localfirst.fm

A podcast about local-first software development

Listen

Conference

Special: Local-First Conf 2026

CfP for Local-first Conf 2026 is open. Adam and Johannes discuss this year’s conference themes.

Mentioned in podcast:

Links:

Transcript

Intro
00:00AI is really shaking up everything.
00:02I think putting software as a entire, like industry a little bit in a different
00:07light now, and, I think this does not make local-first at all, less relevant.
00:13I think this is, as for most things, a huge tailwind.
00:16But given that, both of us have been really like, immersed in that new AI
00:21space as well, now, I think we can come with a perspective where we can speak
00:26as insiders from the local-first space, but also as insiders from the AI space.
00:32And with that fresh perspective, I think we wanted to give this year's
00:35Local-First Conf a new, code of paint.
00:39Hey Adam, so great to have you back.
00:41How are you doing?
00:42Hey Johannes.
00:42I'm good.
00:43Looking forward to the arrival of Spring here.
00:46So I was just joking that the two of us have been on the last
00:51localfirst.fm episode since it hasn't really been on a regular cadence
00:56over the last couple of months.
00:57I'm taking a bit of a, pause from the podcast to reorient myself
01:02in this new crazy world of AI and coding agents, et cetera.
01:07And like, I wanted to get my own hands even more dirty again.
01:11And yeah, little bit recalibrate, and I think that's also gonna be quite the theme
01:16for, an announcement we're making today.
01:19And to foreshadow a little bit.
01:22What this year's Local-First Conference will be about.
01:25So with that out of the way, yeah, curious what's on your mind, Adam?
The Evolution of Local-First Conference
01:31Well, obviously as we ramp up to this Conference and we're,
01:33we're doing something pretty different this year, I think so.
01:35I don't know.
01:36Maybe I can give some, some history for, I espect a lot of your listeners have either
01:40been to the Conference or know about it.
01:42but just for context, this will be our third year and we started in 2024.
01:48we really weren't sure honestly, if there was enough demand for it.
01:52You sort of had the idea of like, Hey, I think there's space for or
01:56interest, you know, a latent interest in local-first and we could build a
01:59conference around it on data ownership and CRDTs and all that goodness is
02:03things that yeah, need a venue and a place to meet in person and share ideas.
02:08And we were genuinely really concerned.
02:09We just wouldn't sell enough tickets and we'd lose money on the
02:12venue and all this kind of thing.
02:13Instead, what happened was, I think our venue held like 150 or 180, something
02:17like that, we sold it out within a week.
02:18It was actually kind of embarrassing because we had to be turning wonderful
02:22people away and plus ones for our speakers and things like that, just
02:25'cause we were at the venue capacity, so.
02:27And for me it was also a big question about bringing together eclectic mix
02:32of people, you had both academic people who had been working on CRDT theorems
02:36and things for a decade and more kind of fringy, you know, Brett Victor, Alan Kay.
02:43Tools for thought, perspectives with more practical software
02:48engineers build React apps, just looking for the newest and greatest
02:51technology to make their jobs easier.
02:53Startups, there was, sync vendors were starting to become a little
02:56bit of a thing then, and, you know, would these people even gel?
02:59And I think the answer was yes.
03:00The event was a huge success one day.
03:02Single track, everyone loved it.
03:05and it was a pretty obvious thing then the following year, 2025 to
03:08expand it, we had a venue for 350.
03:10We did two full days plus a community day, to more talks, more people.
03:14Everything was 2 to 3x the scale.
03:18And I think it really showed that we had this, community here and
03:21that, that there was a lot of cohesion and people wanted to come.
03:24Come back to it at the same time, I think, because sync and sync engines were
03:27having a pretty big moment last year that ended up being a, at least one of the
03:31day, was largely given over to that topic and we explored that pretty thoroughly.
03:36So now we find ourselves thinking about, okay, one, we just enjoy
03:40bringing this group of people together in this community of people together.
03:42We want to get our friends back together in Berlin in the spring or the
03:46summertime to just spend time together.
03:48But at the same time, we also know the industry's changing and we also
03:51know that it would be a little boring to just retread the same ground.
03:55So that's causing us to want to be a little more creative with the
03:58direction we're going this year, which is I think why we wanted to get
04:01together and talk about that today.
04:02Yeah, and I think if you're
04:05just now looking back over the last two to three years, I feel that the
04:10world is just such a different place.
04:12When we look back then, local-first was really like, a wake up call
04:17in this kinda like very almost.
04:19not by today's standard, almost like static world, where things have
04:24feel like seem pretty settled in terms of technologies, et cetera.
04:28And we have like our best practices.
04:30And local-first at that point was much more of sort of like a provocation,
04:35hey, we can think differently about this and like rethink, the
04:40typical like cloud model, et cetera.
04:42And like it stirred up quite a bit of dust.
04:44And I think this is what, what got people really interested about like
04:47insiders who've been thinking about those lines for like many years.
04:51Some even like, decades and have always built software this way.
04:55And, then last year's conference I think has really shown a huge step to the year
05:01before in terms of like how much further, how much mature the ecosystem has gotten.
05:07But, I think it would be, like this year is different.
05:11It's not just like, yes, the ecosystem has gotten a lot more mature and like
05:16all of those technologies are more production ready, there's like more
AI's Impact on Local-First
05:24AI is really shaking up everything.
05:27I think putting software as a entire, like industry a little bit in a different
05:31light now, and, I think this does not make local-first at all, less relevant.
05:37I think this is, as for most things, a huge tailwind.
05:41But given that, both of us have been really like, immersed in that new AI
05:46space as well, now, I think we can come with a perspective where we can speak
05:51as insiders from the local-first space, but also as insiders from the AI space.
05:56And with that fresh perspective, I think we wanted to give this year's
06:00Local-First Conf a new, code of paint.
06:03this is where, you've been this year leading the effort around the CFP.
06:08And yeah, so I'm excited to, talk through the different themes we're expecting
06:13for the conference where we have a set of curated speakers already, but
06:18we think, the best content is really brought up from the community, who is
06:23building things in the local-first space.
06:25And this is, like the people who are maybe listening to this podcast, or who
06:29have attended, the last few conferences.
06:32So yeah, this is, what I would love to talk to you about.
06:35Yeah, well, maybe we can just start with that call to action really upfront, which
06:38is we're broadening the umbrella for what kind of talks we want a lot this year.
06:42And so if you do nothing else, you can just stop listening to the podcast.
06:45Go read the CFP page.
06:47If you think there's any chance of what you're working on, which has something
06:49to do with empowering users or making, computers and software and the internet
06:55freer and more capable for enhancing human life, then you should submit
07:00a talk that's your CTA right there.
07:02To drill, one level deeper.
07:04you already mentioned the AI topic and obviously that feeds into this.
07:07There's also the desire to continue to build and expand on the themes of this
07:11community and respond to other things that are happening in the tech industry.
07:15On the AI side, obviously, where I believe, you know, creation of software
07:19is changing fundamentally with AI assisted coding, and I think people in our
07:22community have a very wide mix of views.
07:25There's people who come from a. very skeptical point of view, I probably
07:28lean more in that direction myself, especially up until recently, there's
07:32some bubble, it's gonna burst soon, it's overhyped, it's, you know, whatever.
07:36And on the other end, there's probably plenty of maximalist, people that,
07:39you know, I think it's not even such a bold thing now to say that 2026 might
07:43be the last year that humans do all that much direct coding in terms of
07:48like writing out programming syntax.
07:50and I suspect people listening fall in a wide range of, views there, and
07:54we wanna make room for all of that.
07:56But, you know, at the same time, we're acknowledging that this industry shift is
08:00happening while also staying grounded in our values and what we're all here for.
08:05For the record, October, 2025 was the last time I wrote a
08:08single line of code by myself.
08:10So, but I'm leaning into that, quite heavily.
08:14But I think as you say that's a wide spectrum.
08:16And nonetheless, I think that makes, even more clear that we need really
08:21good principles for the software we're creating, and that we also want to use, I
08:26think, software through AI, like through the terms like slop software, et cetera,
08:31I think has made it even more, even more important to really care about like what
08:36we're using, not just for the perspective of that "it's nice to use", but also for
08:41the perspective of like, data ownership, privacy, security, all of those.
08:47And I think local-first is still the guiding light for us.
08:52It's kinda like the gold standard for what software should feel like, what
08:57software should be architected around.
09:00And I think this is where we hopefully get the best of both worlds, where we
09:04can have the cake and eat it too in that regard that we can dream about
09:09software that follows those principles.
09:12But now it's a lot more feasible to build those and to build those,
09:17not just for the sake of like building software for potentially
09:22like a million target users, but for yourself, like personal scale software.
09:27And this is where we had this amazing closing keynote from Maggie
09:32Appleton at the first Local-First Conference about Homecooked software.
09:36And I think in a way this year,
09:38barefoot developers, I think was the term she, she used.
09:41That's right.
09:42And certainly what you could do with language model assisted coding
09:44back then was, you know, much.
09:47more limited than what you can do today.
09:49But even then, she predicted that, which I think is a good, a good illustration
09:52of the kinds of people that are in our community that, you know, were
09:55values driven, but very forward facing.
09:57She, she saw what was coming years before.
10:00Exactly.
10:01And like now, this year's conference is basically like, really like double
10:05clicking on her entire vision here where like everyone can dream about
10:10software and like now make it a reality.
10:13And local-first doesn't just make it, more secure and private, but like overall
10:18also like solves a whole bunch of other problems that is still hard with software.
10:23So I'm really excited how like those two or the, how the conference
10:27can be a place where those two different worlds can come together.
10:31All of like the relentless, accelerated AI community, who wants to, build things
10:37really quickly, but also the people who like really care about software
10:41quality and like those principles.
10:43And this is something where I think it can be, a really nice meeting space for folks.
10:48So, Adam, how about you wanna give us a broad overview of the different
User Empowerment as a Central Theme
10:57Yeah, sure.
10:58So the overall theme for the conference this year is user empowerment.
11:01that comes from, actually a panel we had last year.
11:04I host a panel and one of the, panelists was Martin Kleppmann, who
11:07this year is a keynote speaker for us.
11:09Obviously he's really one of the, founding fathers of, local-first.
11:13But the way he described it was, we were talking here a little bit about the kind
11:17of the specific implementation details of sync engines and CRDTs versus the
11:22broader set of values and principles.
11:24And I think he put it really well in saying.
11:26It's all about user empowerment.
11:27If something gives users more empowerment, more control, more capabilities, more
11:33options, more freedom, then that to him falls under the umbrella of local-first.
11:38And if it's something that sort of takes agency away, which there are times
11:41where that makes sense, especially in consumer software or whatever, that
11:45people want less control over how they're using their computers, that's fine.
11:49But that, leaves it outside the realm of local-first.
11:51So we're borrowing that idea and expanding on it.
11:54User empowerment, greater agency, obviously a lot of that tends
11:58to flow through data ownership.
12:00'cause the data ultimately when you're creating documents on a
12:03computer, the data and feeling like it belongs to you is, tends to be
12:07core to that feeling of empowerment.
12:09So with that overall theme, took together with our, collaborator, Eileen Wagner,
12:13also a long time member of the community who is, helping us here with the CFP.
12:19But she basically sorted us into three categories here.
12:21We've got new territories, which of course includes plenty of AI stuff.
12:25We've got local-first maturity, which is trying to see, you
12:29know, now we're pretty far along.
12:30There's a lot of great sync engine vendors.
12:32There's some real world success stories, but there's also
12:35interesting new challenges to face user experience challenges, legal
12:39challenges, or legal frameworks.
12:41And then where we're really trying to expand here is the larger ecosystem.
12:45And there's a lot of exciting stuff happening in, whether it's in something
12:49like the area of game developments, but also something like social media.
12:53I think, you know, with ATproto and to a lesser extent sort of the masked
12:57on ActivityPub world of things is having their own moment and kind
13:00of causing people to ask questions about data ownership and freedom and
13:05decentralization in this realm that traditionally has been a very kind of
13:09centralized, closed, big tech, you know, inscrutable algorithm world of things.
13:15And to what extent are data ownership and local-first values applicable there.
13:19So those three frames, the kind of new territories, the maturity of local-first
13:24and the larger ecosystem are where we have kind of a few subtopics for each
13:28one, and we're trying to hope, hoping to get a good coverage on, CFP submissions.
13:33Awesome.
13:33Thanks so much for providing this great overview.
13:36I would actually love to double click on each of those and dig a
13:41little bit more in, since I think each topic has, so much potential.
13:46And I think we could, like, if we'd had a lot more time and space for the
13:49conference, we could almost like make each of those their explicit like theme day.
13:54And so now we only have, well, those two days for the conference.
13:59for the talks in one extra day.
14:01So we need to curate and pick.
14:03But I think this gives us like a really nice broad buffet of like different,
14:08topics and talks to choose from.
14:10And ideally, those, themes speak to people who want to submit a talk.
14:16So yeah, let's go through them.
14:18So maybe starting with the, local-first maturity, maybe that is, where we, can
14:23reconnect the most from where we've left off the first two years of the conference
14:29and also Sync Conf and, I think this is like local-first maturity is really like
14:34that that is the essence of what last year's conferences, were about, where we
Local-First Maturity and Success Stories
14:43The UX challenges of local-first.
14:45This is, you've mentioned Eileen before.
14:47She's given an amazing talk about that last year legal framework.
14:51Maybe a little bit more top of mind this year around, let's start with
14:56those like local-first success stories.
14:58Anything that comes to mind for you in particular.
15:01Well, for sure the success stories should be, there's obviously people building
15:05libraries, you know, the sync engine vendors who are well funded and have a
15:08lot of customers and that sort of thing.
15:09But what I really wanna hear about is the end user applications.
15:13You mentioned Linear, Linear CTO Tuomas has been kind enough to give us
15:17excellent talks the last two years but I also think of on a smaller scale,
15:21there's something like, we had a talk last year that was about someone
15:24building a forestry application.
15:26So this is where someone walking around with a phone that needs to basically
15:30tag each tree in the forest with what's going on with it, whether it has a bug
15:34infestation, whether it needs whatever else and it's a pretty simple thing.
15:38And you know, I like that talk a lot because also they basically said,
15:42look, our data modeling is very simple.
15:45They built their own sync that needs to, of course, not only work
15:48offline, but for a big chunk of time.
15:50But it's a pretty simple queue.
15:51And unlike full text editing, that is a rich text editing that's very
15:55complicated you actually can accomplish all the goals of local-first with not a
16:01huge amount of engineering for some of these simpler, use cases and domains.
16:05So I think there's plenty of people in the community that have their
16:07hobby projects or whatever that they use together with their friends.
16:12But what I like to see we're far enough along now is what is at scale.
16:16I wanna see a Linear and Obsidian or some of these smaller scale things
16:20like the forestry app that have been doing this for a long time.
16:23You have real world data and you can talk about the challenges of debugging.
16:26You can talk about the challenges of accumulating data history.
16:29You can talk about what are the ugly and difficult parts of this
16:33that we're grappling with from using this frontier technology.
16:37Yeah, definitely.
16:38And I think there's, just thinking through like the software and new
16:41software that I've been using over the last like year or so, I've been
16:45using like Rayon for example, quite a bit, which is like a Figma like,
16:50software, but for more like architecture use cases as I'm like remodeling
16:55the house I grew up in and that's been like really nice to get sort of the
17:00benefits that I'm like spoiled with of like using production software
17:05like Figma now for more use cases.
17:07And I think that's just like one example for like, that seeping into more and
17:11more parts of the software we're using.
17:14And I think that will just, continue and like, particularly
17:18now where, AI is another driver of like personalized software.
17:24I think it's just like very clear that we want those benefits of
17:28like collaboration, that things should work offline, et cetera.
17:32Like if anything, those problems are now even more pressing, where when
17:37we have software that is like meant for individuals, and to have like
17:41that software, be shared, across like family members or, friends.
17:46This is where you don't like, you, probably don't build it out as like
17:50a SaaS architecture from the get go, but you're building this like
17:54in a different way to begin with.
17:56SQLite certainties having a bit of a moment again as well.
17:59So I think we're gonna see a lot more of that.
18:02Files are, are, are are the hot new thing again.
18:06Unix
18:06tools.
18:06Yeah.
18:07Honestly.
18:08Honestly, files could have almost been like another category, in itself.
18:12Like it should really also, and not to sound like a broken record,
18:16but like, yeah, with the AI hat on, files and folders are like the best
18:22foundation for agents to do good work.
18:25And it has, intentionally or not really, like, become, again, a main way of like
18:31how data collaboration works, et cetera.
18:34And so, yeah, I think there's a lot to be explored there.
18:37A lot of creative people doing weird, interesting things.
18:41So if you're, if you're like that, please submit a talk and, yeah, I
18:46think there's like whatever, you've been working on in regards to like
18:50traditional local-first in the sense of like the previous conferences, we would
18:54love to hear from you, but also around the UX challenges, of, local-first.
19:00I think we're still just scratching the surface of like trying to
19:04understand that, trying to like deal with the implications of that.
19:09I think we're allowing ourselves to dream a little bit bigger or a little
19:13bit more nichey with software now.
19:16And now that might require like new user experience paradigms and like,
19:22particularly also when it comes to extending collaboration just beyond
19:27people, but also like collaborating with like other agents, is another
19:33aspect that I think squarely fits into the realm of local-first as it's
19:38another like kind of trust boundary.
19:40You wanna understand what has happened.
19:43Do you want to collaborate?
19:44Do you wanna like detach, be by yourself?
19:47Do you want to like revoke authentication, authorization, et cetera.
19:52Maybe real time collaborate.
19:54I think there's a lot here, whether you want to look at us for the AI lens or
19:59not, but I think that area is like super underexplored and if you've been exploring
20:03that, we'd love to hear from you.
20:06Yeah.
20:06The UX challenges for me as part of the.
20:09I don't wanna say quite say fun part, but sort of it is because it's a chance
20:13to rethink those computing primitives.
20:15Right?
20:16And the files example is a good one to me, where files and the Unix
20:19methodology and everything as a tool and pipes and redirects was really kind
20:25of a first love for me in many ways.
20:27In computing, especially creative com, you know, computing that's
20:30designed for creating things.
20:32And then, you know, mobile basically abstracted files away.
20:36And that's kind of a win for most of the consumer use cases.
20:39but actually is kind of disempowering for creative work that you wanna do.
20:43And now again, the files are, somehow having a moment again.
20:46But now we live in this world of things like real-time collaboration, version
20:50control pull requests, yeah, live collaboration, async collaboration, and
20:54thinking about how all those people pieces fit together or how you can use those
20:58building blocks for the particular piece of software you happen to be building.
21:01That is very interesting to me.
Legal Frameworks and Data Ownership
21:04So another one is legal frameworks.
21:07So we are working with, like, we're building those systems, maybe just
21:11for us, maybe for others, but, like we're not building this in a vacuum.
21:15Sure, I can build some software that might just run like locally
21:19on my, phone or on my computer.
21:22Or like on like a little shared server in my home.
21:25But in that regard, no one cares.
21:28But like once I put it a little bit more out there for the world to see and to use,
21:33or possibly that it's like still meant to be used for a smaller group, but it's
21:37still accessible for others, possibly.
21:40So you need to think about the implications.
21:42You need to possibly lock it down.
21:43You need to constrain it, you need to set expectations.
21:47And one like way of setting expectations or being constrained is through like
21:53our legal system, whatever legislation we're falling into, like Germany, where
21:58we're doing this conference in like.
22:00Very particular set of legal constraints, also like European constraints, et cetera.
22:06But we're not building this in a vacuum.
22:08We need to like fit into the constraints, realities of the real world.
22:13One big question is like, where does data live?
22:16But also of like, other questions?
22:18Well, like what sort of rights do I as a user of software have?
22:22Like, GDPR is certainly something that really changed, the ecosystem
22:27quite a bit for better or for worse?
22:30I think actually one, funny aspect of GDPR, like maybe one intended or
22:35unintended consequence that I'm quite excited about is that GDPR has had a
22:40massive impact on data compatibility.
22:43Where you can now go to like pretty much every service and
22:47through the right to be forgotten.
22:49Like you can still export your data, but that exported data now actually
22:53gives you all the data for that service.
22:56So you can now like go to LinkedIn.
22:57It's like, hey, like give me all my data or like, go to x.com, give me all
23:02my data, and like now you can export that and like build tools with it.
23:08So I think that's like one interesting ways of like how a legal framework
23:12has unlocked data compatibility, even though that like, and not in
23:17a very rigorous way, but at least something better than nothing.
23:20And so that just like as one anecdote of like.
23:23How something has changed and unlocked new opportunities.
23:27And I think there will be a lot more like that.
23:30Like good developments, bad developments.
23:33Another real world analogy might be if you're going to Linear and you
23:37create a team, you need to set like, Hey, where should the data be stored?
23:41Do you want to have it stored in the US?
23:42You don't wanna have it stored in EU.
23:44Like, that's another way to like really model that.
23:47And I think that's just like, scratching the surface of how
23:51we need to bake those realities into the software we're building.
23:55And, I think there, this is an area that that's really worthwhile thinking through
24:01and be hopefully inspired by, new ideas.
24:06Yeah.
24:06I'm weirdly interested in the legal frameworks area and I hope we get
24:09some good submissions on that.
24:11I think it is natural for people in the tech world to dismiss or
24:15downplay that aspect of things.
24:17Especially maybe if you come from the Silicon Valley, American
24:21Maverick laws just get in my way.
24:23I'm gonna ignore them perspective.
24:25but yeah, we're, we're both here in Europe.
24:27I, two, two things that come to mind for me on that, that are
24:30interesting and of this moment.
24:32One is, yeah, you mentioned GDPR and much maligned in some ways.
24:37You know, the cookie banner thing is an unfortunate consequence of
24:42how that all played out, but it also has provided real front guidelines
24:46for handling of person identifying information that I think genuinely has
24:51made the internet a safer place, right?
24:54It's full of threat and attacks and that's only getting worse
24:58now that, that stuff can all be superpowered with language models.
25:01But, you know, when I was working on an AI powered browser called Dia
25:05last year, it was actually genuinely helpful to us to basically go and ask
25:10the question, what is GDPR recommended?
25:12And it basically comes down to, okay, we wanna be able to train on user data
25:17that's incredibly valuable to our users in the same way that, you know, Google's
25:21magic results and auto complete works so well because they're training it on
25:25what other people are searching for.
25:26And that's how it can seem to kind of read your mind.
25:28That's the same thing is true with, Cursor and the auto complete.
25:31So there's huge value in training on user data, even though there's, you know, but,
25:35but there's obviously some, things you need to be really cautious about there.
25:38And GDPR gives actually really good guidance.
25:41It basically says if you strip all information that can tie
25:45it back to a particular user.
25:46Certainly anything personally identifiable, like a name or an email,
25:49but also you're decoupling it from the user ID at the moment of collection
25:53and anonymizing there and you throw it away within 30 days, you're all good.
25:57That's a very simple framework and actually we could do a lot with that.
26:00We could really, go pretty far with training our own,
26:04models with data in that frame.
26:06And it was actually nice to have that, not just from a legal
26:08perspective, but from just a best practices guidance perspective.
26:12so GDPR has been a mixed bag in some ways, but has also I
26:16think provided genuine value.
26:18The other thing I would throw out that is also probably kind of
26:20from the Europe perspective that's top of mind now is sovereignty.
26:24Basically data sovereignty and software sovereignty.
26:27And I think a lot of the European Union as a whole, but also individual countries
26:31are thinking about how to decouple themselves a little bit from US providers.
26:35And I think we can put aside, or we don't wanna make a political conference
26:39or be too tied to whatever current event happens to be in the news at the moment.
26:43But I think in general, this is a good perspective for nations to have
26:47and it is just a scaled up version of local-first data ownership.
26:51It is just a version of the same reason I say it's good for me to have control
26:55over my data and have the ability to make choices about it, not only in
26:59the short term, but in the long term.
27:00Where do I back it up?
27:01Can I delete it, can I duplicate it, et cetera.
27:04There's a version of that for nations.
27:06And I think that's a little bit what, Europe is grappling with right now.
27:09And I think totally separate from anything happened in the current moment.
27:13That's a good thing to think about.
27:15That's a long-term thing that helps create greater agency and autonomy
27:19and security for your nation, just as it does for an individual.
27:23And I think there's also really like a window of opportunity, around that, where
27:27like, whether, what do you wanna call, like, big players like Monopoly or not?
27:32that aside, I think there are now real cases of precedent where I think certain
27:37German states, for example, have made the switch away from Microsoft to like
27:42some, open source software, et cetera.
27:44And like, I think that sets a, like me as someone who has like, very good
27:50feelings towards open source and wants to see open source software succeed,
27:54I think there's new momentum here.
27:56And I think local-first can really also through that perspective,
27:59play a more mainstream role.
28:02Since like you don't just, yes, you can have like your open source react
28:05components, but that does make like fully functional software where you
28:09can replace other software with it.
28:10Yeah.
28:10Like you need to think about data and if you're making that painful
28:14switch from one software stack, that might even like include all the way
28:18to the operating system, you probably like experience a lot of pain there.
28:23And ideally, you now choose a foundation that gives you more
28:27optionality in the future when you might want to do another switch.
28:31And this is where things like, data compatibility, et cetera,
28:35all like play into this.
28:37And where ideally, like legal frameworks set a good foundation and like encourage
28:43things like data compatibility, et cetera, but in a thoughtful way where it does
28:49not, introduce so much overhead and like bureaucracy that it grinds down like the
28:55velocity of like hurricane ship software.
28:57And I think this is like the best of all worlds that
29:00local-first wants to empower that.
29:03Where if you build with like those local-first data layers.
29:06The data layer takes care of that for you, and you can focus on the software.
29:11So I think this is all like a very rich area, to talk about and I think that
29:15what's makes it really interesting is to hear it through the lens of like one
29:20particular story, whether it is like a company having made that switch, whether
29:26you've built software and like you've made your software more compatible and
29:30interoperable in some way, or, yeah, there's like all of those stories that
29:36I've heard over the years and I think those deserve a broader, audience.
29:40And so if you have something to share in that regard, please definitely take
29:44the step and like, submit it to the CFP.
29:46That's a great point.
29:47It almost makes me wanna make data portability or something to that regard.
29:52a top level topic.
29:53But I think this comes back to the principle in local-first,
29:56which we call the Long Now.
29:57Which is, yeah, once you've been around for a while, you
30:00can see the gray in my beard.
30:01I've been working with computers for many decades now, and I've
30:05seen many products come and go.
30:06Software platforms come and go.
30:08But being able to preserve that data for a variety of reasons.
30:11'cause it's important for your company.
30:13'cause it's important for you personally.
30:14There's emotional meaning and saving your old family photos or the, master thesis
30:18he wrote in university or whatever.
30:20And then of course it is natural that especially the startup industry,
30:24which skews young and people are building products and moving fast
30:26and thinking about tomorrow and next quarter and not 10 years from now.
30:30That's, that's fine.
30:31and well and good, but maybe that is a role.
30:34Governments can play a little bit, but it's also a role that, yeah, maybe our
30:37community can play, which is encourage people to think a little bit about that
30:41long term, both as users, what's going to make my data be a little safer, give me
30:46that long term optionality, as you said.
30:48but also product makers to think in terms of like, okay, may maybe
30:52my product won't be hero, will have changed a lot in five or 10 years.
30:55Am I empowering my users, to take their data and go to where they need
30:59to go when the time comes for that?
31:01And I think it's just like a wonderful way to think about the world a bit more, like
31:05positive sum, where I think we've, like in some areas of software, we've already
31:10figured us out, like in a pretty good way when we have like languages and like
31:14ideally like typed languages, et cetera.
31:16Like you have an API that you can compatible with and then
31:19you can like reuse modules.
31:21We even got that all the way to like UI, et cetera.
31:25But for data, this is where we're struggling still the most.
31:27And I think this is like one of the core themes around local-first where
31:32there has been tremendous progress to, to make, progress in that regard.
31:36And in that theme of like user agency, et cetera, personal software,
31:40I think this will play an even bigger role, where we are gonna
31:44have like more software than ever.
31:46Whether it's like we've built this ourselves, whether it's agents
31:48have built it, and then those things can, can talk to each other.
31:53But I think it's also like changing in an interesting way where the
31:56contract between that does not have to be like absolutely perfect anymore.
32:01But like now AI can also like, shoulder a bit of that burden of like talking
32:07to each other, where now we can like get away with having fuzzy software.
32:11Where it has like CSV over here and like a JSON API over there, and the
32:16agent can like very easily either ad hoc one-off, like convert something or
32:21create a script that talks to each other.
32:23So I think we're like, we're getting closer and closer having like some
32:28bridges and ideally over time we'll get to like really having very reliable
32:33bridges that just, in our minds when we hear a story like over there and
32:38like we write it down over there, like we are a bridge and just works.
32:42And ideally software can be like that in the future.
32:45I like that a lot and I like that point that software is becoming
32:49cheaper to make, so therefore it will be more ephemeral, more throwaway.
32:53You'll be more likely to make something that's purpose-built for
32:56a, a shorter period of time, whether it's an internal tool at a company
32:59or a personal app or something.
33:01Something like that.
33:02And in the past we've really relied, or to date you could say we've
33:05really relied on, my Gmail is still readable 20 years ago because Gmail
33:10still exists and is well maintained.
33:12but instead, maybe if we have a system that is more based on, well, I can
33:16use an email tool that's a little bit bespoke and gets stops being relevant
33:21or stops being maintained two years from now, and it's no big deal because
33:24getting my email out or whatever data came from that tool and into the next
33:28thing that I'm going to use is pretty straightforward to do from through some
33:32combination of export data portability.
33:34But also, as you said, that importing and ingesting data and making data
33:38coherent with the help of AI is easier.
33:41So I like that idea of a lot.
33:43This, the software is kind of ephemeral and fits the need you have in the exact
33:47moment, and then you can move on from it.
33:49The data is something that has a longer lifespan.
33:52It can move with you from, product to product.
33:56Exactly.
33:57Great.
33:57So this was local-first maturity, and we've already teased a little
34:01bit of like some adjacent topics, but the, the other, big pillars here,
34:05as Elene put it, is like the new territories and the larger ecosystem.
34:10So maybe if we spend a bit of time on the new territories where we have like,
34:14designing for agency, then around like AI, local and open weight models, and
Designing for Agency
34:23Maybe starting with the first one, design for agency.
34:26What should people think about here?
34:27Yeah.
34:28For me, this would come to the topic of, I like computers as a tool for thought,
34:34a way to augment, thinking, a way to make humans more creative, more capable,
34:39able to accomplish our goals better.
34:41I sometimes think, especially as consumer products have have come to
34:45dominate, for example, just sort of usage and revenue in the computing world.
34:49There ends up being this weird race to the bottom of, hey,
34:52the computer will think for me.
34:54that's what we like about AI.
34:55Good.
34:55I don't need to think anymore.
34:56The computer can do it for me.
34:57And that to me is a very sad perspective.
35:01And rather I would think, wanna think that, AI or other capabilities
35:05that we continue to develop as, we continue to explore the computing
35:08space are things that help us do more and better the things we enjoy
35:14doing, being creative and thinking.
35:16and so I think the design for agency topic is the idea of, okay.
35:20So if you're doing something with language models, with computer vision,
35:24with agents, whatever it is, what can be part of your design, in the sense of how
35:30your product works, that really engages and creates more agency and more deeper
35:36understanding from your users, right?
35:38The negative scenario or the people often quote is that Star Trek
35:41next generation episode, right?
35:43Where people have this, these omniscient computers that do everything and
35:46then they lose all the capability to understand not only just how to
35:50fix the machines, but just really how the world around them works.
35:53And then when those computers start to fail, they're suddenly
35:55like helpless children.
35:57And I think for me, good computing tools of all sorts, and that includes everything
36:02to do with AI and language models deepens our understanding of the world, deepens
36:06our understanding of the technology by freeing us from some of the fussy details
36:10of where exactly does the semicolon go, and allowing me to spend more time on the
36:15mental models and the frameworks and the principles and how things fit together and
36:20how we use those to accomplish our goals.
36:23Yeah.
36:24And I think this is, I mean, for, for me, AI has been like one of the biggest
36:29unlocks to like just learn new things.
36:32I've gotten into so many new areas, like I've gotten deeper on like
36:36some hardware related things.
36:38I've gotten deeper on understanding how wifi works,
36:41like band channels, et et cetera.
36:43Like gave me a deeper and a broader understanding of the world, like
36:48I have been on an accelerated pace of, learning and understanding.
36:52and I think that's like the positive sum, version of that or like the
36:56flip side or the other, scenario, of an like the utopia maybe of that
37:02Star Trek, anecdote you shared.
37:05So, one other much more concrete topic that I think has been, a very
37:10new one that's come, about over the last few years as, like off the
37:15shelf models became available at all.
37:18And now we also have like, open weight models and they're even small enough
37:23that you can run them on your own device, whether it's on a very beefy, desktop
37:29grade machine or whether it's like, scaled down models that might even work in the
37:34browser, that might even work on like a phone or other much lower end devices.
37:40I think that opens a whole new category of software and use cases, et cetera, whether
37:47it is, like that you built a little system that relies on like detecting
37:52certain shapes to do a certain thing.
37:54Maybe you're building, like maybe you have chicken at home and you wanna like
37:58automate a little gate that like, opens when a chicken walks through or whether
38:03you have like some other use cases.
38:05But this is you ideally wanna do that like reliably, locally, efficiently, et cetera.
38:10And there's so many, like we can, we can just like dream so much bigger now
38:14of like what we can build in a way that does not rely on like Anthropic currently
38:20not having an outage and us like paying whatever, a lot of money each time.
38:25We are like using a bit of like API quota and, I think this is
38:29gonna be like right now, this is.
38:31Probably similar to the early days of like computers being room
38:35sized, and not very capable yet.
38:38And I think that's just gonna, like at some point our little like devices like
38:42this, will be able to do like really, really impressive, inference capabilities.
38:48And I think this is like, now is the time to start thinking about
38:52that and start already embracing that in all sorts of different
38:56scenarios, form factors, et cetera.
38:59So I think this area could be about like if you're building
39:02infrastructure, building tool, like maybe you're the company behind some
39:06of those open models and you wanna show people what's already possible.
39:10I think that's really great.
39:12if you are building applications, systems, et cetera, that already
39:16use, open models, local models.
39:19that is like a story we'd love to hear.
39:21We had last year, Thomas from Google already share, some earlier versions
39:26of like how you can run, Gemini Nano in Chrome and showed some really cool demos.
39:32And all of this has come a really long way and, I think across all sorts of
39:37different devices, platforms, et cetera.
39:40So if you're doing cool things in that regard, please definitely apply to speak.
39:45Yeah, I think so much attention is on the frontier models and the
39:49sort of one giant chat bot that can do everything, multimodal models,
39:53ChatGPT, and Claude and so forth.
39:55And those are of course impressive and very useful and there's a lot to that.
40:00But I really have, opened up to the value of things like
40:02lightweight classifier models.
40:05Yeah, embeddings, including, there's plenty of text embedding, models that
40:08you can just run in a Python server with PyTorch that are great, really
40:12almost as good as basically as good as what you can get in the big providers.
40:16Things like computer vision and OCR have actually been good for a pretty long time.
40:20I think in some ways, you know, maybe myself, I almost feel a little,
40:23surprised, you know, something about the AI hype, you know, around chatbots
40:28and large language models then caused me to take a closer look at a lot of
40:31these other ML technologies that have been getting good for a long time.
40:35A lot of the text embedding stuff and semantic search and things like
40:38that have been, you know, really had their breakthroughs 10 years ago.
40:42Computer vision has also been very good for a long time.
40:44We know that because captures keep getting harder and harder.
40:47And you can do a lot with these, in those very specialized cases, like what you
40:51were describing with, yeah, whether it's a robot in your house or, something like
40:54that, but also software where you can have the model run on each keystroke, right?
40:58You can do all kinds of things.
41:00They're very different versus needing to wait several seconds for
41:04a heavy inference to come back from the big language model provider.
41:07And just broadly, the story of computing is one of taking these
41:11computing things that are originally are heavy and slow and as you said,
41:15require a room size computer and making them smaller and cheaper and faster.
41:19And that just opens up all kinds of new use cases.
41:22And there's really great open models out there from Quinn
41:24to Llama to many, many others.
41:26And yeah, I think if you're working either on open weight models, or open
41:30models of any kind or the libraries behind it, like Transformer.js or
41:34PyTorch or something like that, we'd love to hear a talk from you.
AI Assistance for Local First
41:37Awesome.
41:38So the next topic in the new territories, area is AI assistance for local-first.
41:45This is, I guess more around like the application of those new materials
41:50to those new building blocks we got.
41:52And I think that can be really, like wide ranging, can be going in
41:56all sorts of different directions.
41:58Like, and I think the exciting part is actually the stuff that we can't
42:01even, like, think about right now, but that maybe some of you are,
42:05actively exploring and have built like surprising things on top of this.
42:10But just some things that came, to mind for me as I've been like,
42:14experimenting myself or thinking this through, like just to, generate ideas.
42:19How can AI address like some of those open problems in the local-first
42:24space or like some challenges that are so like kind of hard to crack
42:29and if we're allowing ourselves.
42:31To go beyond the realm of like deterministic, traditional software.
42:36And like, we're using those fuzzy materials a bit more.
42:39And like we're thinking of like, hey, if we're actually, embedding
42:42them into our local software stack, what becomes possible through this?
42:47So like, just one way, for example is let's say you're building a power tool and
42:52like you have a very, very sophisticated set of like system settings, et cetera.
42:57Like for me, for example, the Apple system settings come to mind and
43:01they're really like, I don't think they're the pinnacle of modern software.
43:05And, I always need to use search.
43:07And like five different things come up and then it's like the 15th one
43:11that I was actually looking for.
43:13And like, it's just like very painful.
43:16Now where we are so spoiled by like using things like ChatGPT et cetera,
43:22like it really feels like worlds a part of like how good something could be.
43:27So why not use something like a, ChatGPT chat interface to interact with your
43:34own app that like, you can, instead of have to click the clock settings,
43:39like why not allow your entire app to begin configurable through AI models.
43:45And so to make that actually happen now you need to bring in like that
43:49fuzzy LM piece into your software.
43:52But like exploring that, like figuring out best practices, figuring out
43:56like how can you actually, constrain that, that you don't, give the, users
44:01so much, power that they can like accidentally tear apart the app and
44:06it's no longer functional, et cetera.
44:08I think there's a lot to be explored there.
44:11So just as one area.
44:13Another one is to give even more power to the user or build systems in a way
44:18that wasn't really feasible before.
44:20We talked about the data interoperability before data exports, et cetera.
44:25Now we can actually build like fuzzy bridges between different
44:29software that were never meant to be talking to each other.
44:32Whether it's like in real time or whether it's just like making
44:36sort of, export interfaces a little bit more ergonomic to use.
44:41And now you could imagine whatever, like using your, local supermarket
44:46that might have like a certain, online shopping, like not even API,
44:51but just like a maybe connecting that with your own home build to-do app.
44:57And, there is not really like the end all be all integration API, yet
45:02you can make those things happen now.
45:03So this is just like one of many, many ideas where hopefully you all have
45:08been exploring some interesting use cases that we'd love to hear about.
45:12So this is what comes to mind for me about AI assistance for local-first.
45:16Yeah, using these new technologies to make the world more local-first,
45:21especially maybe parts that have been resistant to that.
45:23I think breaking down data silos is one of the biggest ones that's been such a long
45:27running topic at you could switch with projects like Cambria, but almost always
45:31it involves kind of a convince everyone to build their software on a completely
45:36different stack, which sort of Yeah.
45:40Is very much an uphill battle.
45:41Right.
45:42And as you point out, there's a great agency in our personal lives.
45:45I've experienced this already with, yeah, being able to pull out, I don't know.
45:49You know, there's some website that has a list of things and I want
45:52to put that into a spreadsheet, but it doesn't offer a CSV export.
45:56Okay.
45:56I basically could do that with a screenshot or a PDF and let the language
45:59model do the hard work of, you know, turning it into something coherent.
46:04Similar idea there is just downloading.
46:06I like to download my transactions from my bank account on a quarterly basis, so I
46:10have a record of that that's separate from whatever clunky interface the bank has.
46:15But inevitably, if you have a couple of accounts and then they're all in slightly
46:19different formats and trying to like bring them all together into any kind
46:22of unified ledger is nearly impossible.
46:24It's just not worth the bother for an individual or whatever.
46:28But actually now with language model, it's pretty easy to do an ad hoc thing where I
46:32can go do the download effort myself once a quarter, but then pull that all into a
46:36spreadsheet with a normalized data set.
46:39That feels very empowering for me.
46:40It's like, okay, now I have all my transactions in a place that I control
46:44in a format that I want that isn't dependent on, you know, my bank letting
46:48me, you know, I go to login to download something and it tells me, oh, sorry,
46:52we only support transactions, you know, back six months or whatever thing, right?
46:57Very disempowering user experience that instead it's something that data
47:02that feels like it belongs to me, my banking and transaction data, and I
47:06have it in a place and a format that I can control for the long term.
47:10Exactly.
47:11it's just so cool to see like how far people take, like how much they embrace
47:16the new forms of like, agencies that they, I guess they, they've been always like
47:20longing for, but now, like it becomes feasible to, like really go for it.
47:25And I think it's so, also so liberating that you can build software for
47:29yourself unburdened by like making it enterprise grade, et cetera.
47:34And so you can like, you know, I need to hold it this way and it's
47:38gonna work, otherwise it falls apart.
47:39But that's fine.
47:40And I think that's like the, beauty of like building tools for yourself
47:45and like local-first gives us the glue that they actually can bear like
47:49real weight and like work together.
47:52So that is the new territories, which I think is very related to the larger
47:58ecosystem where we have, a set of like other, very interesting topics.
48:02So like some of them we've teased before, but just to run through like on a high
48:06level and then we can go into them like again, we have like self-sovereign
48:10identity, we have social media and data ownership, encrypted messaging,
48:15game development with local-first and malleable software in the age of LLMs.
48:19So maybe starting with the self-sovereign identity, what
48:23should people think about here?
48:25Which sort of stories might we be interested in having at the conference?
48:30Yeah.
48:31Well in general, this category of the larger ecosystem is one I'm excited
48:34about because I think there are so many adjacent communities that are
48:37interested in, yeah, for example, things like freedom, you know, the open source
48:41community, obviously the Linux world.
48:44The Open Web and open standards, as well as things like the end-to-end encryption,
48:48kind of privacy, nerds out there.
48:51You know, there's, that's not exactly what local-first is
48:53about, but it's very adjacent.
48:55There's a lot of shared technologies and things, and this is an invitation
48:58for folks in those communities.
48:59And I'm even sending out some emails and so forth to folks to try to invite them in
49:04to say, look, let's see where we overlap.
49:06Let's, find the edges there.
49:08yeah.
49:09The identity topic though, this is a big one for me.
49:11It's something I've been fussing about my whole career.
49:14I have whole, article about it in my making computers better
49:17essay, from a few years ago.
49:19But it basically boils down to, I think the way we identify ourselves to
49:22computers is kind of an embarrassment, usernames and passwords, identity theft.
49:27Now you've got, I don't know, magic login links and login with Google
49:30passkeys and whatever, and it's just endlessly confusing for most
49:35users and even for power users.
49:38Yeah, the amount of time in my day, I end up spite typing in two factor
49:41codes and digging magic links out of emails and just trying to convince
49:44the computer that I am who I say I am.
49:47And I know why it's that way.
49:48Of course, as a professional that's worked in this field and particularly
49:51in the security, space, on and off, over the course of my career.
49:54But it really is a shame to me.
49:57And then in the meantime, you have this more and more control and
50:00more and more ownership over our identities that go to the companies.
50:04And basically it's been a kind of a good deal to say like, look, I'll
50:07just let Google manage my identity and then I can click the login with Google
50:10button everywhere, because it is such a hard problem and I can just kind
50:14of outsource that to them, sort of.
50:17But there, it's kind of like a little bit of a deal with the devil there
50:19comes with a lot of trade offs.
50:21and local-first is a chance to, rethink that, reinvent that.
50:25and of course there are also a lot of adjacent communities.
50:27Things like Open ID for example.
50:29We'll talk a little bit more about, app proto and ActivityPub in a moment.
50:33But they have their own concepts of identity and how you give people ownership
50:36over your online handle, in a way that's sort of secure and safe, but doesn't
50:40put the user in the position of needing to manage a bunch of cryptography keys,
50:44which they're inevitably going to lose.
50:46so I think anyone working on anything in this space, whether it feels
50:50local-firsty or not, if you're working with passkeys, if you're working
50:52with, any kind of identity system, yeah, we'd love to hear from you.
50:56Yeah, I think this is a highly.
50:59unsolved area.
51:00I wouldn't say under explored.
51:02I think, people are really exploring it.
51:04It's a hard problem in really interesting ways.
51:06It is a hard problem and I think we've, we've also come a long way already.
51:11Like for example, like if you're depending on, which pact with the
51:14devil you've made, you can already have a slightly more convenient life.
51:18So me being, in pact with the Apple ecosystem, I am very thankful for things
51:24like Touch Id, et cetera, that are then like nicely working together with things
51:29like 1Password, Passkeys, et cetera.
51:31and like being able to understand how those things work and like thinking
51:36through the implications, et cetera.
51:38Like let's me be for, for this point in time, in a relative sweet
51:41spot between, like living a secure life, but also having convenience.
51:46But there's still like great levels of like lock in, et cetera.
51:50And I think this is also like all of that aside, like we're now also,
51:55having this like phase shift in regards to software where we are no
51:58longer the only ones who are doing things on our behalf, with computers.
52:03Like if we're, now whatever, like maybe ordering, like some food or
52:08we're doing some other things, maybe we're, like scheduling something on,
52:13the calendar or whatever it might be.
52:16It's no longer just us, but there's like another thing that needs
52:20identity and even more fine grains.
52:22And I think one aspect here as well is like if you're going from beyond
52:26authentication to authorization, so far we have like, all off scopes, et cetera,
52:32but like it's very all or nothing.
52:34And it's all like, okay, you, now get access forever.
52:38But maybe for like those agentic use cases, you might want to think about it
52:43like, okay, you get to do this like one time, or you get to do this like for five
52:48minutes, or you get to do this like only for this email address or for this scope.
52:52So I think we want to make this like much more fine-grained, but in a way
52:57where you like, it's, it's frankly like going through like the permission
53:02granting on like a Google OAuth flow is already overwhelming for most people.
53:07And now making this even more fine granular
53:09Yeah,
53:09I think is like really
53:11tough challenge.
53:11Thinking of a
53:11giant
53:12array of check
53:12check
53:12boxes on the GitHub, token generation and there's a reason for it.
53:16But then yes, again, that trade off of, convenience and security is ever present.
53:22I think it's that intersection of like user experience
53:26design and also cryptography.
53:28Like how do you make it actually secure?
53:30and I think that to, come together in the best of both worlds.
53:35I think that is like a area that's like highly deserving of like
53:39more resources, more investment, more smart people working on it.
53:43So if you're working on that, we'd love to hear from you.
53:46ideally where you cover quite a bit of that intersection.
53:50I think this is probably not the best, conference to share
53:53like your new breakthrough of like some cryptographic nuance.
53:57But I think where we can like bring those two things together where you can also
54:01with the focus of like, or one aspect can be the solve sovereign, aspect.
54:06I think it doesn't have to be a prerequisite, but I think if
54:09something points in that direction.
54:12And also I think another, tricky part here is like that we need to kind of
54:18create a bridge from like today, where people are like, defacto using things
54:24like login with Google, et cetera, because it strikes A, acceptable balance
54:29of like convenience and security.
54:31And if there's like a much to esoteric, version out there that doesn't have a
54:36credible path to becoming mainstream, I think that also is something that
54:40needs to be taken into consideration.
54:42So this is a very nuanced topic and I think as one of
54:46the harder nuts to crack here.
54:48but if you're working on that, we'd love to hear from you.
54:51And I think with that it is like a natural transition to, things like
54:56BlueSky ATproto where we talk about, social media and data ownership.
55:01So you've already mentioned initially.
55:04That Martin Kleppmann is affiliated with the ATproto ecosystem, and I think that's
55:10just like one example for a broader, set of design challenges and like goals
55:16that I think is like nicely adjacent or partially overlapping with local-first.
Self-Sovereign Identity and Data Ownership
55:24Yeah, for sure.
55:25I think social media and data ownership is so interesting because social
55:28media is the classic centralization and the classic opaque appliance
55:34with a mysterious algorithm that you can't and don't know how it works.
55:38And in general, I would say we could, obviously could, and indeed I have
55:44in the past done whole podcasts on social media and the both the
55:47positive and negatives that that has.
55:48Impacts that has had on society and us as individuals, but narrowing in
55:53a little more on this kind of, yeah, user empowerment around, and data
55:57ownership in our computing lives.
56:00It is the nature of the classics like an Instagram or a Facebook or even a
56:04Twitter back in its heyday that, you know, it, it is a very centralized
56:07closed, you have little ownership.
56:09You can just participate in this stream of, content.
56:13and now we have this new crop of, creations.
56:17I think Mastodon and ActivityPub was, kind of early out the gate,
56:21but now there's some really good energy around, BlueSky and ATproto.
56:24Actually, I think even as right as we record this, the ATmosphere
56:27Conferences is happening.
56:29which yeah, just a lot of good energy in there and it's
56:33a different set of problems.
56:34I think the classic local-first app is something more like Linear
56:39or Notion or Figma, which is a document editor, heavy creative work.
56:43You're collaborating with a small team.
56:45You really care a lot about the output of that data.
56:48Social media where it's sort of global town square, little fragments of
56:51data, everyone gets merged into a big stream is on the surface, a completely
56:56different, both technical and kind of design or user experience, challenge.
57:01But at the same time, when you look at what ActivityPub and ATproto have
57:04been grappling with, including things like personal data servers and how
57:07you have ownership over your identity, how you make your handle portable,
57:10but still kind of globally unique and especially verifiable, trustable, right?
57:14not easy to impersonate that you're the New York Times or a
57:17politician or something like that.
57:19So there's some really interesting technical and design work that has
57:22been done in both of those communities.
57:24I think we already have, some folks lined up to speak on some aspects of those.
57:28But I think there's, that's such a rich area that if you're working on something
57:31in this space with either one of those communities or otherwise in the kind
57:35of broader, how do we make social media something more empowering for users,
57:40yeah, we'd love to hear that, we'd love to see a, talk submission from you.
57:43Yeah.
57:43In particular also like using those protocols, those ideas to go beyond
57:49social media where you can basically like to, draw that bridge to the
57:53previous, like, identity question.
57:55Like right now I think this is, we we're blind to it at this point, that
57:58when there's new software, like, okay, of course I need to create an account.
58:03And like, depending on, which sort of patterns you use, maybe
58:07you automatically like, look for that sign in with Google button.
58:11Or if it's a developer oriented thing, like sign in with, GitHub, I'm more in
58:15the camp of like, where I always try to like just, create a standalone account
58:19just in case, something were to happen.
58:22but does it have to be this way?
58:24Are there other ways and like, and not just for.
58:29social media use cases, but for something that spans a much, wider range of
58:34scenarios, whether that is like something, I don't know, like maybe there's something
58:39in between where it's something in between Figma, Linear and, and social
58:44media, whether it's like a, smaller scale community that you are part of where
58:49you do some very interesting things.
58:51Maybe a community for a school that you volunteer in, or you have a much
58:57richer set of different, kinds of media you deal with, whether it's documents,
59:02whether it's chats, whether it's like some, posts, some forum like things.
59:07Then you also have like things like identity.
59:09And I think this is just like an example for thinking about
59:12this a little bit more broadly.
59:14So if you're using, some of those technologies or exploring that.
59:19And see how we can bridge those worlds.
59:21we'd love to hear from you and, going from one, one, tricky technical,
59:26nuanced topic to another one.
59:28Another like evergreen in local-first and beyond is like
59:33encryption and encrypted messaging.
59:36So, I think highly related to the areas that we talked about before.
59:41And I think is another, aspect of software.
59:44I think for a lot of software, kinda like still an optional part because it's hard,
59:49but ideally over time it becomes just like a no-brainer to have to have encryption,
59:54for your, data, for your applications.
59:57It is, I think also dependent on the kind of like app you're working with.
1:00:02So for example, if you're working in the finance space or the
1:00:05healthcare space, I think it's much more, table stakes already.
1:00:09And then there are other areas where it's much less common.
1:00:13but ideally this is a quality that does not, quadruple the effort it takes
1:00:18to build the app, but it's some just something, a data library, for example,
1:00:23like Automerge takes care of for you.
1:00:26And so yeah, those are like some preliminary thoughts
1:00:29on this topic from, my end.
1:00:31But yeah, curious to hear yours, Adam.
1:00:33I mean, end to end encryption is first of all just such a powerful technology.
1:00:39You know, I think wars have been one with, it is one frame you could use yet remains
1:00:45hard to put into practice in a way that is approachable for users because, you
1:00:49might think of it, everything boils down to a key management problem, but managing
1:00:53keys is really difficult and incredibly abstract, even for very power users.
1:00:58And certainly for a more mainstream computer user.
1:01:02Yet you see all these places where encryption, when it has been brought to
1:01:05bear effectively in the computing world, it is a complete game changer, I think
1:01:08of like SSH for connecting to servers versus Telnet, which came before it.
1:01:13I think of HTTPS for website encryption has, I think, literally brought trillions
1:01:17about dollars of value into the economy through enabling things like e-commerce
1:01:21and online, stock brokerages and so on.
1:01:25Something like password managers and how they can provide a secure, setup,
1:01:30based on their, encryption arrangement.
1:01:32And then more recently you've got, for example, encrypted messaging like Signal.
1:01:36And I think when we find ways to bring those to bear in a way that
1:01:41is approachable for end users, it can just unlock so much value.
1:01:45But it remains the case that there are a lot of areas where we haven't really
1:01:49effectively brought it to bear fully.
1:01:51And yeah, if you're working on any of these, whether it's something
1:01:55relatively well established like encrypted messaging or something a
1:01:58little more, on the cutting edge.
1:01:59Like last year we heard a, an amazing talk from Brooklyn Zelenka about
1:02:03Keyhive, which is a kind of adding an identity encryption layer to Automerge.
1:02:08This just remains an area that is very hard, but I think again, just has the
Game Development and Local-First
1:02:15Awesome.
1:02:16so another adjacent topic, a little bit more lighthearted, and yet probably one
1:02:21of the most, OG adopters of local-first, and they've always built software
1:02:27this way, which is game development.
1:02:30I think most games, particularly in the era of like internet not
1:02:37being as ubiquitous as it is today, have like, out of necessity.
1:02:42They needed to be local-first.
1:02:44So you can play single player campaigns or whatever, but also like
1:02:49in that gold, era you had, a lot of games also having multiplayer modes.
1:02:54And this is where you had like LAN parties, et cetera.
1:02:56And like, software was just like built in a different way where like internet
1:03:01a lot of games, even didn't like support it or like was one, feature
1:03:05that was added a little bit later.
1:03:07And it's just like a.
1:03:09I think an interesting case study for how like a different parallel universe evolved
1:03:15and like did very complex data management with a lot of like similar challenges
1:03:19where you needed to have like, reactivity you needed to be like, very efficient.
1:03:23You needed to like care about persistence.
1:03:27so I think there's like a lot of parallel universe lessons learned
1:03:30that maybe never really, fully percolated into the web ecosystem.
1:03:35Probably sometimes for, the worst.
1:03:37I think I always feel, I can learn a lot from the game ecosystem overall, as
1:03:44a lot of like lessons have been already much, earlier been learned there.
1:03:48And I think now it's like that, that is still true today.
1:03:52And video games have always been like the, one of the most demanding use cases
1:03:57and where like latency really matters, where like all of those things, matter.
1:04:02And, I think it's just about time that we bridge the world from game
1:04:07development to the local-first ecosystem.
1:04:10So if you're, doing things in the game development space and you have
1:04:15some interesting stories to share about how your game has always
1:04:19been local-first, then, please definitely consider submitting a talk.
1:04:23So that, that's kind of my, my perspective on it.
1:04:26But I think you have even more nuanced takes on this, Adam.
1:04:29Well, yeah.
1:04:30I don't know about nuance, but it's definitely close to my heart.
1:04:32I actually began my career in video games.
1:04:34I probably like a lot of little kids, I basically got inspired to
1:04:38learn to program, to make games.
1:04:40And, began my career there.
1:04:41I worked for four or five years in the game industry, including on
1:04:44PlayStation 2 games and Dreamcast games.
1:04:46And that was kind of the era then.
1:04:48And it's an incredible world of technology that's very separate from the productivity
1:04:53software, web operating system, thing where you're just drawing 60 frames or
1:04:58120 frames per second on a blank canvas.
1:05:02The APIs to the underlying system are just a very, very different thing
1:05:06than perhaps what we're used to, in some ways easier to work with, in some
1:05:08ways harder, but yeah, the performance and being able to use it wherever.
1:05:14even certainly in the age before internet connection was ubiquitous,
1:05:17but even nowadays, yeah, you expect if you have a game on your device, that
1:05:21you should be able to play it on the plane and not be worried about whether
1:05:23you have internet in that moment.
1:05:25Even think something like, you know, even once you start to add some
1:05:27internet connectivity, Hey, I wanna post my high score to a global thing,
1:05:31but you don't expect the game will just stop working the moment that the
1:05:34high score API is not available, so.
1:05:37And then you have so much going on in the world of these different engines like
1:05:41Unreal and Unity and Gadot engine slash game development tools and yeah, I just
1:05:47feel like there's, maybe a lot these communities can learn from each other
1:05:51and yeah, if you're someone in that, tough space, we'd like to hear from you.
1:05:54Yeah, I think there are just so many amazing stories waiting to be heard in
1:05:59the local-first ecosystem, whether it's someone who like struggled to scale
1:06:04World of Warcraft, to like massively multiplayer scenarios where I mean,
1:06:09it's already hard to build multiplayer.
1:06:12like systems for 10 users, but now, like, think about this, like in the, tens of
1:06:17thousands of users, in like, in much harsher, real time scenarios, et cetera.
1:06:22So, or whether you're building like, more modern games that, they're currently
1:06:27like scaling things even further.
1:06:29so yeah, we'd love to hear from you.
1:06:31if you think this is interesting for the local-first community and as the
1:06:35last topic, in the list, we, we probably like the, kinda like the key topic that
1:06:41is the, center for us this year is like malleable software in the age of LLMs or.
1:06:47malleable software like at all.
1:06:49But then also in the age of LLMs, that's new.
1:06:52That was not really like the, we got like the early glimpses of that last
1:06:56year, but at last year's conference, but I think it was still so early, I don't
1:07:00think like Claude Code was released yet.
1:07:03ChatGPT was out there, of course, but I think it was not yet at a
1:07:07point where it like really clicked for everyone that this is real and
1:07:12not just like a niche use case.
1:07:14And I think now it's like, already so ubiquitously available that we can
1:07:19make this a part of our application.
1:07:22And I think there's like really one of the broadest topics here, almost like a
1:07:25catchall for a lot of, different, talks that could fit into the conference.
1:07:30the anchoring point here is like, uh, what we've talked about before with
1:07:34like Maggie's, closing keynote from 2024 was it, with like barefoot developers.
1:07:42And I think this can, cast a really wide net.
1:07:46And, what are your thoughts on, this topic?
1:07:49Yeah, we'll probably first just defining malleable software.
1:07:51For those not familiar, this is the idea that software systems are ones
1:07:56that could be modified by their users as we're using it or to adapt to our needs.
1:08:01and of course, Ink & Switch has a, I dunno, 8,000 word essay up on the website
1:08:06if you search for malleable software, that can dive into detail on that.
1:08:10But, you know, how does language model assisted coding it seems to enable so
1:08:14much the ability to create new software and especially for people who are on the,
1:08:18not necessarily professional software developers, even if they're technical and
1:08:22good at thinking about software systems.
1:08:24But when you think of the classic, let's call it vibe coded app, it's
1:08:29let me make a one-off website that I can post, but that's very different
1:08:32from being able to modify the tools that I'm relying on in my daily life.
1:08:37My text editor, my email client, my web browser, et cetera.
1:08:41so I think on one hand, language model coding should, in theory,
1:08:45really open up some major new possibilities for malleable software.
1:08:51On the other hand, I think we have some of the same old problems we have always had
1:08:53around data siloing around kind of closed ecosystems, closed appliance heavily.
1:09:00You know, the control, the locus of control lies completely with the software
1:09:04vendors and very little with the users.
1:09:07and how has this world evolved or what new possibilities are opened by the
1:09:11existence of language model assisted coding or other AI things or other things
1:09:16that have changed in the technology world since we last convened in 2025?
Malleable Software in the Age of LLMs
1:09:21Yeah, I think this is, I mean, this topic in itself has been sort of like a core
1:09:26pillar for this year's conference, and I think we've like added quite a lot other
1:09:31ones just because, it felt like the right moment to, to broaden the net a bit.
1:09:37But I think here, like malleable software, this was, a theme that I
1:09:41mean, is is one of the key tracks of, the Ink & Switch, family.
1:09:45But I think now with AI models, et cetera, it's really something
1:09:50that's becoming mainstream.
1:09:51And where before I think it was like very something that developers
1:09:56maybe had romantic thoughts about.
1:09:58Like, oh yeah, this was like, also on my list of side projects I wanted to build.
1:10:03And now you can actually build it for yourself.
1:10:06Or maybe you built software for like someone in your family, or it's
1:10:10just like a, almost as an act of kindness of like supporting someone
1:10:14like this is through your abilities.
1:10:16And now it becomes so easy.
1:10:19I think now that will lead to so much, creation of, new software that will also
1:10:25have like interesting new consequences.
1:10:28whether it's that we need to wrestle with, the quality of the software,
1:10:31or the interoperability of it.
1:10:33And also like, yeah, how much control do we give to, users?
1:10:37How much control, do we, like where do we draw the line between,
1:10:42a developer building the software versus a user using the software?
1:10:47Like a lot of power tools are very customizable.
1:10:50But now I think that line gets more and more blurry.
1:10:53We'd love to hear from people who have been wrestling with, those
1:10:57blurry lines and trying out different ideas and sharing your learnings.
1:11:02What sort of software you've been building as a barefoot developer.
1:11:05Yeah.
1:11:06I think this is, quite the portfolio of, different themes and
1:11:10different topics that we're gonna have for, this year's conference.
1:11:15That's right.
1:11:15We maybe even run the risk of being too scattershot or too wide
1:11:19ranging, but in some ways I'd rather.
1:11:21take the risk of having too many ideas, too many interesting,
1:11:24intellectual, avenues to pursue, then get stuck in a rut.
1:11:28And now seems like a great time to do it with the changes happening, in the
1:11:32technology world, but in general because of the foundation of success we've had
1:11:36in the event to date and the community we've built around, that just seems
1:11:40like the perfect time to, broaden and bring new people in and fresh ideas.
1:11:44And like all of that is not to say that we ran out of ideas with like the,
1:11:49traditional way how we did the Local-First Conference, the last two years before.
1:11:54But I think what it's always been all about for me and for I'm sure for you as
1:11:58well is like the people who actually come and like the conversations we're having.
1:12:03And it is like no one in that ecosystem has been there only because they
1:12:08only care about the data sync engine.
1:12:10It's always been in service of like some broader vision and like some specific
1:12:16use case that really like, had real impact and real meaning for their life.
1:12:22And like I think all of those different themes we now talked
1:12:24about is like one manifestation, one direction of that and local-first.
1:12:30It's like the foundation of that.
1:12:31It's kinda like what brings us all together.
1:12:34And I think what all of those people who have attended and have contributed to
1:12:37the conferences have, in common is that they have a very, deep understanding of
1:12:42the, technical world and of the world more broadly, but, also a broad interest
1:12:47and, are always keen to be inspired to think a little bit different about,
1:12:53like be beyond their current horizon.
1:12:56And I think this is, what we, try to, also, facilitate through the conference
1:13:01where those, different talks are hopefully little points of inspiration that can
1:13:06lead to further exploration by everyone who's attending and paying interest.
1:13:11Well said.
1:13:13Well, I think we hammered in the, you know, submit a talk call to action.
1:13:16we'll say it one more time.
1:13:17The CFPs open till I think, end of April.
1:13:20So you have a little bit of time, but not too much time.
1:13:23But also if, even if you don't wanna give a talk, buy a ticket,
1:13:26just come listen to all the ideas.
1:13:28we got sponsorship booths available as well.
1:13:31If, the folks at the, an event like this are interesting to, talk
1:13:35to for your, for your company.
1:13:37So we'd love to have you come participate in any of those three forms.
Outro
1:13:42Awesome.
1:13:42Well, Adam, thanks so much for talking this story together.
1:13:46I feel like this is also already Yeah.
1:13:49Given me so much excitement for, what we're gonna hear, at the conference
1:13:54and like in a way where we still don't fully know which stories we'll hear,
1:13:58but we'll know about the broad topics.
1:14:01So that've got me very excited already and I can't wait for, the conference and
1:14:07to meet a lot of familiar and new faces.
1:14:10Same here.
1:14:11I'm excited for the new themes.
1:14:12I'm excited to see old friends, and I'm excited to enjoy beautiful,
1:14:16beautiful Berlin in the summertime.
1:14:21All righty.
1:14:21Thanks so much.
1:14:22Take care.
1:14:23Bye.
1:14:23Bye.