Special: Local-First Conf 2026

1 - > AI is really shaking up everything. 2 - > I think putting software as a entire, like industry a little bit in a different 3 - > light now, and, I think this does not make local-first at all, less relevant. 4 - > I think this is, as for most things, a huge tailwind. 5 - > But given that, both of us have been really like, immersed in that new AI 6 - > space as well, now, I think we can come with a perspective where we can speak 7 - > as insiders from the local-first space, but also as insiders from the AI space.

8 - > And with that fresh perspective, I think we wanted to give this year's 9 - > Local-First Conf a new, code of paint. 10 - > Hey Adam, so great to have you back. 11 - > How are you doing? 12 - > Hey Johannes.

13 - > I'm good. 14 - > Looking forward to the arrival of Spring here. 15 - > So I was just joking that the two of us have been on the last 16 - > localfirst.fm episode since it hasn't really been on a regular cadence 17 - > over the last couple of months.

18 - > I'm taking a bit of a, pause from the podcast to reorient myself 19 - > in this new crazy world of AI and coding agents, et cetera. 20 - > And like, I wanted to get my own hands even more dirty again. 21 - > And yeah, little bit recalibrate, and I think that's also gonna be quite the theme 22 - > for, an announcement we're making today. 23 - > And to foreshadow a little bit.

24 - > What this year's Local-First Conference will be about. 25 - > So with that out of the way, yeah, curious what's on your mind, Adam? 26 - > Well, obviously as we ramp up to this Conference and we're, 27 - > we're doing something pretty different this year, I think so. 28 - > I don't know.

29 - > Maybe I can give some, some history for, I espect a lot of your listeners have either 30 - > been to the Conference or know about it. 31 - > but just for context, this will be our third year and we started in 2024. 32 - > we really weren't sure honestly, if there was enough demand for it. 33 - > You sort of had the idea of like, Hey, I think there's space for or 34 - > interest, you know, a latent interest in local-first and we could build a 35 - > conference around it on data ownership and CRDTs and all that goodness is 36 - > things that yeah, need a venue and a place to meet in person and share ideas.

37 - > And we were genuinely really concerned. 38 - > We just wouldn't sell enough tickets and we'd lose money on the 39 - > venue and all this kind of thing. 40 - > Instead, what happened was, I think our venue held like 150 or 180, something 41 - > like that, we sold it out within a week. 42 - > It was actually kind of embarrassing because we had to be turning wonderful 43 - > people away and plus ones for our speakers and things like that, just 44 - > 'cause we were at the venue capacity, so.

45 - > And for me it was also a big question about bringing together eclectic mix 46 - > of people, you had both academic people who had been working on CRDT theorems 47 - > and things for a decade and more kind of fringy, you know, Brett Victor, Alan Kay. 48 - > Tools for thought, perspectives with more practical software 49 - > engineers build React apps, just looking for the newest and greatest 50 - > technology to make their jobs easier. 51 - > Startups, there was, sync vendors were starting to become a little 52 - > bit of a thing then, and, you know, would these people even gel?

53 - > And I think the answer was yes. 54 - > The event was a huge success one day. 55 - > Single track, everyone loved it. 56 - > and it was a pretty obvious thing then the following year, 2025 to 57 - > expand it, we had a venue for 350.

58 - > We did two full days plus a community day, to more talks, more people. 59 - > Everything was 2 to 3x the scale. 60 - > And I think it really showed that we had this, community here and 61 - > that, that there was a lot of cohesion and people wanted to come. 62 - > Come back to it at the same time, I think, because sync and sync engines were 63 - > having a pretty big moment last year that ended up being a, at least one of the 64 - > day, was largely given over to that topic and we explored that pretty thoroughly.

65 - > So now we find ourselves thinking about, okay, one, we just enjoy 66 - > bringing this group of people together in this community of people together. 67 - > We want to get our friends back together in Berlin in the spring or the 68 - > summertime to just spend time together. 69 - > But at the same time, we also know the industry's changing and we also 70 - > know that it would be a little boring to just retread the same ground. 71 - > So that's causing us to want to be a little more creative with the 72 - > direction we're going this year, which is I think why we wanted to get 73 - > together and talk about that today.

74 - > Yeah, and I think if you're 75 - > just now looking back over the last two to three years, I feel that the 76 - > world is just such a different place. 77 - > When we look back then, local-first was really like, a wake up call 78 - > in this kinda like very almost. 79 - > not by today's standard, almost like static world, where things have 80 - > feel like seem pretty settled in terms of technologies, et cetera. 81 - > And we have like our best practices.

82 - > And local-first at that point was much more of sort of like a provocation, 83 - > hey, we can think differently about this and like rethink, the 84 - > typical like cloud model, et cetera. 85 - > And like it stirred up quite a bit of dust. 86 - > And I think this is what, what got people really interested about like 87 - > insiders who've been thinking about those lines for like many years. 88 - > Some even like, decades and have always built software this way.

89 - > And, then last year's conference I think has really shown a huge step to the year 90 - > before in terms of like how much further, how much mature the ecosystem has gotten. 91 - > But, I think it would be, like this year is different. 92 - > It's not just like, yes, the ecosystem has gotten a lot more mature and like 93 - > all of those technologies are more production ready, there's like more 94 - > players, et cetera, but the world around it is like, entirely different.

95 - > AI is really shaking up everything. 96 - > I think putting software as a entire, like industry a little bit in a different 97 - > light now, and, I think this does not make local-first at all, less relevant. 98 - > I think this is, as for most things, a huge tailwind. 99 - > But given that, both of us have been really like, immersed in that new AI 100 - > space as well, now, I think we can come with a perspective where we can speak 101 - > as insiders from the local-first space, but also as insiders from the AI space.

102 - > And with that fresh perspective, I think we wanted to give this year's 103 - > Local-First Conf a new, code of paint. 104 - > this is where, you've been this year leading the effort around the CFP. 105 - > And yeah, so I'm excited to, talk through the different themes we're expecting 106 - > for the conference where we have a set of curated speakers already, but 107 - > we think, the best content is really brought up from the community, who is 108 - > building things in the local-first space.

109 - > And this is, like the people who are maybe listening to this podcast, or who 110 - > have attended, the last few conferences. 111 - > So yeah, this is, what I would love to talk to you about. 112 - > Yeah, well, maybe we can just start with that call to action really upfront, which 113 - > is we're broadening the umbrella for what kind of talks we want a lot this year. 114 - > And so if you do nothing else, you can just stop listening to the podcast.

115 - > Go read the CFP page. 116 - > If you think there's any chance of what you're working on, which has something 117 - > to do with empowering users or making, computers and software and the internet 118 - > freer and more capable for enhancing human life, then you should submit 119 - > a talk that's your CTA right there. 120 - > To drill, one level deeper. 121 - > you already mentioned the AI topic and obviously that feeds into this.

122 - > There's also the desire to continue to build and expand on the themes of this 123 - > community and respond to other things that are happening in the tech industry. 124 - > On the AI side, obviously, where I believe, you know, creation of software 125 - > is changing fundamentally with AI assisted coding, and I think people in our 126 - > community have a very wide mix of views. 127 - > There's people who come from a. very skeptical point of view, I probably 128 - > lean more in that direction myself, especially up until recently, there's 129 - > some bubble, it's gonna burst soon, it's overhyped, it's, you know, whatever.

130 - > And on the other end, there's probably plenty of maximalist, people that, 131 - > you know, I think it's not even such a bold thing now to say that 2026 might 132 - > be the last year that humans do all that much direct coding in terms of 133 - > like writing out programming syntax. 134 - > and I suspect people listening fall in a wide range of, views there, and 135 - > we wanna make room for all of that. 136 - > But, you know, at the same time, we're acknowledging that this industry shift is 137 - > happening while also staying grounded in our values and what we're all here for.

138 - > For the record, October, 2025 was the last time I wrote a 139 - > single line of code by myself. 140 - > So, but I'm leaning into that, quite heavily. 141 - > But I think as you say that's a wide spectrum. 142 - > And nonetheless, I think that makes, even more clear that we need really 143 - > good principles for the software we're creating, and that we also want to use, I 144 - > think, software through AI, like through the terms like slop software, et cetera, 145 - > I think has made it even more, even more important to really care about like what 146 - > we're using, not just for the perspective of that "it's nice to use", but also for 147 - > the perspective of like, data ownership, privacy, security, all of those.

148 - > And I think local-first is still the guiding light for us. 149 - > It's kinda like the gold standard for what software should feel like, what 150 - > software should be architected around. 151 - > And I think this is where we hopefully get the best of both worlds, where we 152 - > can have the cake and eat it too in that regard that we can dream about 153 - > software that follows those principles. 154 - > But now it's a lot more feasible to build those and to build those, 155 - > not just for the sake of like building software for potentially 156 - > like a million target users, but for yourself, like personal scale software.

157 - > And this is where we had this amazing closing keynote from Maggie 158 - > Appleton at the first Local-First Conference about Homecooked software. 159 - > And I think in a way this year, 160 - > barefoot developers, I think was the term she, she used. 161 - > That's right. 162 - > And certainly what you could do with language model assisted coding 163 - > back then was, you know, much.

164 - > more limited than what you can do today. 165 - > But even then, she predicted that, which I think is a good, a good illustration 166 - > of the kinds of people that are in our community that, you know, were 167 - > values driven, but very forward facing. 168 - > She, she saw what was coming years before. 169 - > Exactly.

170 - > And like now, this year's conference is basically like, really like double 171 - > clicking on her entire vision here where like everyone can dream about 172 - > software and like now make it a reality. 173 - > And local-first doesn't just make it, more secure and private, but like overall 174 - > also like solves a whole bunch of other problems that is still hard with software. 175 - > So I'm really excited how like those two or the, how the conference 176 - > can be a place where those two different worlds can come together.

177 - > All of like the relentless, accelerated AI community, who wants to, build things 178 - > really quickly, but also the people who like really care about software 179 - > quality and like those principles. 180 - > And this is something where I think it can be, a really nice meeting space for folks. 181 - > So, Adam, how about you wanna give us a broad overview of the different 182 - > kinda like, themes and then we walk through them one at a time? 183 - > Yeah, sure.

184 - > So the overall theme for the conference this year is user empowerment. 185 - > that comes from, actually a panel we had last year. 186 - > I host a panel and one of the, panelists was Martin Kleppmann, who 187 - > this year is a keynote speaker for us. 188 - > Obviously he's really one of the, founding fathers of, local-first.

189 - > But the way he described it was, we were talking here a little bit about the kind 190 - > of the specific implementation details of sync engines and CRDTs versus the 191 - > broader set of values and principles. 192 - > And I think he put it really well in saying. 193 - > It's all about user empowerment. 194 - > If something gives users more empowerment, more control, more capabilities, more 195 - > options, more freedom, then that to him falls under the umbrella of local-first.

196 - > And if it's something that sort of takes agency away, which there are times 197 - > where that makes sense, especially in consumer software or whatever, that 198 - > people want less control over how they're using their computers, that's fine. 199 - > But that, leaves it outside the realm of local-first. 200 - > So we're borrowing that idea and expanding on it. 201 - > User empowerment, greater agency, obviously a lot of that tends 202 - > to flow through data ownership.

203 - > 'cause the data ultimately when you're creating documents on a 204 - > computer, the data and feeling like it belongs to you is, tends to be 205 - > core to that feeling of empowerment. 206 - > So with that overall theme, took together with our, collaborator, Eileen Wagner, 207 - > also a long time member of the community who is, helping us here with the CFP. 208 - > But she basically sorted us into three categories here. 209 - > We've got new territories, which of course includes plenty of AI stuff.

210 - > We've got local-first maturity, which is trying to see, you 211 - > know, now we're pretty far along. 212 - > There's a lot of great sync engine vendors. 213 - > There's some real world success stories, but there's also 214 - > interesting new challenges to face user experience challenges, legal 215 - > challenges, or legal frameworks. 216 - > And then where we're really trying to expand here is the larger ecosystem.

217 - > And there's a lot of exciting stuff happening in, whether it's in something 218 - > like the area of game developments, but also something like social media. 219 - > I think, you know, with ATproto and to a lesser extent sort of the masked 220 - > on ActivityPub world of things is having their own moment and kind 221 - > of causing people to ask questions about data ownership and freedom and 222 - > decentralization in this realm that traditionally has been a very kind of 223 - > centralized, closed, big tech, you know, inscrutable algorithm world of things.

224 - > And to what extent are data ownership and local-first values applicable there. 225 - > So those three frames, the kind of new territories, the maturity of local-first 226 - > and the larger ecosystem are where we have kind of a few subtopics for each 227 - > one, and we're trying to hope, hoping to get a good coverage on, CFP submissions. 228 - > Awesome. 229 - > Thanks so much for providing this great overview.

230 - > I would actually love to double click on each of those and dig a 231 - > little bit more in, since I think each topic has, so much potential. 232 - > And I think we could, like, if we'd had a lot more time and space for the 233 - > conference, we could almost like make each of those their explicit like theme day. 234 - > And so now we only have, well, those two days for the conference. 235 - > for the talks in one extra day.

236 - > So we need to curate and pick. 237 - > But I think this gives us like a really nice broad buffet of like different, 238 - > topics and talks to choose from. 239 - > And ideally, those, themes speak to people who want to submit a talk. 240 - > So yeah, let's go through them.

241 - > So maybe starting with the, local-first maturity, maybe that is, where we, can 242 - > reconnect the most from where we've left off the first two years of the conference 243 - > and also Sync Conf and, I think this is like local-first maturity is really like 244 - > that that is the essence of what last year's conferences, were about, where we 245 - > had local-first success stories, companies like Linear, et cetera come to mind. 246 - > The UX challenges of local-first.

247 - > This is, you've mentioned Eileen before. 248 - > She's given an amazing talk about that last year legal framework. 249 - > Maybe a little bit more top of mind this year around, let's start with 250 - > those like local-first success stories. 251 - > Anything that comes to mind for you in particular.

252 - > Well, for sure the success stories should be, there's obviously people building 253 - > libraries, you know, the sync engine vendors who are well funded and have a 254 - > lot of customers and that sort of thing. 255 - > But what I really wanna hear about is the end user applications. 256 - > You mentioned Linear, Linear CTO Tuomas has been kind enough to give us 257 - > excellent talks the last two years but I also think of on a smaller scale, 258 - > there's something like, we had a talk last year that was about someone 259 - > building a forestry application.

260 - > So this is where someone walking around with a phone that needs to basically 261 - > tag each tree in the forest with what's going on with it, whether it has a bug 262 - > infestation, whether it needs whatever else and it's a pretty simple thing. 263 - > And you know, I like that talk a lot because also they basically said, 264 - > look, our data modeling is very simple. 265 - > They built their own sync that needs to, of course, not only work 266 - > offline, but for a big chunk of time.

267 - > But it's a pretty simple queue. 268 - > And unlike full text editing, that is a rich text editing that's very 269 - > complicated you actually can accomplish all the goals of local-first with not a 270 - > huge amount of engineering for some of these simpler, use cases and domains. 271 - > So I think there's plenty of people in the community that have their 272 - > hobby projects or whatever that they use together with their friends. 273 - > But what I like to see we're far enough along now is what is at scale.

274 - > I wanna see a Linear and Obsidian or some of these smaller scale things 275 - > like the forestry app that have been doing this for a long time. 276 - > You have real world data and you can talk about the challenges of debugging. 277 - > You can talk about the challenges of accumulating data history. 278 - > You can talk about what are the ugly and difficult parts of this 279 - > that we're grappling with from using this frontier technology.

280 - > Yeah, definitely. 281 - > And I think there's, just thinking through like the software and new 282 - > software that I've been using over the last like year or so, I've been 283 - > using like Rayon for example, quite a bit, which is like a Figma like, 284 - > software, but for more like architecture use cases as I'm like remodeling 285 - > the house I grew up in and that's been like really nice to get sort of the 286 - > benefits that I'm like spoiled with of like using production software 287 - > like Figma now for more use cases.

288 - > And I think that's just like one example for like, that seeping into more and 289 - > more parts of the software we're using. 290 - > And I think that will just, continue and like, particularly 291 - > now where, AI is another driver of like personalized software. 292 - > I think it's just like very clear that we want those benefits of 293 - > like collaboration, that things should work offline, et cetera. 294 - > Like if anything, those problems are now even more pressing, where when 295 - > we have software that is like meant for individuals, and to have like 296 - > that software, be shared, across like family members or, friends.

297 - > This is where you don't like, you, probably don't build it out as like 298 - > a SaaS architecture from the get go, but you're building this like 299 - > in a different way to begin with. 300 - > SQLite certainties having a bit of a moment again as well. 301 - > So I think we're gonna see a lot more of that. 302 - > Files are, are, are are the hot new thing again.

303 - > Unix 304 - > tools. 305 - > Yeah. 306 - > Honestly. 307 - > Honestly, files could have almost been like another category, in itself.

308 - > Like it should really also, and not to sound like a broken record, 309 - > but like, yeah, with the AI hat on, files and folders are like the best 310 - > foundation for agents to do good work. 311 - > And it has, intentionally or not really, like, become, again, a main way of like 312 - > how data collaboration works, et cetera. 313 - > And so, yeah, I think there's a lot to be explored there. 314 - > A lot of creative people doing weird, interesting things.

315 - > So if you're, if you're like that, please submit a talk and, yeah, I 316 - > think there's like whatever, you've been working on in regards to like 317 - > traditional local-first in the sense of like the previous conferences, we would 318 - > love to hear from you, but also around the UX challenges, of, local-first. 319 - > I think we're still just scratching the surface of like trying to 320 - > understand that, trying to like deal with the implications of that. 321 - > I think we're allowing ourselves to dream a little bit bigger or a little 322 - > bit more nichey with software now.

323 - > And now that might require like new user experience paradigms and like, 324 - > particularly also when it comes to extending collaboration just beyond 325 - > people, but also like collaborating with like other agents, is another 326 - > aspect that I think squarely fits into the realm of local-first as it's 327 - > another like kind of trust boundary. 328 - > You wanna understand what has happened. 329 - > Do you want to collaborate? 330 - > Do you wanna like detach, be by yourself?

331 - > Do you want to like revoke authentication, authorization, et cetera. 332 - > Maybe real time collaborate. 333 - > I think there's a lot here, whether you want to look at us for the AI lens or 334 - > not, but I think that area is like super underexplored and if you've been exploring 335 - > that, we'd love to hear from you. 336 - > Yeah.

337 - > The UX challenges for me as part of the. 338 - > I don't wanna say quite say fun part, but sort of it is because it's a chance 339 - > to rethink those computing primitives. 340 - > Right? 341 - > And the files example is a good one to me, where files and the Unix 342 - > methodology and everything as a tool and pipes and redirects was really kind 343 - > of a first love for me in many ways.

344 - > In computing, especially creative com, you know, computing that's 345 - > designed for creating things. 346 - > And then, you know, mobile basically abstracted files away. 347 - > And that's kind of a win for most of the consumer use cases. 348 - > but actually is kind of disempowering for creative work that you wanna do.

349 - > And now again, the files are, somehow having a moment again. 350 - > But now we live in this world of things like real-time collaboration, version 351 - > control pull requests, yeah, live collaboration, async collaboration, and 352 - > thinking about how all those people pieces fit together or how you can use those 353 - > building blocks for the particular piece of software you happen to be building. 354 - > That is very interesting to me. 355 - > So another one is legal frameworks.

356 - > So we are working with, like, we're building those systems, maybe just 357 - > for us, maybe for others, but, like we're not building this in a vacuum. 358 - > Sure, I can build some software that might just run like locally 359 - > on my, phone or on my computer. 360 - > Or like on like a little shared server in my home. 361 - > But in that regard, no one cares.

362 - > But like once I put it a little bit more out there for the world to see and to use, 363 - > or possibly that it's like still meant to be used for a smaller group, but it's 364 - > still accessible for others, possibly. 365 - > So you need to think about the implications. 366 - > You need to possibly lock it down. 367 - > You need to constrain it, you need to set expectations.

368 - > And one like way of setting expectations or being constrained is through like 369 - > our legal system, whatever legislation we're falling into, like Germany, where 370 - > we're doing this conference in like. 371 - > Very particular set of legal constraints, also like European constraints, et cetera. 372 - > But we're not building this in a vacuum. 373 - > We need to like fit into the constraints, realities of the real world.

374 - > One big question is like, where does data live? 375 - > But also of like, other questions? 376 - > Well, like what sort of rights do I as a user of software have? 377 - > Like, GDPR is certainly something that really changed, the ecosystem 378 - > quite a bit for better or for worse?

379 - > I think actually one, funny aspect of GDPR, like maybe one intended or 380 - > unintended consequence that I'm quite excited about is that GDPR has had a 381 - > massive impact on data compatibility. 382 - > Where you can now go to like pretty much every service and 383 - > through the right to be forgotten. 384 - > Like you can still export your data, but that exported data now actually 385 - > gives you all the data for that service. 386 - > So you can now like go to LinkedIn.

387 - > It's like, hey, like give me all my data or like, go to x.com, give me all 388 - > my data, and like now you can export that and like build tools with it. 389 - > So I think that's like one interesting ways of like how a legal framework 390 - > has unlocked data compatibility, even though that like, and not in 391 - > a very rigorous way, but at least something better than nothing. 392 - > And so that just like as one anecdote of like.

393 - > How something has changed and unlocked new opportunities. 394 - > And I think there will be a lot more like that. 395 - > Like good developments, bad developments. 396 - > Another real world analogy might be if you're going to Linear and you 397 - > create a team, you need to set like, Hey, where should the data be stored?

398 - > Do you want to have it stored in the US? 399 - > You don't wanna have it stored in EU. 400 - > Like, that's another way to like really model that. 401 - > And I think that's just like, scratching the surface of how 402 - > we need to bake those realities into the software we're building.

403 - > And, I think there, this is an area that that's really worthwhile thinking through 404 - > and be hopefully inspired by, new ideas. 405 - > Yeah. 406 - > I'm weirdly interested in the legal frameworks area and I hope we get 407 - > some good submissions on that. 408 - > I think it is natural for people in the tech world to dismiss or 409 - > downplay that aspect of things.

410 - > Especially maybe if you come from the Silicon Valley, American 411 - > Maverick laws just get in my way. 412 - > I'm gonna ignore them perspective. 413 - > but yeah, we're, we're both here in Europe. 414 - > I, two, two things that come to mind for me on that, that are 415 - > interesting and of this moment.

416 - > One is, yeah, you mentioned GDPR and much maligned in some ways. 417 - > You know, the cookie banner thing is an unfortunate consequence of 418 - > how that all played out, but it also has provided real front guidelines 419 - > for handling of person identifying information that I think genuinely has 420 - > made the internet a safer place, right? 421 - > It's full of threat and attacks and that's only getting worse 422 - > now that, that stuff can all be superpowered with language models.

423 - > But, you know, when I was working on an AI powered browser called Dia 424 - > last year, it was actually genuinely helpful to us to basically go and ask 425 - > the question, what is GDPR recommended? 426 - > And it basically comes down to, okay, we wanna be able to train on user data 427 - > that's incredibly valuable to our users in the same way that, you know, Google's 428 - > magic results and auto complete works so well because they're training it on 429 - > what other people are searching for.

430 - > And that's how it can seem to kind of read your mind. 431 - > That's the same thing is true with, Cursor and the auto complete. 432 - > So there's huge value in training on user data, even though there's, you know, but, 433 - > but there's obviously some, things you need to be really cautious about there. 434 - > And GDPR gives actually really good guidance.

435 - > It basically says if you strip all information that can tie 436 - > it back to a particular user. 437 - > Certainly anything personally identifiable, like a name or an email, 438 - > but also you're decoupling it from the user ID at the moment of collection 439 - > and anonymizing there and you throw it away within 30 days, you're all good. 440 - > That's a very simple framework and actually we could do a lot with that. 441 - > We could really, go pretty far with training our own, 442 - > models with data in that frame.

443 - > And it was actually nice to have that, not just from a legal 444 - > perspective, but from just a best practices guidance perspective. 445 - > so GDPR has been a mixed bag in some ways, but has also I 446 - > think provided genuine value. 447 - > The other thing I would throw out that is also probably kind of 448 - > from the Europe perspective that's top of mind now is sovereignty. 449 - > Basically data sovereignty and software sovereignty.

450 - > And I think a lot of the European Union as a whole, but also individual countries 451 - > are thinking about how to decouple themselves a little bit from US providers. 452 - > And I think we can put aside, or we don't wanna make a political conference 453 - > or be too tied to whatever current event happens to be in the news at the moment. 454 - > But I think in general, this is a good perspective for nations to have 455 - > and it is just a scaled up version of local-first data ownership.

456 - > It is just a version of the same reason I say it's good for me to have control 457 - > over my data and have the ability to make choices about it, not only in 458 - > the short term, but in the long term. 459 - > Where do I back it up? 460 - > Can I delete it, can I duplicate it, et cetera. 461 - > There's a version of that for nations.

462 - > And I think that's a little bit what, Europe is grappling with right now. 463 - > And I think totally separate from anything happened in the current moment. 464 - > That's a good thing to think about. 465 - > That's a long-term thing that helps create greater agency and autonomy 466 - > and security for your nation, just as it does for an individual.

467 - > And I think there's also really like a window of opportunity, around that, where 468 - > like, whether, what do you wanna call, like, big players like Monopoly or not? 469 - > that aside, I think there are now real cases of precedent where I think certain 470 - > German states, for example, have made the switch away from Microsoft to like 471 - > some, open source software, et cetera. 472 - > And like, I think that sets a, like me as someone who has like, very good 473 - > feelings towards open source and wants to see open source software succeed, 474 - > I think there's new momentum here.

475 - > And I think local-first can really also through that perspective, 476 - > play a more mainstream role. 477 - > Since like you don't just, yes, you can have like your open source react 478 - > components, but that does make like fully functional software where you 479 - > can replace other software with it. 480 - > Yeah. 481 - > Like you need to think about data and if you're making that painful 482 - > switch from one software stack, that might even like include all the way 483 - > to the operating system, you probably like experience a lot of pain there.

484 - > And ideally, you now choose a foundation that gives you more 485 - > optionality in the future when you might want to do another switch. 486 - > And this is where things like, data compatibility, et cetera, 487 - > all like play into this. 488 - > And where ideally, like legal frameworks set a good foundation and like encourage 489 - > things like data compatibility, et cetera, but in a thoughtful way where it does 490 - > not, introduce so much overhead and like bureaucracy that it grinds down like the 491 - > velocity of like hurricane ship software.

492 - > And I think this is like the best of all worlds that 493 - > local-first wants to empower that. 494 - > Where if you build with like those local-first data layers. 495 - > The data layer takes care of that for you, and you can focus on the software. 496 - > So I think this is all like a very rich area, to talk about and I think that 497 - > what's makes it really interesting is to hear it through the lens of like one 498 - > particular story, whether it is like a company having made that switch, whether 499 - > you've built software and like you've made your software more compatible and 500 - > interoperable in some way, or, yeah, there's like all of those stories that 501 - > I've heard over the years and I think those deserve a broader, audience.

502 - > And so if you have something to share in that regard, please definitely take 503 - > the step and like, submit it to the CFP. 504 - > That's a great point. 505 - > It almost makes me wanna make data portability or something to that regard. 506 - > a top level topic.

507 - > But I think this comes back to the principle in local-first, 508 - > which we call the Long Now. 509 - > Which is, yeah, once you've been around for a while, you 510 - > can see the gray in my beard. 511 - > I've been working with computers for many decades now, and I've 512 - > seen many products come and go. 513 - > Software platforms come and go.

514 - > But being able to preserve that data for a variety of reasons. 515 - > 'cause it's important for your company. 516 - > 'cause it's important for you personally. 517 - > There's emotional meaning and saving your old family photos or the, master thesis 518 - > he wrote in university or whatever.

519 - > And then of course it is natural that especially the startup industry, 520 - > which skews young and people are building products and moving fast 521 - > and thinking about tomorrow and next quarter and not 10 years from now. 522 - > That's, that's fine. 523 - > and well and good, but maybe that is a role. 524 - > Governments can play a little bit, but it's also a role that, yeah, maybe our 525 - > community can play, which is encourage people to think a little bit about that 526 - > long term, both as users, what's going to make my data be a little safer, give me 527 - > that long term optionality, as you said.

528 - > but also product makers to think in terms of like, okay, may maybe 529 - > my product won't be hero, will have changed a lot in five or 10 years. 530 - > Am I empowering my users, to take their data and go to where they need 531 - > to go when the time comes for that? 532 - > And I think it's just like a wonderful way to think about the world a bit more, like 533 - > positive sum, where I think we've, like in some areas of software, we've already 534 - > figured us out, like in a pretty good way when we have like languages and like 535 - > ideally like typed languages, et cetera.

536 - > Like you have an API that you can compatible with and then 537 - > you can like reuse modules. 538 - > We even got that all the way to like UI, et cetera. 539 - > But for data, this is where we're struggling still the most. 540 - > And I think this is like one of the core themes around local-first where 541 - > there has been tremendous progress to, to make, progress in that regard.

542 - > And in that theme of like user agency, et cetera, personal software, 543 - > I think this will play an even bigger role, where we are gonna 544 - > have like more software than ever. 545 - > Whether it's like we've built this ourselves, whether it's agents 546 - > have built it, and then those things can, can talk to each other. 547 - > But I think it's also like changing in an interesting way where the 548 - > contract between that does not have to be like absolutely perfect anymore.

549 - > But like now AI can also like, shoulder a bit of that burden of like talking 550 - > to each other, where now we can like get away with having fuzzy software. 551 - > Where it has like CSV over here and like a JSON API over there, and the 552 - > agent can like very easily either ad hoc one-off, like convert something or 553 - > create a script that talks to each other. 554 - > So I think we're like, we're getting closer and closer having like some 555 - > bridges and ideally over time we'll get to like really having very reliable 556 - > bridges that just, in our minds when we hear a story like over there and 557 - > like we write it down over there, like we are a bridge and just works.

558 - > And ideally software can be like that in the future. 559 - > I like that a lot and I like that point that software is becoming 560 - > cheaper to make, so therefore it will be more ephemeral, more throwaway. 561 - > You'll be more likely to make something that's purpose-built for 562 - > a, a shorter period of time, whether it's an internal tool at a company 563 - > or a personal app or something. 564 - > Something like that.

565 - > And in the past we've really relied, or to date you could say we've 566 - > really relied on, my Gmail is still readable 20 years ago because Gmail 567 - > still exists and is well maintained. 568 - > but instead, maybe if we have a system that is more based on, well, I can 569 - > use an email tool that's a little bit bespoke and gets stops being relevant 570 - > or stops being maintained two years from now, and it's no big deal because 571 - > getting my email out or whatever data came from that tool and into the next 572 - > thing that I'm going to use is pretty straightforward to do from through some 573 - > combination of export data portability.

574 - > But also, as you said, that importing and ingesting data and making data 575 - > coherent with the help of AI is easier. 576 - > So I like that idea of a lot. 577 - > This, the software is kind of ephemeral and fits the need you have in the exact 578 - > moment, and then you can move on from it. 579 - > The data is something that has a longer lifespan.

580 - > It can move with you from, product to product. 581 - > Exactly. 582 - > Great. 583 - > So this was local-first maturity, and we've already teased a little 584 - > bit of like some adjacent topics, but the, the other, big pillars here, 585 - > as Elene put it, is like the new territories and the larger ecosystem.

586 - > So maybe if we spend a bit of time on the new territories where we have like, 587 - > designing for agency, then around like AI, local and open weight models, and 588 - > then also AI assistance for local-first. 589 - > Maybe starting with the first one, design for agency. 590 - > What should people think about here? 591 - > Yeah.

592 - > For me, this would come to the topic of, I like computers as a tool for thought, 593 - > a way to augment, thinking, a way to make humans more creative, more capable, 594 - > able to accomplish our goals better. 595 - > I sometimes think, especially as consumer products have have come to 596 - > dominate, for example, just sort of usage and revenue in the computing world. 597 - > There ends up being this weird race to the bottom of, hey, 598 - > the computer will think for me.

599 - > that's what we like about AI. 600 - > Good. 601 - > I don't need to think anymore. 602 - > The computer can do it for me.

603 - > And that to me is a very sad perspective. 604 - > And rather I would think, wanna think that, AI or other capabilities 605 - > that we continue to develop as, we continue to explore the computing 606 - > space are things that help us do more and better the things we enjoy 607 - > doing, being creative and thinking. 608 - > and so I think the design for agency topic is the idea of, okay. 609 - > So if you're doing something with language models, with computer vision, 610 - > with agents, whatever it is, what can be part of your design, in the sense of how 611 - > your product works, that really engages and creates more agency and more deeper 612 - > understanding from your users, right?

613 - > The negative scenario or the people often quote is that Star Trek 614 - > next generation episode, right? 615 - > Where people have this, these omniscient computers that do everything and 616 - > then they lose all the capability to understand not only just how to 617 - > fix the machines, but just really how the world around them works. 618 - > And then when those computers start to fail, they're suddenly 619 - > like helpless children. 620 - > And I think for me, good computing tools of all sorts, and that includes everything 621 - > to do with AI and language models deepens our understanding of the world, deepens 622 - > our understanding of the technology by freeing us from some of the fussy details 623 - > of where exactly does the semicolon go, and allowing me to spend more time on the 624 - > mental models and the frameworks and the principles and how things fit together and 625 - > how we use those to accomplish our goals.

626 - > Yeah. 627 - > And I think this is, I mean, for, for me, AI has been like one of the biggest 628 - > unlocks to like just learn new things. 629 - > I've gotten into so many new areas, like I've gotten deeper on like 630 - > some hardware related things. 631 - > I've gotten deeper on understanding how wifi works, 632 - > like band channels, et et cetera.

633 - > Like gave me a deeper and a broader understanding of the world, like 634 - > I have been on an accelerated pace of, learning and understanding. 635 - > and I think that's like the positive sum, version of that or like the 636 - > flip side or the other, scenario, of an like the utopia maybe of that 637 - > Star Trek, anecdote you shared. 638 - > So, one other much more concrete topic that I think has been, a very 639 - > new one that's come, about over the last few years as, like off the 640 - > shelf models became available at all.

641 - > And now we also have like, open weight models and they're even small enough 642 - > that you can run them on your own device, whether it's on a very beefy, desktop 643 - > grade machine or whether it's like, scaled down models that might even work in the 644 - > browser, that might even work on like a phone or other much lower end devices. 645 - > I think that opens a whole new category of software and use cases, et cetera, whether 646 - > it is, like that you built a little system that relies on like detecting 647 - > certain shapes to do a certain thing.

648 - > Maybe you're building, like maybe you have chicken at home and you wanna like 649 - > automate a little gate that like, opens when a chicken walks through or whether 650 - > you have like some other use cases. 651 - > But this is you ideally wanna do that like reliably, locally, efficiently, et cetera. 652 - > And there's so many, like we can, we can just like dream so much bigger now 653 - > of like what we can build in a way that does not rely on like Anthropic currently 654 - > not having an outage and us like paying whatever, a lot of money each time.

655 - > We are like using a bit of like API quota and, I think this is 656 - > gonna be like right now, this is. 657 - > Probably similar to the early days of like computers being room 658 - > sized, and not very capable yet. 659 - > And I think that's just gonna, like at some point our little like devices like 660 - > this, will be able to do like really, really impressive, inference capabilities. 661 - > And I think this is like, now is the time to start thinking about 662 - > that and start already embracing that in all sorts of different 663 - > scenarios, form factors, et cetera.

664 - > So I think this area could be about like if you're building 665 - > infrastructure, building tool, like maybe you're the company behind some 666 - > of those open models and you wanna show people what's already possible. 667 - > I think that's really great. 668 - > if you are building applications, systems, et cetera, that already 669 - > use, open models, local models. 670 - > that is like a story we'd love to hear.

671 - > We had last year, Thomas from Google already share, some earlier versions 672 - > of like how you can run, Gemini Nano in Chrome and showed some really cool demos. 673 - > And all of this has come a really long way and, I think across all sorts of 674 - > different devices, platforms, et cetera. 675 - > So if you're doing cool things in that regard, please definitely apply to speak. 676 - > Yeah, I think so much attention is on the frontier models and the 677 - > sort of one giant chat bot that can do everything, multimodal models, 678 - > ChatGPT, and Claude and so forth.

679 - > And those are of course impressive and very useful and there's a lot to that. 680 - > But I really have, opened up to the value of things like 681 - > lightweight classifier models. 682 - > Yeah, embeddings, including, there's plenty of text embedding, models that 683 - > you can just run in a Python server with PyTorch that are great, really 684 - > almost as good as basically as good as what you can get in the big providers. 685 - > Things like computer vision and OCR have actually been good for a pretty long time.

686 - > I think in some ways, you know, maybe myself, I almost feel a little, 687 - > surprised, you know, something about the AI hype, you know, around chatbots 688 - > and large language models then caused me to take a closer look at a lot of 689 - > these other ML technologies that have been getting good for a long time. 690 - > A lot of the text embedding stuff and semantic search and things like 691 - > that have been, you know, really had their breakthroughs 10 years ago.

692 - > Computer vision has also been very good for a long time. 693 - > We know that because captures keep getting harder and harder. 694 - > And you can do a lot with these, in those very specialized cases, like what you 695 - > were describing with, yeah, whether it's a robot in your house or, something like 696 - > that, but also software where you can have the model run on each keystroke, right? 697 - > You can do all kinds of things.

698 - > They're very different versus needing to wait several seconds for 699 - > a heavy inference to come back from the big language model provider. 700 - > And just broadly, the story of computing is one of taking these 701 - > computing things that are originally are heavy and slow and as you said, 702 - > require a room size computer and making them smaller and cheaper and faster. 703 - > And that just opens up all kinds of new use cases. 704 - > And there's really great open models out there from Quinn 705 - > to Llama to many, many others.

706 - > And yeah, I think if you're working either on open weight models, or open 707 - > models of any kind or the libraries behind it, like Transformer.js or 708 - > PyTorch or something like that, we'd love to hear a talk from you. 709 - > Awesome. 710 - > So the next topic in the new territories, area is AI assistance for local-first.

711 - > This is, I guess more around like the application of those new materials 712 - > to those new building blocks we got. 713 - > And I think that can be really, like wide ranging, can be going in 714 - > all sorts of different directions. 715 - > Like, and I think the exciting part is actually the stuff that we can't 716 - > even, like, think about right now, but that maybe some of you are, 717 - > actively exploring and have built like surprising things on top of this. 718 - > But just some things that came, to mind for me as I've been like, 719 - > experimenting myself or thinking this through, like just to, generate ideas.

720 - > How can AI address like some of those open problems in the local-first 721 - > space or like some challenges that are so like kind of hard to crack 722 - > and if we're allowing ourselves. 723 - > To go beyond the realm of like deterministic, traditional software. 724 - > And like, we're using those fuzzy materials a bit more. 725 - > And like we're thinking of like, hey, if we're actually, embedding 726 - > them into our local software stack, what becomes possible through this?

727 - > So like, just one way, for example is let's say you're building a power tool and 728 - > like you have a very, very sophisticated set of like system settings, et cetera. 729 - > Like for me, for example, the Apple system settings come to mind and 730 - > they're really like, I don't think they're the pinnacle of modern software. 731 - > And, I always need to use search. 732 - > And like five different things come up and then it's like the 15th one 733 - > that I was actually looking for.

734 - > And like, it's just like very painful. 735 - > Now where we are so spoiled by like using things like ChatGPT et cetera, 736 - > like it really feels like worlds a part of like how good something could be. 737 - > So why not use something like a, ChatGPT chat interface to interact with your 738 - > own app that like, you can, instead of have to click the clock settings, 739 - > like why not allow your entire app to begin configurable through AI models. 740 - > And so to make that actually happen now you need to bring in like that 741 - > fuzzy LM piece into your software.

742 - > But like exploring that, like figuring out best practices, figuring out 743 - > like how can you actually, constrain that, that you don't, give the, users 744 - > so much, power that they can like accidentally tear apart the app and 745 - > it's no longer functional, et cetera. 746 - > I think there's a lot to be explored there. 747 - > So just as one area. 748 - > Another one is to give even more power to the user or build systems in a way 749 - > that wasn't really feasible before.

750 - > We talked about the data interoperability before data exports, et cetera. 751 - > Now we can actually build like fuzzy bridges between different 752 - > software that were never meant to be talking to each other. 753 - > Whether it's like in real time or whether it's just like making 754 - > sort of, export interfaces a little bit more ergonomic to use. 755 - > And now you could imagine whatever, like using your, local supermarket 756 - > that might have like a certain, online shopping, like not even API, 757 - > but just like a maybe connecting that with your own home build to-do app.

758 - > And, there is not really like the end all be all integration API, yet 759 - > you can make those things happen now. 760 - > So this is just like one of many, many ideas where hopefully you all have 761 - > been exploring some interesting use cases that we'd love to hear about. 762 - > So this is what comes to mind for me about AI assistance for local-first. 763 - > Yeah, using these new technologies to make the world more local-first, 764 - > especially maybe parts that have been resistant to that.

765 - > I think breaking down data silos is one of the biggest ones that's been such a long 766 - > running topic at you could switch with projects like Cambria, but almost always 767 - > it involves kind of a convince everyone to build their software on a completely 768 - > different stack, which sort of Yeah. 769 - > Is very much an uphill battle. 770 - > Right. 771 - > And as you point out, there's a great agency in our personal lives.

772 - > I've experienced this already with, yeah, being able to pull out, I don't know. 773 - > You know, there's some website that has a list of things and I want 774 - > to put that into a spreadsheet, but it doesn't offer a CSV export. 775 - > Okay. 776 - > I basically could do that with a screenshot or a PDF and let the language 777 - > model do the hard work of, you know, turning it into something coherent.

778 - > Similar idea there is just downloading. 779 - > I like to download my transactions from my bank account on a quarterly basis, so I 780 - > have a record of that that's separate from whatever clunky interface the bank has. 781 - > But inevitably, if you have a couple of accounts and then they're all in slightly 782 - > different formats and trying to like bring them all together into any kind 783 - > of unified ledger is nearly impossible. 784 - > It's just not worth the bother for an individual or whatever.

785 - > But actually now with language model, it's pretty easy to do an ad hoc thing where I 786 - > can go do the download effort myself once a quarter, but then pull that all into a 787 - > spreadsheet with a normalized data set. 788 - > That feels very empowering for me. 789 - > It's like, okay, now I have all my transactions in a place that I control 790 - > in a format that I want that isn't dependent on, you know, my bank letting 791 - > me, you know, I go to login to download something and it tells me, oh, sorry, 792 - > we only support transactions, you know, back six months or whatever thing, right?

793 - > Very disempowering user experience that instead it's something that data 794 - > that feels like it belongs to me, my banking and transaction data, and I 795 - > have it in a place and a format that I can control for the long term. 796 - > Exactly. 797 - > it's just so cool to see like how far people take, like how much they embrace 798 - > the new forms of like, agencies that they, I guess they, they've been always like 799 - > longing for, but now, like it becomes feasible to, like really go for it.

800 - > And I think it's so, also so liberating that you can build software for 801 - > yourself unburdened by like making it enterprise grade, et cetera. 802 - > And so you can like, you know, I need to hold it this way and it's 803 - > gonna work, otherwise it falls apart. 804 - > But that's fine. 805 - > And I think that's like the, beauty of like building tools for yourself 806 - > and like local-first gives us the glue that they actually can bear like 807 - > real weight and like work together.

808 - > So that is the new territories, which I think is very related to the larger 809 - > ecosystem where we have, a set of like other, very interesting topics. 810 - > So like some of them we've teased before, but just to run through like on a high 811 - > level and then we can go into them like again, we have like self-sovereign 812 - > identity, we have social media and data ownership, encrypted messaging, 813 - > game development with local-first and malleable software in the age of LLMs.

814 - > So maybe starting with the self-sovereign identity, what 815 - > should people think about here? 816 - > Which sort of stories might we be interested in having at the conference? 817 - > Yeah. 818 - > Well in general, this category of the larger ecosystem is one I'm excited 819 - > about because I think there are so many adjacent communities that are 820 - > interested in, yeah, for example, things like freedom, you know, the open source 821 - > community, obviously the Linux world.

822 - > The Open Web and open standards, as well as things like the end-to-end encryption, 823 - > kind of privacy, nerds out there. 824 - > You know, there's, that's not exactly what local-first is 825 - > about, but it's very adjacent. 826 - > There's a lot of shared technologies and things, and this is an invitation 827 - > for folks in those communities. 828 - > And I'm even sending out some emails and so forth to folks to try to invite them in 829 - > to say, look, let's see where we overlap.

830 - > Let's, find the edges there. 831 - > yeah. 832 - > The identity topic though, this is a big one for me. 833 - > It's something I've been fussing about my whole career.

834 - > I have whole, article about it in my making computers better 835 - > essay, from a few years ago. 836 - > But it basically boils down to, I think the way we identify ourselves to 837 - > computers is kind of an embarrassment, usernames and passwords, identity theft. 838 - > Now you've got, I don't know, magic login links and login with Google 839 - > passkeys and whatever, and it's just endlessly confusing for most 840 - > users and even for power users. 841 - > Yeah, the amount of time in my day, I end up spite typing in two factor 842 - > codes and digging magic links out of emails and just trying to convince 843 - > the computer that I am who I say I am.

844 - > And I know why it's that way. 845 - > Of course, as a professional that's worked in this field and particularly 846 - > in the security, space, on and off, over the course of my career. 847 - > But it really is a shame to me. 848 - > And then in the meantime, you have this more and more control and 849 - > more and more ownership over our identities that go to the companies.

850 - > And basically it's been a kind of a good deal to say like, look, I'll 851 - > just let Google manage my identity and then I can click the login with Google 852 - > button everywhere, because it is such a hard problem and I can just kind 853 - > of outsource that to them, sort of. 854 - > But there, it's kind of like a little bit of a deal with the devil there 855 - > comes with a lot of trade offs. 856 - > and local-first is a chance to, rethink that, reinvent that. 857 - > and of course there are also a lot of adjacent communities.

858 - > Things like Open ID for example. 859 - > We'll talk a little bit more about, app proto and ActivityPub in a moment. 860 - > But they have their own concepts of identity and how you give people ownership 861 - > over your online handle, in a way that's sort of secure and safe, but doesn't 862 - > put the user in the position of needing to manage a bunch of cryptography keys, 863 - > which they're inevitably going to lose. 864 - > so I think anyone working on anything in this space, whether it feels 865 - > local-firsty or not, if you're working with passkeys, if you're working 866 - > with, any kind of identity system, yeah, we'd love to hear from you.

867 - > Yeah, I think this is a highly. 868 - > unsolved area. 869 - > I wouldn't say under explored. 870 - > I think, people are really exploring it.

871 - > It's a hard problem in really interesting ways. 872 - > It is a hard problem and I think we've, we've also come a long way already. 873 - > Like for example, like if you're depending on, which pact with the 874 - > devil you've made, you can already have a slightly more convenient life. 875 - > So me being, in pact with the Apple ecosystem, I am very thankful for things 876 - > like Touch Id, et cetera, that are then like nicely working together with things 877 - > like 1Password, Passkeys, et cetera.

878 - > and like being able to understand how those things work and like thinking 879 - > through the implications, et cetera. 880 - > Like let's me be for, for this point in time, in a relative sweet 881 - > spot between, like living a secure life, but also having convenience. 882 - > But there's still like great levels of like lock in, et cetera. 883 - > And I think this is also like all of that aside, like we're now also, 884 - > having this like phase shift in regards to software where we are no 885 - > longer the only ones who are doing things on our behalf, with computers.

886 - > Like if we're, now whatever, like maybe ordering, like some food or 887 - > we're doing some other things, maybe we're, like scheduling something on, 888 - > the calendar or whatever it might be. 889 - > It's no longer just us, but there's like another thing that needs 890 - > identity and even more fine grains. 891 - > And I think one aspect here as well is like if you're going from beyond 892 - > authentication to authorization, so far we have like, all off scopes, et cetera, 893 - > but like it's very all or nothing.

894 - > And it's all like, okay, you, now get access forever. 895 - > But maybe for like those agentic use cases, you might want to think about it 896 - > like, okay, you get to do this like one time, or you get to do this like for five 897 - > minutes, or you get to do this like only for this email address or for this scope. 898 - > So I think we want to make this like much more fine-grained, but in a way 899 - > where you like, it's, it's frankly like going through like the permission 900 - > granting on like a Google OAuth flow is already overwhelming for most people.

901 - > And now making this even more fine granular 902 - > Yeah, 903 - > I think is like really 904 - > tough challenge. 905 - > Thinking of a 906 - > giant 907 - > array of check 908 - > check 909 - > boxes on the GitHub, token generation and there's a reason for it. 910 - > But then yes, again, that trade off of, convenience and security is ever present. 911 - > I think it's that intersection of like user experience 912 - > design and also cryptography.

913 - > Like how do you make it actually secure? 914 - > and I think that to, come together in the best of both worlds. 915 - > I think that is like a area that's like highly deserving of like 916 - > more resources, more investment, more smart people working on it. 917 - > So if you're working on that, we'd love to hear from you.

918 - > ideally where you cover quite a bit of that intersection. 919 - > I think this is probably not the best, conference to share 920 - > like your new breakthrough of like some cryptographic nuance. 921 - > But I think where we can like bring those two things together where you can also 922 - > with the focus of like, or one aspect can be the solve sovereign, aspect. 923 - > I think it doesn't have to be a prerequisite, but I think if 924 - > something points in that direction.

925 - > And also I think another, tricky part here is like that we need to kind of 926 - > create a bridge from like today, where people are like, defacto using things 927 - > like login with Google, et cetera, because it strikes A, acceptable balance 928 - > of like convenience and security. 929 - > And if there's like a much to esoteric, version out there that doesn't have a 930 - > credible path to becoming mainstream, I think that also is something that 931 - > needs to be taken into consideration.

932 - > So this is a very nuanced topic and I think as one of 933 - > the harder nuts to crack here. 934 - > but if you're working on that, we'd love to hear from you. 935 - > And I think with that it is like a natural transition to, things like 936 - > BlueSky ATproto where we talk about, social media and data ownership. 937 - > So you've already mentioned initially.

938 - > That Martin Kleppmann is affiliated with the ATproto ecosystem, and I think that's 939 - > just like one example for a broader, set of design challenges and like goals 940 - > that I think is like nicely adjacent or partially overlapping with local-first. 941 - > So do you want to give a little bit more detail on this topic? 942 - > Yeah, for sure. 943 - > I think social media and data ownership is so interesting because social 944 - > media is the classic centralization and the classic opaque appliance 945 - > with a mysterious algorithm that you can't and don't know how it works.

946 - > And in general, I would say we could, obviously could, and indeed I have 947 - > in the past done whole podcasts on social media and the both the 948 - > positive and negatives that that has. 949 - > Impacts that has had on society and us as individuals, but narrowing in 950 - > a little more on this kind of, yeah, user empowerment around, and data 951 - > ownership in our computing lives. 952 - > It is the nature of the classics like an Instagram or a Facebook or even a 953 - > Twitter back in its heyday that, you know, it, it is a very centralized 954 - > closed, you have little ownership.

955 - > You can just participate in this stream of, content. 956 - > and now we have this new crop of, creations. 957 - > I think Mastodon and ActivityPub was, kind of early out the gate, 958 - > but now there's some really good energy around, BlueSky and ATproto. 959 - > Actually, I think even as right as we record this, the ATmosphere 960 - > Conferences is happening.

961 - > which yeah, just a lot of good energy in there and it's 962 - > a different set of problems. 963 - > I think the classic local-first app is something more like Linear 964 - > or Notion or Figma, which is a document editor, heavy creative work. 965 - > You're collaborating with a small team. 966 - > You really care a lot about the output of that data.

967 - > Social media where it's sort of global town square, little fragments of 968 - > data, everyone gets merged into a big stream is on the surface, a completely 969 - > different, both technical and kind of design or user experience, challenge. 970 - > But at the same time, when you look at what ActivityPub and ATproto have 971 - > been grappling with, including things like personal data servers and how 972 - > you have ownership over your identity, how you make your handle portable, 973 - > but still kind of globally unique and especially verifiable, trustable, right?

974 - > not easy to impersonate that you're the New York Times or a 975 - > politician or something like that. 976 - > So there's some really interesting technical and design work that has 977 - > been done in both of those communities. 978 - > I think we already have, some folks lined up to speak on some aspects of those. 979 - > But I think there's, that's such a rich area that if you're working on something 980 - > in this space with either one of those communities or otherwise in the kind 981 - > of broader, how do we make social media something more empowering for users, 982 - > yeah, we'd love to hear that, we'd love to see a, talk submission from you.

983 - > Yeah. 984 - > In particular also like using those protocols, those ideas to go beyond 985 - > social media where you can basically like to, draw that bridge to the 986 - > previous, like, identity question. 987 - > Like right now I think this is, we we're blind to it at this point, that 988 - > when there's new software, like, okay, of course I need to create an account. 989 - > And like, depending on, which sort of patterns you use, maybe 990 - > you automatically like, look for that sign in with Google button.

991 - > Or if it's a developer oriented thing, like sign in with, GitHub, I'm more in 992 - > the camp of like, where I always try to like just, create a standalone account 993 - > just in case, something were to happen. 994 - > but does it have to be this way? 995 - > Are there other ways and like, and not just for. 996 - > social media use cases, but for something that spans a much, wider range of 997 - > scenarios, whether that is like something, I don't know, like maybe there's something 998 - > in between where it's something in between Figma, Linear and, and social 999 - > media, whether it's like a, smaller scale community that you are part of where 1000 - > you do some very interesting things.

1001 - > Maybe a community for a school that you volunteer in, or you have a much 1002 - > richer set of different, kinds of media you deal with, whether it's documents, 1003 - > whether it's chats, whether it's like some, posts, some forum like things. 1004 - > Then you also have like things like identity. 1005 - > And I think this is just like an example for thinking about 1006 - > this a little bit more broadly. 1007 - > So if you're using, some of those technologies or exploring that.

1008 - > And see how we can bridge those worlds. 1009 - > we'd love to hear from you and, going from one, one, tricky technical, 1010 - > nuanced topic to another one. 1011 - > Another like evergreen in local-first and beyond is like 1012 - > encryption and encrypted messaging. 1013 - > So, I think highly related to the areas that we talked about before.

1014 - > And I think is another, aspect of software. 1015 - > I think for a lot of software, kinda like still an optional part because it's hard, 1016 - > but ideally over time it becomes just like a no-brainer to have to have encryption, 1017 - > for your, data, for your applications. 1018 - > It is, I think also dependent on the kind of like app you're working with. 1019 - > So for example, if you're working in the finance space or the 1020 - > healthcare space, I think it's much more, table stakes already.

1021 - > And then there are other areas where it's much less common. 1022 - > but ideally this is a quality that does not, quadruple the effort it takes 1023 - > to build the app, but it's some just something, a data library, for example, 1024 - > like Automerge takes care of for you. 1025 - > And so yeah, those are like some preliminary thoughts 1026 - > on this topic from, my end. 1027 - > But yeah, curious to hear yours, Adam.

1028 - > I mean, end to end encryption is first of all just such a powerful technology. 1029 - > You know, I think wars have been one with, it is one frame you could use yet remains 1030 - > hard to put into practice in a way that is approachable for users because, you 1031 - > might think of it, everything boils down to a key management problem, but managing 1032 - > keys is really difficult and incredibly abstract, even for very power users. 1033 - > And certainly for a more mainstream computer user.

1034 - > Yet you see all these places where encryption, when it has been brought to 1035 - > bear effectively in the computing world, it is a complete game changer, I think 1036 - > of like SSH for connecting to servers versus Telnet, which came before it. 1037 - > I think of HTTPS for website encryption has, I think, literally brought trillions 1038 - > about dollars of value into the economy through enabling things like e-commerce 1039 - > and online, stock brokerages and so on.

1040 - > Something like password managers and how they can provide a secure, setup, 1041 - > based on their, encryption arrangement. 1042 - > And then more recently you've got, for example, encrypted messaging like Signal. 1043 - > And I think when we find ways to bring those to bear in a way that 1044 - > is approachable for end users, it can just unlock so much value. 1045 - > But it remains the case that there are a lot of areas where we haven't really 1046 - > effectively brought it to bear fully.

1047 - > And yeah, if you're working on any of these, whether it's something 1048 - > relatively well established like encrypted messaging or something a 1049 - > little more, on the cutting edge. 1050 - > Like last year we heard a, an amazing talk from Brooklyn Zelenka about 1051 - > Keyhive, which is a kind of adding an identity encryption layer to Automerge. 1052 - > This just remains an area that is very hard, but I think again, just has the 1053 - > potential to provide so much value.

1054 - > Awesome. 1055 - > so another adjacent topic, a little bit more lighthearted, and yet probably one 1056 - > of the most, OG adopters of local-first, and they've always built software 1057 - > this way, which is game development. 1058 - > I think most games, particularly in the era of like internet not 1059 - > being as ubiquitous as it is today, have like, out of necessity. 1060 - > They needed to be local-first.

1061 - > So you can play single player campaigns or whatever, but also like 1062 - > in that gold, era you had, a lot of games also having multiplayer modes. 1063 - > And this is where you had like LAN parties, et cetera. 1064 - > And like, software was just like built in a different way where like internet 1065 - > a lot of games, even didn't like support it or like was one, feature 1066 - > that was added a little bit later. 1067 - > And it's just like a.

1068 - > I think an interesting case study for how like a different parallel universe evolved 1069 - > and like did very complex data management with a lot of like similar challenges 1070 - > where you needed to have like, reactivity you needed to be like, very efficient. 1071 - > You needed to like care about persistence. 1072 - > so I think there's like a lot of parallel universe lessons learned 1073 - > that maybe never really, fully percolated into the web ecosystem. 1074 - > Probably sometimes for, the worst.

1075 - > I think I always feel, I can learn a lot from the game ecosystem overall, as 1076 - > a lot of like lessons have been already much, earlier been learned there. 1077 - > And I think now it's like that, that is still true today. 1078 - > And video games have always been like the, one of the most demanding use cases 1079 - > and where like latency really matters, where like all of those things, matter. 1080 - > And, I think it's just about time that we bridge the world from game 1081 - > development to the local-first ecosystem.

1082 - > So if you're, doing things in the game development space and you have 1083 - > some interesting stories to share about how your game has always 1084 - > been local-first, then, please definitely consider submitting a talk. 1085 - > So that, that's kind of my, my perspective on it. 1086 - > But I think you have even more nuanced takes on this, Adam. 1087 - > Well, yeah.

1088 - > I don't know about nuance, but it's definitely close to my heart. 1089 - > I actually began my career in video games. 1090 - > I probably like a lot of little kids, I basically got inspired to 1091 - > learn to program, to make games. 1092 - > And, began my career there.

1093 - > I worked for four or five years in the game industry, including on 1094 - > PlayStation 2 games and Dreamcast games. 1095 - > And that was kind of the era then. 1096 - > And it's an incredible world of technology that's very separate from the productivity 1097 - > software, web operating system, thing where you're just drawing 60 frames or 1098 - > 120 frames per second on a blank canvas. 1099 - > The APIs to the underlying system are just a very, very different thing 1100 - > than perhaps what we're used to, in some ways easier to work with, in some 1101 - > ways harder, but yeah, the performance and being able to use it wherever.

1102 - > even certainly in the age before internet connection was ubiquitous, 1103 - > but even nowadays, yeah, you expect if you have a game on your device, that 1104 - > you should be able to play it on the plane and not be worried about whether 1105 - > you have internet in that moment. 1106 - > Even think something like, you know, even once you start to add some 1107 - > internet connectivity, Hey, I wanna post my high score to a global thing, 1108 - > but you don't expect the game will just stop working the moment that the 1109 - > high score API is not available, so.

1110 - > And then you have so much going on in the world of these different engines like 1111 - > Unreal and Unity and Gadot engine slash game development tools and yeah, I just 1112 - > feel like there's, maybe a lot these communities can learn from each other 1113 - > and yeah, if you're someone in that, tough space, we'd like to hear from you. 1114 - > Yeah, I think there are just so many amazing stories waiting to be heard in 1115 - > the local-first ecosystem, whether it's someone who like struggled to scale 1116 - > World of Warcraft, to like massively multiplayer scenarios where I mean, 1117 - > it's already hard to build multiplayer.

1118 - > like systems for 10 users, but now, like, think about this, like in the, tens of 1119 - > thousands of users, in like, in much harsher, real time scenarios, et cetera. 1120 - > So, or whether you're building like, more modern games that, they're currently 1121 - > like scaling things even further. 1122 - > so yeah, we'd love to hear from you. 1123 - > if you think this is interesting for the local-first community and as the 1124 - > last topic, in the list, we, we probably like the, kinda like the key topic that 1125 - > is the, center for us this year is like malleable software in the age of LLMs or.

1126 - > malleable software like at all. 1127 - > But then also in the age of LLMs, that's new. 1128 - > That was not really like the, we got like the early glimpses of that last 1129 - > year, but at last year's conference, but I think it was still so early, I don't 1130 - > think like Claude Code was released yet. 1131 - > ChatGPT was out there, of course, but I think it was not yet at a 1132 - > point where it like really clicked for everyone that this is real and 1133 - > not just like a niche use case.

1134 - > And I think now it's like, already so ubiquitously available that we can 1135 - > make this a part of our application. 1136 - > And I think there's like really one of the broadest topics here, almost like a 1137 - > catchall for a lot of, different, talks that could fit into the conference. 1138 - > the anchoring point here is like, uh, what we've talked about before with 1139 - > like Maggie's, closing keynote from 2024 was it, with like barefoot developers. 1140 - > And I think this can, cast a really wide net.

1141 - > And, what are your thoughts on, this topic? 1142 - > Yeah, we'll probably first just defining malleable software. 1143 - > For those not familiar, this is the idea that software systems are ones 1144 - > that could be modified by their users as we're using it or to adapt to our needs. 1145 - > and of course, Ink & Switch has a, I dunno, 8,000 word essay up on the website 1146 - > if you search for malleable software, that can dive into detail on that.

1147 - > But, you know, how does language model assisted coding it seems to enable so 1148 - > much the ability to create new software and especially for people who are on the, 1149 - > not necessarily professional software developers, even if they're technical and 1150 - > good at thinking about software systems. 1151 - > But when you think of the classic, let's call it vibe coded app, it's 1152 - > let me make a one-off website that I can post, but that's very different 1153 - > from being able to modify the tools that I'm relying on in my daily life.

1154 - > My text editor, my email client, my web browser, et cetera. 1155 - > so I think on one hand, language model coding should, in theory, 1156 - > really open up some major new possibilities for malleable software. 1157 - > On the other hand, I think we have some of the same old problems we have always had 1158 - > around data siloing around kind of closed ecosystems, closed appliance heavily. 1159 - > You know, the control, the locus of control lies completely with the software 1160 - > vendors and very little with the users.

1161 - > and how has this world evolved or what new possibilities are opened by the 1162 - > existence of language model assisted coding or other AI things or other things 1163 - > that have changed in the technology world since we last convened in 2025? 1164 - > Yeah, I think this is, I mean, this topic in itself has been sort of like a core 1165 - > pillar for this year's conference, and I think we've like added quite a lot other 1166 - > ones just because, it felt like the right moment to, to broaden the net a bit.

1167 - > But I think here, like malleable software, this was, a theme that I 1168 - > mean, is is one of the key tracks of, the Ink & Switch, family. 1169 - > But I think now with AI models, et cetera, it's really something 1170 - > that's becoming mainstream. 1171 - > And where before I think it was like very something that developers 1172 - > maybe had romantic thoughts about. 1173 - > Like, oh yeah, this was like, also on my list of side projects I wanted to build.

1174 - > And now you can actually build it for yourself. 1175 - > Or maybe you built software for like someone in your family, or it's 1176 - > just like a, almost as an act of kindness of like supporting someone 1177 - > like this is through your abilities. 1178 - > And now it becomes so easy. 1179 - > I think now that will lead to so much, creation of, new software that will also 1180 - > have like interesting new consequences.

1181 - > whether it's that we need to wrestle with, the quality of the software, 1182 - > or the interoperability of it. 1183 - > And also like, yeah, how much control do we give to, users? 1184 - > How much control, do we, like where do we draw the line between, 1185 - > a developer building the software versus a user using the software? 1186 - > Like a lot of power tools are very customizable.

1187 - > But now I think that line gets more and more blurry. 1188 - > We'd love to hear from people who have been wrestling with, those 1189 - > blurry lines and trying out different ideas and sharing your learnings. 1190 - > What sort of software you've been building as a barefoot developer. 1191 - > Yeah.

1192 - > I think this is, quite the portfolio of, different themes and 1193 - > different topics that we're gonna have for, this year's conference. 1194 - > That's right. 1195 - > We maybe even run the risk of being too scattershot or too wide 1196 - > ranging, but in some ways I'd rather. 1197 - > take the risk of having too many ideas, too many interesting, 1198 - > intellectual, avenues to pursue, then get stuck in a rut.

1199 - > And now seems like a great time to do it with the changes happening, in the 1200 - > technology world, but in general because of the foundation of success we've had 1201 - > in the event to date and the community we've built around, that just seems 1202 - > like the perfect time to, broaden and bring new people in and fresh ideas. 1203 - > And like all of that is not to say that we ran out of ideas with like the, 1204 - > traditional way how we did the Local-First Conference, the last two years before.

1205 - > But I think what it's always been all about for me and for I'm sure for you as 1206 - > well is like the people who actually come and like the conversations we're having. 1207 - > And it is like no one in that ecosystem has been there only because they 1208 - > only care about the data sync engine. 1209 - > It's always been in service of like some broader vision and like some specific 1210 - > use case that really like, had real impact and real meaning for their life.

1211 - > And like I think all of those different themes we now talked 1212 - > about is like one manifestation, one direction of that and local-first. 1213 - > It's like the foundation of that. 1214 - > It's kinda like what brings us all together. 1215 - > And I think what all of those people who have attended and have contributed to 1216 - > the conferences have, in common is that they have a very, deep understanding of 1217 - > the, technical world and of the world more broadly, but, also a broad interest 1218 - > and, are always keen to be inspired to think a little bit different about, 1219 - > like be beyond their current horizon.

1220 - > And I think this is, what we, try to, also, facilitate through the conference 1221 - > where those, different talks are hopefully little points of inspiration that can 1222 - > lead to further exploration by everyone who's attending and paying interest. 1223 - > Well said. 1224 - > Well, I think we hammered in the, you know, submit a talk call to action. 1225 - > we'll say it one more time.

1226 - > The CFPs open till I think, end of April. 1227 - > So you have a little bit of time, but not too much time. 1228 - > But also if, even if you don't wanna give a talk, buy a ticket, 1229 - > just come listen to all the ideas. 1230 - > we got sponsorship booths available as well.

1231 - > If, the folks at the, an event like this are interesting to, talk 1232 - > to for your, for your company. 1233 - > So we'd love to have you come participate in any of those three forms. 1234 - > Awesome. 1235 - > Well, Adam, thanks so much for talking this story together.

1236 - > I feel like this is also already Yeah. 1237 - > Given me so much excitement for, what we're gonna hear, at the conference 1238 - > and like in a way where we still don't fully know which stories we'll hear, 1239 - > but we'll know about the broad topics. 1240 - > So that've got me very excited already and I can't wait for, the conference and 1241 - > to meet a lot of familiar and new faces. 1242 - > Same here.

1243 - > I'm excited for the new themes. 1244 - > I'm excited to see old friends, and I'm excited to enjoy beautiful, 1245 - > beautiful Berlin in the summertime. 1246 - > All righty. 1247 - > Thanks so much.

1248 - > Take care. 1249 - > Bye. 1250 - > Bye.