 Hey folks, thanks for coming all here to my talk and being interested in Memexy stuff. I assume some of you may heard about that term before in the past, raise your hands. Okay, so for those of you who don't, Memex is a concept that has been populated or made popular by Vannevar Bush. He was a scientist from mid 20th century and he envisioned the device that would let you organize, connect and share your knowledge. Back then it was a microfilm, obviously not really feasible from a hardware point of view, but now we're in a stage where we can do that and so we started to build what Vannevar wanted to build, hopefully. So who of you knows these problems? Please raise your hands and let them up if you're still in it, okay. Okay, let them up. Who of you knows this problem when trying to save stuff? And who of you knows this problem? Okay, yeah, then you probably like what I'm about to show you. So the thing is, when we're thinking about our knowledge, the way our brain stores information is always contextual to other things. So when you're looking at, for example, that article that you saw while browsing in the back rows of the talk at FUSDEM, you remember some words from that article. You remember the date, you remember it was at an event, but these are not the things we can search for. And that's because the way our software currently is structured doesn't allow the kind of fluid, associative memory to be taken advantage of that we have, right? Like if we think about it, we use 15th century tools to organize the kind of 21st century digital knowledge. So we first had libraries that were very siloed. So you had to go to a library and find a book. Then you had to go to the digital age. You had to go to organizations like Facebook, like your knowledge management, task management, your Google search, everything is disconnected. So it's basically like a library again. But what we think is the future is connected knowledge. And you may have heard also about an application called Notion before. Who of you has used that? One person? It's a really, really great tool because it allows you to connect knowledge and data objects with each other in a very customizable way because it is so connected. And what Notion is for kind of wikis, knowledge management in that regard, we want to do for web research. So let me show you how that works in practice. So now I'm reading this article. This is the article, by the way, about MemEx, the original concept, not ours. And what do I remember? I maybe remember the word recollection, right? So I had no chance to find this again if I just remember that word. I could re-google it maybe, but that takes so much time. So what you can do with MemEx is you type in W, space into the address bar, the browser, the word recollection, and then you have all the results from every website that you visited with that word in it. You can also say, please show me everything after like 20 minutes ago and filter by time. You have also an overview where you can, for example, tag stuff. You can sort it into collections because not needing to organize doesn't mean you can't. Obviously, you need to have some sort of flexibility in terms of how you organize your knowledge. The second issue that the tool tackles is one of note-taking. So we all know whenever we take notes about the web stuff we see we have to copy paste it out. We have to use the web clipper and make a snapshot of the page in Evernote and then write some inline text and you cannot share it. You cannot reuse it. It's locked into this data silo again. So what you can do here is you already saw the little tooltip. You can just highlight any piece of text and add notes to it. And another thing that you might also encounter constantly, specifically if you're in the journalistic or doc writers for open source projects, that you want to show people specific pieces of an article and not the whole thing. So right now you need to copy paste out that specific section and then you need to copy paste again the URL. The other person needs to do a whole shanning again. So what you can do is you can just highlight any piece and click on the second button that you saw. It's maybe too low. I'm putting it up here. So here hopefully the internet works. Yeah. Okay, it worked. So now you get a link that you can share with anyone. And this is what happens. You can really quickly link stuff with just a click. So that's the current state. That's actually my production version. That's my private memex. You can download that already. You go on this website, if the internet loads, worldbrain.io. If you want to do that, I repeat it later again, if you like. So really important. This tool is obviously because I'm here at FOSTA entirely open source and also we value data ownership and privacy a lot. So we designed this application to be offline first. And it runs by default all on your computer. No data is sent to any server except you want to share stuff. Then there obviously other people get specific access to specific sub data. For the linking, we... Hello? Cable? I'll try it again. How to take a moment. Maybe I should disable flux or so. Okay. Just don't touch it. Just don't touch it. Yeah, so all your data is stored locally. And if we make these links, we for example only store the highlighted text, the URL, but no IPs, who gets it. We actually never looked into the database of those links before. We really don't want to analyze any of your data. And that's also really important for the kind of business approach we have because it's a for profit project. We try to build a SAS model out of it. But we intentionally don't take venture capital because we know that if we want... Once we do that, we would at some point need to sacrifice your data. It's inevitable because we... For the sake of growth that is required to have a venture-backed business, we at some point need to build lock-ins. We need to somehow monetize this huge chunk of data that is potentially on our servers for the sharing services. So that's not what we want to do. Instead, we use a model that is called Steward Ownership. You can also read about that on our website, which is basically a way of incentivizing collaborators and investors with some sort of profit sharing model that doesn't require us to grow in order, like to grow infinitely in order to give people a fair share for the risk they're taking. Also, the vision of where we want to go is that this MemeX is just one MemeX. It is really like every person on this planet has a completely different way of doing knowledge management. Everyone uses different apps that they want to connect with each other. Everyone has different workflows. So we really want to enable a whole ecosystem of MemeX devices that can be adapted to different workflows. So a developer might use a different MemeX than a student, than a journalist, than an entrepreneur. And that's also a reason why we don't take venture capital, is to be able to say, hey, the ecosystem can generate 95% plus of the wealth, of the revenue of the users in the ecosystem, and we would still be okay because we don't need to grow infinitely in that ecosystem. So, yeah, interoperability is at the core of what we do. It means you can easily switch to another MemeX and modify it and do whatever you want. You're not locked into our services. On that note, what we launched this week is the ability to back up and restore your data to any cloud provider. So we have built a little local server app for all main operating systems that would allow you to basically upload the backup to all cloud providers that have a local file system integration. So you stay full control of your backups, too, if you like to do that. But obviously, you can also choose Google Drive. That's for the majority of people. We haven't done that yet for priority reasons. We're a relatively small team, but it's on a roadmap. And when we have, for the syncing, for the syncing, the kind of relay, because we offer, like, asynchronous syncing, then we encrypt it on travel. But it will be deleted after it's synced. So it's possible on the basis of that. We haven't implemented it yet. But it's thought about to just point to a server, have some authentication, and just say make a append-only backup log there. So that's possible. With a little bit of contribution, that's certainly possible, yeah. So yeah. And then the next step for us is peer-to-peer sync between your devices. Means you can have, yeah, you can also use it on mobile phone. What we offer, then, as a service is to say for everyone who doesn't want to run a server or doesn't know how to run a server, we do it for them. And for everyone who knows that and wants to be independent, they can host their own handshake server, signaling server, to make the peer-to-peer sync possible. For that, we also developed a library that we call Storex, which is a library that would allow us to gradually replace our backend infrastructure. So, for example, we may think to implement an IPFS backend or a dot backend to make more distributed sharing possible. And so, because we're not sure yet which one of those to take, we have built this kind of modular database abstraction ecosystem for JavaScript. Please check it out. And maybe it's useful to you too. Which brings us to another next step that we're doing. That is collaboration. And so that's also the initial premise of the talk, is how do we actually be able to collaborate more in our web research? Because right now, there are so many limitations because our data is also locked up and it's very hard to reshare information. Because for one, you need to re-google it, right? Then you need to find all your notes that are relevant to it that you may have left in a chat log somewhere or in another note-taking application or in an email thread. And so it becomes so scattered that it's very hard to share. And so, one of the, like these are the three features that we're focusing on. And for example, the first one, co-curation of collections, means that you can create these folders that you then can share with other people and collaborate in curating them. This might be useful for entrepreneurs to do some customer research or for journalists to do the story research together. So, hey, this is all the stuff that I read about climate change here. Please, you can look through it and you can also full-text search through them. So, that's part of the publishing collections. You can basically say, maybe as a thought leader term, like someone who likes to share knowledge and who likes to share thoughts, maybe giving them the ability to say, look, here's the thousand articles that I ever read about climate change. And I don't want to just share a list. You can discover what I've searched for or what I looked at, for example, by just typing, what did this guy say about ocean acidification? And then basically being able to search in this kind of shared subset of his knowledge or her knowledge, plus all the annotations that this person has left there and wanted to share with the community or peer to peer. There is no central platform. It's really you shared with your people and your tribe and you find other people that you want to look into that, like, into the knowledge of those other people too. And the last point is annotation threats. So, having conversations, like really responding to annotations and highlights that people have made inside an article to exchange thoughts about that and be very precise about this. So, yeah, I wanted to end the talk a bit early, perfect, because I wanted to give a bit of room for questions, because it may be that it's complicated. So, so please. So it's a federation built in or on the road, because if I have my memory instance, and I have all my annotations, then if I have an other one too, and I said also, but they didn't know how to set it up, they can come to me, but then the data becomes centralized on my notes. So, is there any way for it not to commute? So, the thing is, that's also reason why we built Storex. In the beginning, we might have to start more centralized, because it's very difficult to build a federated system that at the end is useful to non-technical audiences. And so, because our tool is made for developers, but not exclusively, we probably start with more centralized ways, then thinking about how we make data interoperable. So, for example, by using a distributed hash table, like duck does it, so that addressability, that might be the technical now, that addressability becomes interoperable, but authentication server might be more centralized. So, we need to find, we need to figure that out still. I don't have a definite answer. But it's on our, it's in our interest to make the system as interoperable, and for the user as free to move as possible. I'm wondering about this, working together on the same knowledge. You mentioned the collaboration stuff that you want to do, but is that production ready? No, it's like in the roadmap. Yes. Okay. So, before that, as far as I understood, is the server that you set up somewhere, you make it run, and you send data to it, right? Is it possible for multiple people to set up the same server, and then they share everything freely, 100% with no... Potentially, yes. Now, it's more, right now, the version is more on a single player focus. So, multiplayer with having multiple people on the same server, that would be the difficulty that we're talking about, because that would mean we have a mastodon-like system where you need to sign up for a new node, to be a part of that node, and that is difficult in a more decentralized collaboration. At some node, etc., does it store the actual current version of the site, or just some metadata, and when I look at it after a year, and the site changes it meanwhile, if it does not change? So, what we... Yes, for that use case, that would be a problem right now. So, it doesn't... It doesn't persist the page, but it's on our roadmap to make snapshotting possible, and we're going to work with the Internet Archive. They said they gave their goal that we basically say, with every annotation that is done in a certain time frame, we make a snapshot at the Internet Archive, and add a link to an alternative snapshot to the page for that particular annotation, so that in any case you could at least go back to the Internet Archive, and then afterwards, being able to snapshot it for yourself, if you like to, but the thing is that this space-wise can get a bit tricky. This year. So, it will certainly come before we do the collaboration, because there it becomes more important to have a central source of truth when you're talking about an article. Sorry, yeah? I understood that the product is still early, like there's a lot of things to do, but I didn't hear of the search in China, or two, two, three, or the next we can save, and like two, two, two, two. Also, in the timeline, like for example, your reference can change, like you can find first reference, and then another one, and the new one, change how you think about the first one, all these kind of things you can, how you access all of your OpenMX, like you've been told about that. I'm not sure I've understood the question. How do you search through all the contents you connect? How we solve that? You search. How do you search all the contents? So, the technology underneath? Yeah. Yes, we built a search layer on top of taxiJS that runs in a browser, it's a custom search engine, and that right now we tested about 200,000 articles, and it didn't have, like, it has a little bit of flow down, but not measurable. But you didn't have down any specific word on this? Oh, no, no, I didn't do anything. Every website you visit is automatically indexed in the background, and if you have a decent, fast computer, you don't feel it at all, slower. Okay, him because he was for you. Yeah, to export it, we actually use hypothesis annotation client, so it's the same library. We don't export it yet as an open annotation data model, but it's planned to do so, yeah, totally. Yeah, it's a browser extension, but I haven't tested it on Chrome OS. Because the device which they use, not only for browsing, it's easy, fast boot, and works nice. If a Chrome extension works on Chrome OS, it should work. Just back in the page, because it's inside of the browser, it shouldn't be a problem. I guess so. We haven't just packaged it yet. Okay, does that mean you're going to go to Firefox? Yeah, it's Firefox, all browsers that support a web extension date, standards, so cool.