that’s a whole different line of work.” And so I think it’s fun to explore that, the individualization, understanding how we’re addressing that through AI and everything, but also surprised by these other opportunities for the masses.

Tom Occhino: Yeah, when Jordan and I introduced it at that conference, I remember it vividly. One of the things we wanted to focus on was the sort of syntax, and how this thing that was very familiar to us we thought would be also familiar to the rest of the world. It’s like “Look, your HTML, the way that you want to think about your components - you don’t have to use these nested JSON structures, and all this stuff. Look, it just [unintelligible 00:07:46.01] But the sort of conversation ended up becoming about this mixing of technologies. And at the time, it was like “No, no, no. The whole industry is moving towards a separation of concerns, and you guys are mashing them together.” And we’re like “Wait, wait, wait… This is one concern. It’s my component. It’s my UI.”

[00:08:05.22] And so I think the biggest thing we did wrong there was we didn’t actually start with the problems that we were trying to solve, we just started with our solution. So luckily, Pete Hunt, I think later the same year at JSConf EU, did a much better job of teeing up the problem. Here’s what we were experiencing - as our applications got more complicated, as our models got super-big, people didn’t want to make changes to them… We wanted a loose coupling between components, so that those components could be used in lots of places, but high cohesion between them. So how do you do that? And it’s with the seam between components, which is the props boundary, and all these other things. So he did a much better job of teeing up the problem. And then I think it was more well accepted. People were willing to kind of give it five minutes after that point. But I’m glad that there wasn’t like a stampeding herd of adoption. This is amazing.

At no point in React’s history have we ever said “Behold, this is the answer! We’ve finally done it!” It’s always been about sort of like “Here’s the problems we’re facing. Here’s one solution that worked pretty well for us. If you have the same problems, let’s chat. If not, no worries. There’s lots of good solutions out there.” But I’m glad that Pete came through and restored faith in the sort of engineering team that Jordan and I were attempting to represent. Because the team was great, and the tech was novel. Something interesting and important.

Theo Browne: The absolute best type. I love him so much. He’s also like the nicest dude. We chat all the time. I love Will. He missed OG Maker Faire and OG VidCon. Nobody was doing anything like that. He was sad. He wanted the ultimate science fair with his nerd YouTube friends, and made it. And he hit me up, [unintelligible 01:12:43.10] I was like “Wait, this is engineering, but how much am I allowed to infect this with software?” He’s like “Software engineering is engineering. Do whatever you want.” So the first year I brought a few of my friends… It was the best conference I’ve ever attended, simply because it was creators and YouTubers who understand that dynamic, and the community management, and all of that well… As well as little things like when I showed up, they gave me a little index card [unintelligible 01:13:08.25] if I needed to get out of the crowd. All those little details they’d figured out. I never felt so understood as a creator at an event before.

So when they did it again this year, I dragged all of my creator tech friends to it. I told them - not mincing words - “You have to be at this one”, and I think that went really well, because all of them were hyped. They’re all trying to help fund it next year, because it was so fun. That’s the one conference I’ll be like “No, you have to go to this one.” I’m getting there with React Miami. That one’s really fun and cool, and I love Michelle so much…

This is my first time at a React Summit, and I can see how I could be convinced there as well… But I’m going to a conference because I have friends there. If I’m surprised by the quality of the experience I have there, I’ll become an advocate for it. But I don’t want to be the first person in hopes that my friends show up later.

Jerod Santo: Yeah. I mean, I think the magical thing as a coder is – it’s just obvious, everything lines up perfectly. And there’s just something very satisfying about that. Obviously, like you said, it’s kind of necessary in code, especially if you work in languages that have significant white space, stuff like this. But even in non-code contexts – of course, you don’t want everything to be monospace everywhere, because there’s other things to reflect, and to show… But there’s just something satisfying about easily lining up letters and numbers perfectly, without any sort of wiggle room.

Zac Smith: Well, I think that they might not own it, but they’re going to cause it to occur as customers. I think they’re very different businesses. Some are high margin, high growth, the other ones are used to being regulated, non-growth businesses… I think the big shift we’re having right now is that energy has gone from – basically, we reached peak load in the ‘90s, and it’s been slowly decreasing, or like growing very minimal, to it’s actually growing by a few percentage points a year, which is crazy, and we’re not used to it. So we might see some energy businesses go private, and turn into like growth businesses, basically. I mean, you can see Constellation, which is the nuclear operator doing that. If they figure out that they can build 10 or 20 more new nuclear plants, which hasn’t happened since like the ‘60s, it’d be a growth business.

So I think we’ll probably just see a lot more – well, we’ve already seen it, even in Ashburn, Virginia, which is the largest data center market in the world… Just the local planning board saying “Hey, I’m not going to approve this building permit, because you’re going to bring in 500 diesel generators, or you’re going to use our water. You’re going to use 50 million gallons of water a month. We don’t want that, actually. That’s a municipal thing that we need. We want no diesel generators next to our school, and we need the water for our other stuff.” And so I think these things are mixing, only because it’s just become a bigger part of our lives. And it’s a critical part of our lives, not just a nice to have. It’s a must have, I think, in most ways. So that just means more involvement between public policy and private companies.

And maybe that results in just straight up regulation, or maybe that just results in better partnership, or that being on the talking points of our next local election, is how we’re going to support or not support these types of initiatives. And I think that’s become – the lingua franca of digital infrastructure is like way more available than it was a couple of years ago, where even though I’ve been in this industry for 25 years, my mom didn’t know what a data center was, and now she knows all about data centers and chips, and it’s not from me.

Daniel Whitenack: I’m guessing that you all – you know, the fact that you’re supporting all of these use cases on top of Argilla on the data side makes me think… Like you say, there’s so many things to optimize in terms of that RAG process, but there’s also so many AI workflows that are being thought of, whether that be code generation or assistance, or content generation, information extraction… But then you kind of go beyond that. David, you mentioned text classification, and of course there’s image use cases… So I’m wondering, from you all, at this point – you know, one of the things Chris and I have talked about on the show a bit is… You know, we’re still big proponents and believe that in enterprises a lot of times there is a lot of mixing of rule-based systems, and more kind of traditional, I guess, if you want to think about it that way, machine learning, and smaller models… And then bringing in these larger gen AI models as kind of orchestrators, or inner query layer things… And that’s a story we’ve been kind of telling, but I think it’s interesting that we have both of you here in the sense that – like, you really, I’m sure there’s certain things that you don’t or can’t track about what you’re doing… But just even anecdotally, out of the users that you’re supporting on Argilla, what have you seen in terms of what is the mix between those using Argilla for maybe what people would consider traditional data science type of models, like text classification or image classification type of things, and these maybe newer workflows, like RAG and other things… How do you see that balance, and do you see people using both, or one or the other? Yeah, any insights there?

Jerod Santo: Right. And it’s out there in Release Candidate. It’s probably out there as we speak, at least in the official 3.13 experimental, as you guys pointed out. Are there early indicators? Are there people who have been just pining for this and they’ve already tried it in the RC, and are saying “Wow, we’re seeing this”? Or no?

Jerod Santo: Cool. Taylor’s pining for bonus points, but he’s not going to get a bonus point for that. We move now to round six. This is non-STEM, and it’s also a bit nonsense… This word is a tough one. Round six. Your word is Quomo Docunquize.

Chris Benson: Gotcha. As you’re mixing data from your own organization with some of this benchmark data, and trying to align that so that you have the benefit of trying to maybe fine-tune on a model and use benchmark data to drive that, but you’re also trying to introduce your own new capabilities based on data that your company has… How do you get those two sets of data to end up being a high-quality dataset without a lot of differences between it? As you’re trying to kind of get the best of both worlds; taking advantage of that which is already there, but bringing some new capabilities out that your company or your organization can leverage. How do you think about merging those two disparate sets of data so that you end up getting a good training set to do some fine-tuning with?

Jerod Santo: The only thing on my list that we haven’t talked about yet is mixing and matching. So oftentimes you’re picking - when we go back to databases, you’re picking a database. And I think that it’s common to believe that you have to just pick one and go with it… And there’s no rule in the rule book for programming that you just have to have a singular database. And I know lots of companies have multivariate, or whatever you call it, multiple data stores, depending on what they’re up to… Pretty common at least to have something a relational database and then also have something else, depending on what you’re up to. Oftentimes that is a key-value store, often used as a caching layer, but can be used for other things as well. Have you ever gone multi-database in projects of your own, or…? You’ve seen people do it, I’m sure.

Joseph Jacks: …and in order to upgrade the software, you would have to order, and they would have to ship you a new series of floppies, from which you would need to kind of like physically install, and you would own those physical like pieces of technology that would give you an upgrade. And then the pricing model was all built around that, too. It was perpetual licensing, where you would like basically buy that version, and then you would have maintenance that you’d pay for the version, and in order to upgrade a new version of the software you’d have to physically spend a new upfront amount for the new version, and then do that over and over again.

[34:18] Most of the – I would say probably 99% with some interpolation synthesis, mixing and matching of ideas… Most of the categories in software, including SaaS, which have become a little bit more like fractal domain-specific… Cal.com has almost like one feature, and wow, it’s a billion-dollar company. That just happens because of economies of scale, and humans using more and more technology. But most of these ideas were implemented in the seventies, eighties and nineties, even in the mainframe days. Scheduling, booking things… I mean, this is like calendar functionality.

I was recently watching the oral history of Mike Markkula, who was the first CEO of Apple. He was a guy that Sequoia founder, Don Valentine, introduced to Steve Wozniak and Steve Jobs to basically help them with the business plan, and like run the company. And Mike was also a really brilliant programmer, and he wrote Fortran and basic programs that would actually run on the early Apple computers. And the collection of those programs came with a credit, sort of a copyright credit at the bottom, which was Johnny Appleseed. You might remember this from the early Apple computers, this Johnny Appleseed name. So Johnny Appleseed was Mike Markkula. And Mike in the computer history museum describes this – it’s on YouTube, you can watch it. It’s a really great, multi-hour oral history. Mike actually personally wrote the programs for the applications that would ship in the early Apple computers. So including the calculator, calendar app, bookkeeping app… A lot of these basic – a very basic, primitive word processor…

And so I just think back to like the most fundamental, basic things that humans do as part of their work or their lives have been re-implemented and re-implemented so many times in the history of technology, going back to literally the fifties and sixties, that when we have new ways of distributing and innovating in the core technology, what happens is you just reinvent and you re-implement what happened, what came before, with shinier and newer, sexier, more sort of societally-relevant and interesting abstractions. And then those things end up taking off. And oftentimes people are given a choice. There’s sort of like the free version of the thing, maybe it’s open source and it’s like less sophisticated, but you can totally control it… Then there’s the proprietary version of the thing, which is much more polished and sexy and packaged. This is so fascinatingly – for the first time, and sort of skipping to the AI stuff, we’ve seen this incredibly shockingly surprising almost acceleration of equivalency. This actually happened this week - I mean, the latest LLaMA models are state-of-the-art. And by all accounts, the benchmarks are pretty much indistinguishable, and the capital required to produce that is similar. Mark Zuckerberg literally did an interview a couple of days ago saying they’re spending tens of billions of dollars building LLaMA, and they’re giving it away completely. And in fact, they’re making – Yann LeCun was just tweeting about this the other day… They’re making their licensing even more permissive. You can do distillation, quantization… You can use the outputs from LLaMA models to train other models… It’s almost like an Apache or MIT-level permissiveness. The artifact is very different for a neural network that has massive amount of information compression. Not in the Pied Piper sense, by the way, but in the neural network, AI, tokenization of –

Johnny Boursiquot: Exactly, exactly. Things of that nature. I’m lik, okay, so it sounds like the structs package is inching towards sort of more efficient, or some different way of lining up and doing some struct packing, I would imagine…

Michael Gat: I’ll send you a link. But I do want to conclude on one story, or on one little tidbit that I was just thinking about yesterday as I was looking at some of the latest stuff going on in S3. S3, of course, being Simple Storage Service. But if you use it, you know it’s not really very simple. In fact, using it correctly is extraordinarily complicated. And it occurred to me that “Well, nothing here has really changed.” Because you can use S3 very simply, but you won’t be using it optimally.

[48:19] Well, back in the day on a mainframe I had to write stuff to disk. And to do that, you had to know the physical details of the disk, and tell it how to allocate space. Do you want to allocate tracks, a single track around a single platter, or a cylinder, which is a stack of tracks in the same – you know, lining up with each other? Did you want to make them contiguous? Did you not? The more efficient you made it, the less likely you were to find space in the system. You had to know all this weird, arcane stuff.

When I tell people this today, they tell me “Well, that sounds really weird that you would need to know all that.” But I was thinking, “Yeah, well, 30 years from now people are gonna think it’s really weird that I had to think about how big my puts and gets were going to be.” And whether I wanted to do parallel puts and gets, or any other things to make it more efficient… But there’ll be some new level of bizarre, arcane stuff that people have to do to manage storage. So yeah, things changed, but nothing changes.

Daniel Whitenack: Yeah, it’s interesting that you bring up the element of sort of creativity and communication… So there’s probably some parallels, in the sense that I would say there are many people treating these sort of AI models and what they’re building with it as a very creative, new – I don’t know if you’d call it a new canvas on which they’re painting, but definitely they’re trying things that are new and interesting maybe, that haven’t been done before, and are very generative, and… But some of those things, even if you think about something very much on the creative side like the Udio type of thing, the music generator that we talked about a while back… I think you could make an argument “Well, that is an efficiency builder, because you could make a bunch of music really quick for your YouTube videos, or a bunch of music really quick for your ads”, or whatever you’re running online. But I think also some people are using it as a creative element in and of itself, and doing maybe new and different things, or mixing things in ways that people hadn’t done in the past…

And maybe there’s other examples that are better in my mind where it kind of – it’s almost a both/and type of situation. So I’m always maybe a sucker for the third option, where it’s not like clear-cut on one side or the other side… But this third option of “Yes, it is about efficiency gains”, but I think there is an element of net new things that will come out of the AI space with these models, that maybe are hard to predict right now, like they would have been hard to predict in the rise of the web… It was probably hard to predict what an Amazon would become…

Paul Copplestone: Well there’s this other piece that this particular product fits into. When I talked about most of those products, you would have noticed that they’re kind of standalone servers. And then the vector product, even though we launched it as a product, is actually an extension, and it ties together, like I talked about how to use [unintelligible 00:29:58.10] functions… You can use it with cron, pg_cron, you can use it with – we’ve got this thing called Database Webhooks, which is just another extension to trigger events out of your database to call the edge functions.

So these are all what I call or think of as the first-order primitives in terms of a platform. And Postgres is the substrate, then you’ve got these first-order primitives, and they’re the key products. But it turns out then if you combine some of these first-order primitives, you’ve got these second-order primitives. This one like vector is one of those second-order primitives. And they’re kind of jobs to be done by an engineer.

Another good example, which is quite popular on Postgres platforms, is queues. We could do queues inside Postgres, but also because we’ve got these background workers, the edge functions, they can be used too for the queues. Another really good one - we’re doing a lot of content around maps at the moment, PostGIS, and you want to store your… There’s a really awesome open source tool called Protomaps, where you can store all your tiles inside S3… And so because we’ve got storage, you can store them in your S3.

So you can see you can keep mixing and matching the primitives that we’ve called to release a suite of other kind of features or tools that developers need if they’re building. So we focus on what I think of as day one products at the start, and now we can kind of get these other maybe day 10, day 20 products.