 Hey, everybody. How's it going? So, so it's Monday. Okay, so we used to have a mobile showcase back in the day. After the reorg, we created a reading showcase, a discovery had its own showcase, we're like, well, let's merge these two things. And then the questioner was like, well, why don't we have reading, discovery and editing all get together and do a showcase? So, okay, let's give it a shot. So this is our first attempt at doing that. It may or may not go smoothly. Apologies if we have any issues and if it's a bit bumpy. But these showcases are really meant to be informal, people come and show what they've been working on either for their work work, or for their like 10% time where they're doing research projects, which may not apply to media. We're planning to use the etherpad that's attached to the meeting invitation for people to ask questions, so we can actually get through more of this. So if you do a demo, hop in the etherpad after you're done with your demo and respond to questions inside of the etherpad, I don't know if we're going to break etherpad by having 100 people connect to it, but we'll see. Yeah, who knows? So we'll give that a shot. If you are presenting, there is a connection up here, like if you're presenting from the room, there's a connection up here. You'll want to come plug in and speak into this microphone. So I guess I will be the person probably holding the microphone up next to you. If you are speaking from this room, you'll just have to come up and do your demo. If you're on, well, whoever is speaking, that's like who gets recorded and who gets presented to everyone else. I would suggest if you are not presenting that you actually shut down your video right now, just because it results in higher quality video generally speaking. So yeah, again, this is meant to be informal. So you're here with people who are interested to hear what you're talking about. This is not a time to necessarily critique the judge. It's a time to just show off what you've been working on. So with that said, I think Yuri was the first person actually to be doing a demo. Is Yuri on the line? Yes. Okay, I'm going to mute my microphone. Can you guys see me and hear me okay? Yeah, go for it. All right. I'm about to share my entire screen. Okay. Can you see my screen now? I hope you can. I need some feedback. All good? Yes. Okay, excellent. Okay. So I've been working on maps and graphs for a while, and this is the very first little result that I want to share with you. Just two days ago, a mobile team launched their browser, their browser, sorry, browser Android app with the map support in it. And we see a major spike in the blue line. I'm sorry, green line. Green line is what shows how many unique users we have. And as you can see, we're now up to about 1.3 million files a day. Not that much really, but the servers are totally taking about 2% usage. Now, the interesting stuff. I've been working on maps on an interesting maps plugin extension to allow users to insert maps into the pages. So as you can see here, we have a green area with a phone zone, and you see a museum, like exploratorium, blah, blah, blah, blah, and you can even have something more fun like, I don't know, the cats, it's always fun. You can play a video here. It actually plays classical music when you click on it. Next on, this is how it's actually done. It's actually very simple, maps. There's a, oops, nothing is important. I probably shouldn't interrupt. Map tag is inserted. You just specify I want the map of this size at this coordinates, make it interactive, and then here you just write standard GeoJSON with styling it. It's not perfect at this point, but it shows the potential. Also, allow something more interesting like this with hiding everything else out and just showing some useful points of interest for the wiki voyage style. The wiki voyage actually is a big target for this, and because of them, we can allow multiple types of maps. You just define different groups. It says right here group A, group B, and you say here show the full map, here show just the map with group A or these groups, whatever. Now switching to the graphs. Graphs are actually live and have been live for almost a year now. This is this old style graphs. This is what people have been creating. There's a graph chart template that you can use to draw some basic graphs. There's also maps. That's old news, and you can already use it. But now we have new stuff. Well, this is another pie chart. People have been playing with pie charts for different classes. But the fun stuff is something I just did recently. Because of the page view API, you can now draw graphs like this, like usage of the main page, a viewing of the main page or of the current page or whatever. Or you can do interactive graphs, something like click to play. Thanks to Julian and Moise for helping with this and making it much nicer and cleaner interface. And you can say, oh, just zoom in into this area or that area. Or show me the map of the United States and show all the flights, how they connect with the cities. And you can see, oh, this is how this is where all the flights go, like from Ankara. Or something like this. This is my other favorite graph where family size versus productivity rate versus life expectancy. And you can play with it and see how these things are done. And I am done. Thank you very much. So sorry. Thanks, Yuri. It looks like our next demo is on guided path dialogue. Is that in the office or online? That's, can you hear me? Yep. Yeah, so I'm remote. I'll just share my screen with you. I seem to have to add an extension. Can I use your screen sharing? Just a sec. Sorry. Here we go. Can you see my screen? Not yet. Yes. Okay. All right. So, in visual editor, as well as simply editing text, you can edit more exciting things, for example, math equations. There are a bunch of extensions, different types of nodes that you can insert. And to show you what the old math editor looked like, I'm going to give you an example with hieroglyphics. There are basically automatic editors when you click on a node such as this. And you simply have an input box. And if you know the syntax of the thing you're editing, then you can just type away and edit it. However, if you're not familiar with the syntax, then you're kind of on your own now. So we thought with math, we would try to improve on that. The syntax behind editing a math equation is later. And here is what the new dialogue looks like. All right. So let's say I want to add a new equation or a new formula, let's say, sum of x. That was easy. I knew what to write there. So I just typed just as before. Now let's say, actually, no, I don't want to do sum of x. I want to do sum of log of x. However, I don't know how to write log. Now I have this handy menu down here and I can find the log button. That's how we write log. There we go. Sum of log of x. So basically, this menu has a whole bunch of buttons, different symbols that you can insert. Some of them are simple, like the log one simply inserts the command. Some of them do something a little bit more exciting. What if I wanted to do log base x? Well, then I can highlight the x, click on this one, log base x of y, and so on. So the idea is that someone who either is not very familiar with latex or is familiar with latex but doesn't know every single command that they could possibly need, they won't have to look it up. They can just use this menu. In addition, we also have some amount of error notification. You can see if you've typed something that was incorrect. And we have more exciting layouts down at the bottom. Let's say we want to enter a simple matrix and then we can just set it up or something. There are other things you can do with a math node other than inputting the formula. You can change how it displays and so we have an options tab as well. We can give it an idea as well. And that's the new math dialogue. That pretty much concludes what I have to say. Okay, cool. All right. Next up we have Ron on CrossFit notifications. We only have a Mac up here. So it's the option would probably be to go to another room. You want to give that a shot? You want to wait until after the next presenter? Okay. This is how we do it. I think Ron is sprinting right now. Okay. It looks like we have a Ron. You're muted at the moment, but we do see your video. Oh, yeah. Okay. I was muted there. All right. I'm going to try to share my screen here. Can you guys see my screen now? Yes. Yes. Awesome. Okay. So I'm going to do a quick demo of what we're working on in the collaboration team where we have a redesign of the notification panel almost ready to roll out. And that looks like this. Ignore the Chrome rendering bug. I'm working on that. I swear it's a bug in Chrome. We have some new UI stuff to do with notifications that we're going to be rolling out very soon. But the exciting thing that I wanted to show was that I have a three-weekie setup on this laptop. And if I go over here, it will show me that I have messages from a different wiki. And it will show me notifications about things that happened on other wikis. And if I go and think of myself from a different account on a different wiki, then I will hopefully now see that I think myself from a different account. So that's that. That's cross-week notifications. I should not take credit for any of this. The UI work was done by Moriel. And the backend work was mostly done by Kanal. And this will be coming to wiki near you in the month or so. Awesome. Just a second while we get set up. Installing extension. Never mind. Okay. Okay. We're good to go. All right. Here you go. Hello. So this is something that the discovery team has been investigating a little bit last week. When I go to the portal page, the top 10 wikis order is basically based on the week worth of page views. So sorted by the last week page views or whenever we pull the stats week long of stats of page views. And that's what determines the order. But what we think is this is why this is nice. This is a bit done for most users because everybody is very familiar with English Wikipedia. So it works for all of us. But if you're in a much smaller country or you don't speak English, you might never be interested in going to the English Wikipedia. So the idea is we can read navigator languages, which are the users' preferred languages. And we can try to eventually find that there is a wiki for your preferred language and try to put it more prominent on the page. So based on the navigator languages and here this navigator prefer languages are just English. So it actually doesn't change anything on the page. But on my personal computer since I'm French, I actually have English and French. And so for me, if I go here, see English on the top left and French on the top right. So put some examples. And for example, somebody in Hawaii, if he wants to see first the websites in Hawaii and second in English, and that's how the page could look like for him. And somebody in Catalonia in Spain would eventually see the Catalan wiki and the Spanish wiki before anything else. Same thing for, for example, somebody in Ukraine who would like to see in Ukrainian and Russian or somebody in India. So what we noticed is based on stats that we collect 60% of all that go to that page, don't do anything. And 30% of the people actually search something on that page. And 10% will click on a link. So this is the way actually increase that 10% rate. And yeah, I would like to get feedback from everyone. And eventually this is something that we could, we could write an AB test for a week and, and see if it improves something for the user experience. Right? That's it. Okay. So next, better type ahead on portals. Yep. Can everybody hear me? We can hear you. Oh, you can. Good. Cool. So yeah, I'm on and along with Julian also been working on the addict on the portal. Sorry, Chrome saying I have to add an extension to share my screen. Okay, has been added entire screen. Okay. And as Julian said, we do want to increase engagement on the portal page. And one of the things that Julian is tackling is the languages. We already have a better kind of more prominent search bar, maybe help people, you know, search for stuff. Oh, I gave it away. And today I'll be showing you the better type ahead on the search page, which is similar to the mobile page. So if I type in star wars, you have the thumbnail along with the title and description of the, of the wiki. So this has been on here, the cool, as opposed to what it is right now, which is just, just the title of the thing you're searching for. And this is also very similar to the mobile Wikipedia. You styled it a little bit for consistency as well. And here we go. And hopefully this will go in a AB test and we'll see if it actually does improve the conversion rate of the, of the search as well. That is the better type ahead. And that's the better type ahead. Thanks. Next up geo search over a bounding box. I hear you and see your screen. Excellent. So actually what I'm talking about is not ready to show you how it's working. I'm just telling you that I'm getting close to making a commit on a feature that will allow us to search not around a certain point, but within a bounding box. This could make it easier for apps, for example, to display on points of interest in a shape that's more resembles a phone screen, which is, has one side larger than the other one. So yeah, expect more very soon. Thanks. Tillman section usage data. Okay, trying to screen share. Can you hear me? We hear you. Awesome. Yeah, so I want to briefly show some new type of data on what people are reading on Wikipedia. Small thing, but I think it's pretty interesting. This is a byproduct of work with it on mobile web with John Robson. So on mobile web, as you know, you see the article and the subsequent sections are collapsed. So you have to click on to expand on a section of the index. So this is the World War Two article. And it's pretty long. And so we instrumented this. We have an event for the opening and closing actions. First thing we found out is that only about 40% of readers open a section. 60% are happy with only the least action apparently. Tillman, can I interrupt you for just a second? Are you sharing your screen? Yeah, that's a nice screen. Okay. Okay, I try again. I did not know you need the extension. I'm sorry. So that's installing now. I just keep talking while it's trying to install. Okay, thanks. Okay, so Tillman, are you still there? It looks like we may need to switch to the next presentation while Tillman is reset. Okay. It looks like Eric to the bottom of the list, James, singletown. Just a second. Set up. Hello. Oh, good heavens. Zoom level is a little small. Is there any chance we could get the video feed bump from 400 pixels? I'm sure it's somewhere here. It looks okay. You could just. It's not okay. There's enough content on the screen. No content on the screen. All you're seeing is system preferences. Before there was. That's a bit better. Okay. Okay. Okay. So, Visual Editor has existed for a little while now, and ever since the beginning, we added a hack to MediaWiki to totally disrupt user experience and add a second edit tab, which is incredibly ugly, incredibly ugly. We in fact broke Apple's scraping of our content by adding a second edit tab to sections, which was unfortunate. And it's been played by misery and pain and sadness. And we've always wanted to get rid of it for several reasons, most notably because presenting users with two buttons that do basically the same thing is a bad service to our users and it's assigned to our services. And we now feel we're getting towards the point where it's ready to provide that switch. And so we've now created a single edit tab. So, one wiki with two edit tabs magically becomes one wiki with one edit tab. And so this edit button will do one of two things. It will either take you to Visual Editor or the wiki text editor based on your preferences. So if I click edit, I've got it set up by default for new users to take them to the wiki text editor. And it then prompts you with this dialogue, which is being reworked language-wise. It says, hey, you know, ooh, I like you to remember for me. So if I click edit again, it will just open the wiki editor. But if I switch, it switches into the visual editor. And then if I click edit another time, it will again load the visual editor. And this also applies to section edit links. And if I then click an old version of page and click edit, this will also work. But if I do something that we don't currently support, like I want to undo that edit, we'll go, ugh, we don't know how to fix that. And so we dump it into the wiki text editor. Currently slightly inelegantly, but in the future it'll be pretty promise. And you're still able to do switching back and forth between edits as you go. He says, making a change, you can't actually see switching into visual editor. I changed this thing there, honest. Anyway, that's everything. Thank you very much. Okay. Michael, Michael, for getting the video back up. Okay, ready. Michael, we're only hearing clicks and pops over here. And it's real clean. Okay, do you read me? Yes. Okay. And can you see the emulator? Yes. Okay. And is it clear? It is in the room. Yes. Okay, cool. Yeah. So I wanted to show the pop up integrating Wiktionary content, which is one of the Android teams Q2 goals and what I've been working on the last sprint. So to get right into it, let's say I'm reading about a tetric work, and I want to know what some of the words mean. So I'll highlight the word in addition to where we have the copy and share menu items, there will now be a define menu item in the context menu. I'm just using the creative comments icon as a placeholder for now. Katie's created a define icon that we're going to put in probably this afternoon. And so I'll click on that and I'll get a pop up based on the same component as the link preview. It's a native component. Everything in here is native. No WebView. And it'll just give me a stripped down version of the Wiktionary content. Yeah, so let's say I want to know the meaning of valence. Now I've got all the definitions for the current language. I'm browsing in from Wiktionary for this word. So we have a bunch of noun definitions. Some with examples were available. Looks like another group of nouns with a different sense. Just to give you a sense of what's going on on the back end, this is built on the content service. And ultimately what it's doing is the content service is requesting the Parsoid HTML, getting it back, and then stripping a whole lot of stuff away and just sending a really minimal JSON response to the mobile client, which basically only contains what you see here. So it's getting this stuff really efficiently. And that's that. If there are any questions, I'm happy to answer them. Thanks. Let's take the questions to the etherpad. Okay, Tillman, do you want to give it another try? Okay, looks like Tillman has dropped off. Eric. Sure, we'll set up a screen share here. Is that coming across? Yes. So we, within search, we've been working on replacing or coming up with a replacement for the autocomplete. Currently the autocomplete feature that you see on Wikis in the top right corner here is always a pre strict prefix search, which means if you misspell something or anything like that, you basically just don't get any suggestions. So this Thursday we're shipping out a new beta feature called the completion suggestor. It'll show up on your beta features and you just turn it on with a click. And it basically, it allows you to, so we were to So we were to type in like queen on the left side here, I have the old prefix search. So if you're to do queen, you get a strict list. This is also a, there's not much scoring going on there. It's just putting things up. If we do it with a new completion suggestor, I guess there's two new things with this. One is that it handles a fuzziness, which I'll show in a moment. It also handles scoring a bit better and better in that we can, we can generate scores for all these things. And then we sort them based on those generated scores. The scoring algorithms still going to need a bit of work, but it pulls up things that are quite interesting like Queen Elizabeth, Queen the band, Queen's University, whereas Queen on the prefix search. Also interesting, there's some other ones that, like Quercus rober probably should not be the top result for anything, except unless you really type that some other demos though, if you were to type something wrong, like if you were to type green day, but without a space between the green and the day, it'll still find the green day. Even if you drop a letter out of there and you miss spell it, we're still going to find the things that you're looking for. And so we're hoping this is going to help people to find things a little bit better. This is one of the reasons we haven't done too much with this in the past is that performance is a serious concern for this. We do something like a hundred million prefix searches per day based on this. This is a graph from elastic search for somebody that was working on adjusting this. And we can see the completion suggestor can actually be more performant than prefix search can be. That's because this is based entirely on an in memory algorithm called the finite state transducer. I won't get too much into it, but I'll post a link to a blog post that explains how that all works. If anybody's curious, this is kind of what the data structure looks like internally. It's a directed a graph, but yeah, that's our new completion suggestor. It'll be coming out on Thursday, hopefully. We're fully expecting to anyways, and hopefully it'll help people find things a little bit better. Okay, very cool. Thanks everybody for doing your demos. We will get this out onto YouTube and comments after a bit. If you did do a demo or if you have a question, please add details to the showcase on the etherpad. And I think we can actually wrap early, which it's not something we can always do. Thanks again. Have a good day.