How guys from Storyline returned from Silicon Valley to Minsk with $ 770 thousand for a startup
In 2017, Vasily Shinkorenko and Maxim Abramchuk created the Storyline service, where you can develop applications for a voice assistant from Amazon without using programming. The guys participated in two accelerators in Silicon Valley - Boost VC and Y Combinator. In the spring of 2018 they received $ 770 thousand investment. Fillpackart
and I called them and asked about everything - what is under the hood of the service, what is it like to be a developer in a startup, how they lived in the valley and how soon the voice assistants will change the world.
- Are you Belarus now?
Vasily : - Yes, 8 months were in California, and flew back to Minsk in April, as soon as the YC passed. Even there, we were considering options for how best to build a company. There were many options, but decided to return to Minsk.
First of all, of course, because of the resources. It is much easier and cheaper to search for people in Minsk than in the valley. There you compete with big companies, with cool startups that are released every year. And here we are the only ones who passed the YC in the history of Belarus. Here we do not need to raise so much money to pay engineers 200 thousand a year. We are now at the stage where we work 12 hours a day. There is no difference where to work. Therefore, we chose Minsk.
Maxim Abramchuk and Vasily Shinkorenko - Has the
office already been removed, equipped?
Q: - Yes, yes! We have a wonderful office on the fourth floor, overlooking the river, in the city center.
- Are you planning to recruit a big team?
Q: - Now we are 6 people. By the end of the year there will be 10. And then everything depends very much on metrics. To achieve the goals that we set for ourselves for a year, we will need about 10-15 people.
- I’m not very much in the subject line, but I lived in Minsk for a couple of years and I heard it with the edge of my ear, supposedly all the guys there were taken to Wargaming, EPAM and abroad. Have you encountered such a problem?
Q: - Yes, but the trick is that there are still very few food companies - especially startups. And startups that are more or less adequate do something and do not look like half-dead zombies on the fingers of one hand. I do not think that five companies is a big competition.
- Do you already feel that you are turning into a business?
Q: - We can say that we have already groped something, but I cannot say that this is what will lead us to global success.
Before the product appeared, only developers developed applications for Alex, wrote code, hired companies. We have essentially created a new market segment - now people without a technical background are also doing them. And now this segment is growing, and we are growing now, and the smart speaker market as a whole is growing, the number of applications from users is growing. We are already making a little bit, while it is not yet profitable - it’s far from him, but already somewhere on the horizon.
I think, as long as all people fit into one room, this is more a product than a business. That's when you pass for 15-20 people, then you will need to think about the processes and much more.
- When you received investments, what was more - euphoria or fear that you could not cope?
Q: - It must have been a feeling of mailstone. Most startups do not live the first six months. And when you lived, made a thing that is interesting to someone, people use it, and, most importantly, really love it and are ready to pay - then you understand a bit more what to do next.
Well, that is, there was no such feeling that you were dancing on your head and throwing slippers. Just understanding that the next step is now beginning, and in it there are your goals, your KPIs. YC helped a lot with this, said what the product should look like in a year, and what needs to be done to do this. And so we do.
- Tell us, what did you do in Belarus before leaving the valley?
Q: - I wrote the code for a long time myself, but not very well. Then I realized that I was better given communications, all sorts of business things, marketing. I started to do this and accidentally fell into the realm of dialogue interfaces - bots in the Telegram, in other messengers, engines, frameworks for bots. Spin around it. In 2015, we made several products with the guys and even raised some money.
Together with Max, we started to make one product. Although he did not succeed, but we got the experience. Then together they built an agency of 7 people. And by the summer of the 17th year they went back to groceries.
- Max, and you?
Maksim: - Programmed began in the first year. I had some kind of setup in my head, that while I was getting ready for a university - I was teaching physics, mathematics and all sorts of things, but I don’t take up programming in principle. As soon as I saw myself in the lists of BSUIR, I immediately took it, started C ++, all sorts of other things. The first few months was stupid, it was very hard. Then I learned that there is a web, I started sticking to the frontend. Saw Ruby, wrote a couple of months on it. In general, toss, spin, spin and stop at the back end.
Set a goal to find a job. For several months, I raised Ruby, SQL, made a couple of test cases. They took me to one company, and I worked there for half a year. There were about five of us, and we just made different products at an outsourcing. Then I went to another company, I worked there for about a year and a half. And then just started bots platform in the telegram. I don't remember why, but I was so inspired by it all. For a couple of days some saytik has collapsed (it may still be working). We began to make bots to order.
And in the wake of all this, we met Vasya and started making products.
Phill : - That is, are you a stack larger than the backender?
M: - Well, backender is for me some kind of a rocker or javist. And since I am writing backend in Ruby, there is a place to go for a walk. 30-40% of the time had to write to the front, so I can say I am full. Eight or nine months, until we took the first developer, I wrote the whole Storyline myself - both the front, and backing and devops and everything else.
F: You have no desire to stop writing code and become a pure lead?
M: - I like to write code and manage developers, I also like it, but I would not do full time. I have a desire to deal with the product. Test all hypotheses, watch analytics, write roadmaps, communicate with users and so on. This is super interesting. Perhaps in the future, when we get more guys, and their expertise will be stronger, I can pass on everything to them. But I don’t know yet when this will come true.
F: Max, did the Storyline idea seem interesting to you from the point of view of implementation?
M: - I, in fact, am not a fanatic especially to program for the sake of programming. I liked the idea itself. Yes, it was fun to write an interactive editor of applications, it was interesting to cut new features, maintain old ones, fix bugs, grow metrics. And now the front-endder has come to us, and it is very interesting to optimize the canvass together - so that it can be rendered faster, opened up with users, and does not lag.
F: And before that, you didn’t work with canvas and 2D graphics?
M: - Yes, it did not work. But this is what I call it - canvas. In fact, we use d3.js and SVG, because writing on the canvas would be even more difficult. On the Canvas, we would need to say "Take this point with such coordinates, draw a line to some coordinates, with such and such thickness." And it would take a lot of time to draw a block with fill, with shadows and so that you can drag-and-drop. Therefore, we chose SVG, and there are no particular problems with performance.
F: - Do you spend a lot of effort on prototyping?
M: - I don’t know if it’s good or bad, but I’m doing it pretty quickly. Why we moved so quickly - we literally walked away from the idea to the minimum implementation in a day or a half.
Of course, now I am stepping on a rake because of this, so I have to move on to a more standard process. Still, we no longer have hundreds of users, but tens of thousands. They find some specific bugs, use cases that we did not provide. Therefore, we have to slow down, add testing - manual, automated.
But it is very cool when you can come up with a couple of hours, immediately make and immediately receive a feedback. In my opinion this is not enough in the minds of many developers. Too many sawing to saw, for weeks and months. And for me it's very cool - to come up, quickly do and show.
- When the prototype turned out, do you finish it or start rewriting from scratch?
M: - When we started, we had a tense situation - users started coming right away, it was necessary to show that there is traffic. Then we did not rewrite everything, but decided to use the prototype further and continued to cut features on top of it.
Initially, there was such a concept: you write the text that says Alexa, put two slashes and write the text that the user says. Then it was necessary to pervert, to make all sorts of transitions, variables, and in fact there was a bare text that was hard to debug. And our users are not technical. They opened the bracket, forgot to close it, and everything broke. But then we went to YC, we were very drunk - “Let's go! Let's!". Users grew, bugs grew. It seems we are so stuck with this prototype of the month for three.
F: - If my friends and I are going to do a project, we will choose technologies that we know, and not those that could be better suited. Did you have it too?
M: - Yes, in principle, so. But we are very lucky, because JS and Ruby are such a great stack for a startup. In Ruby, everything can be washed down very quickly, and I would not say that everything becomes bad afterwards. Unless, of course, you reach the level of Twitter - so Twitter was rewritten from Ruby to Scala.
While you're a startup, Ruby on Rails is enough for you. But if I wrote in Java, probably, not everything would be so rosy.
F: The problem here is that you can do well only with what you can do. Do you have a clean JS?
M: - We have the usual JavaScript and React with Redux
F: - Why did you choose them?
M: - When we started writing Storyline, I didn't use React. Before that I wrote on Angular 1.4, 1.5. Slightly poked the second Angular, which is on TypeScript. When we started Storyline, React had already won the war, so it was more promising to write on it (and easier, as it seems to me now). So I just googled the React Tutorial, put it in a couple of days, started the default generator from Dan Abramov, and Redux was already there. Now I do not regret, although I did not have much experience with other libraries.
F: - Can you say a few words about architecture?
M: - Everything is quite simple, we have a Single-page application on React. Rails are used for two things. The first is an apish for a client application. The second is hosting for the skills that users create via Storyline.
The front consists of two large components. The first is the Dashboard and everything inside the project, and the second is the editor of the skill. It is already written interspersed with d3.js. Blocks are drawn there - it’s necessary that all this be fast, super interactive and beautiful, so that you can draw all sorts of arrows, animations and so on. It was super important for us, because we essentially have it for visual programming.
And Alexa, you can say, works like a normal bot in Telegram or Facebook. You feed her some endpoint, some servachok. As soon as the user starts talking to Alex, she sends you rekvesta. You work as a proxy - you receive a request, you respond to it. Alex is formatting your answer and pronouncing it. Nothing supernatural - the usual monolithic rail application, everything works on Amazon Web Services. We try to do everything as standard - React, Redux, Ruby on Rails - it's easier to find developers.
F: - What part now seems the most difficult?
M: - Both technically and conceptually difficult, the implementation of cross-platform seems to be. Now we only support Alex, then we will be Google home, Cortana, and other things. To be honest, we are still not technically prepared for this. I think there will be big problems there, and it will be a huge challenge.
F: - Can you tell how many lines of code there are in the largest file on the project?
M: - There are probably 700 lines on the frontend - there is one file there. On the backend, probably smaller (I just write better in Ruby). There may be 250-300.
- In the valley is a cool community, everyone is helped. Is it possible to bring your code to someone from the development tops, show, ask for advice?
M: - I think it will look a bit strange. The same guys from Slack can ask for some advice, for which there is no right answer. You can sit, philosophize. They can only give advice from their bell tower.
But with programming, I don’t know. Still, you can google, protest, find on github. I don't think any super cool dude would be against it, but it doesn't really make sense.
Q: - It makes sense in specific areas. If you do something on machine learning, and there is a dude who has been doing this for 20 years, then yes, it will be super useful.
F: - Max, what do you think, what is it like to be developers in a startup compared to other areas?
M: - I only know what it is to be a founder, as a developer, I hadn’t worked in startups before. Now I don’t consider myself to be a developer, because I have to write less and less code and dive more into a bunch of different processes - almost before buying water in the office.
But still the developer in a startup is very cool. You do not just saw features that the customer told you, but you can influence and invent them yourself. But it is a very big stress. Everyone wants fast, and you yourself too. In an outsourcing, you can say “I spent 10 hours on a task”, although in fact I spent 4, and everything will be okay. You can just sit and sort through technologies, rewrite one class many times, cover everything with tests.
And then you yourself control your time. There is always the interest to do faster, better. In a startup, you need to balance between a bunch of different things.
F: Did you have to somehow confirm your skill in order to receive investments?
M: - It may sound sad and insulting, but it seems to me that the role of technology was not particularly important. I think that 95% of startups are the same, if we don’t talk about complex technology companies that have technology in their essence.
And all the other technologies fade into the background. Write some piece or application - a lot of mind is not necessary. There are a lot of frameworks, tutorials, it's all pretty understandable and simple. And to find a suitable idea and a suitable audience, to have time to make a product, to understand who needs it - at times, many times more difficult.
It seems to me that the role of technology in startups is overrated in the minds of people. Stack, quickly or slowly you write code - it's all not so important. Much more important is your strategy, how do you find solutions.
- What are they looking at more? At the very idea or your ability to grasp it and bring it to the end?
Q: - YC primarily looks at the founders - how many units of work you can perform per unit of time. The main thing is how fast you can do, test, understand what does not work where and continue.
The second thing they look at is the market. If you do it very quickly, but in a bad market you will not swim far away. In very large markets, as a rule, there is high competition, so they try to find small but rapidly growing ones. We have, I believe, was the perfect timing. We made the product as soon as it became possible to make it. Timing is one of the most important pieces.
And the third is the product itself. It is important for them that people love him. It is better to make a product that hundreds love, than a product that thousands like. If you just like it, it will be difficult to convert. Much better to find a small community. We grew from zero to 14 thousand users organically simply because some people told us about others.
Just last week in our group on Facebook wrote a man with his real estate business. Now a popular phenomenon is when people look at home not with the realtor, but themselves. He wanted to make a skill for Alexa, put it in all the houses according to the device, so that clients could ask her about everything themselves. He appealed to the agency to develop such a skill, and there they asked him for $ 45,000.
He was upset and went to look for how to do such a thing himself. I found our product and made this thing - for which the agency requested 45 thousand - in just a few days. And so he bought our paid version, although he didn’t even need the features that were there (a good reason for us to think about new segments). He just loved the product and supported us.
- You were interested in the topic of linguistics - to teach a computer to speak, all that?
Q: - Yes, it is very cool. I remember in Startrek there was a computer with which they spoke. It was fun, but it seemed so futuristic. And this is actually happening.
There is a story. I'm going from work to training, I sit in Uber, and then the driver is talking to someone. I look - in the car there is nobody. It turned out he spoke with Alice in Yandex Navigator. And here behind him was some dude. The driver began to swear, and Alice thought that he was swearing at her - telling him, “how are you talking to me?”. The driver is such, "shut up there, generally not up to you." It looked so real. You think that the future is somewhere far away, but it is here, it has already come.
Of course, the extent to which devices and helpers can respond contextually is far from ideal. But speech recognition is getting better. If now Alex correctly understands 8 words out of 10 (according to my feelings), then in a few years he will understand all 10. This will greatly affect how people will talk with their device. Siri is super stupid, doesn’t understand anything at all, and this is very outraged. Voice assistant such a product - in order for it to work well, the technology itself must be good.
- You would not want to tie some kind of training, so that Alex understands not literally what you say to her, but use the context?
Q: - Partly it works in Storyline now. For example, there are intentions YES / NO, and a bunch of synonyms on the Amazon side are automatically added to them. If you set Yes, Alex will understand Ok, Fine, True, and so on. All this is done at the level of Alexa, and we use it as an interface.
“When will cars ever talk better than we do?”
Q: - This is a question about artificial intelligence in general. Turing test is now quite possible to pass, specifically to train the system for a narrow task. But speaking better than a person means understanding the context, emotions, intonation, combining everything to give an answer. I don't know, it's very hard for me to imagine.
I may be completely wrong, but it seems to me not earlier than in 10-15 years.
and I called them and asked about everything - what is under the hood of the service, what is it like to be a developer in a startup, how they lived in the valley and how soon the voice assistants will change the world.
- Are you Belarus now?
Vasily : - Yes, 8 months were in California, and flew back to Minsk in April, as soon as the YC passed. Even there, we were considering options for how best to build a company. There were many options, but decided to return to Minsk.
First of all, of course, because of the resources. It is much easier and cheaper to search for people in Minsk than in the valley. There you compete with big companies, with cool startups that are released every year. And here we are the only ones who passed the YC in the history of Belarus. Here we do not need to raise so much money to pay engineers 200 thousand a year. We are now at the stage where we work 12 hours a day. There is no difference where to work. Therefore, we chose Minsk.
Maxim Abramchuk and Vasily Shinkorenko - Has the
office already been removed, equipped?
Q: - Yes, yes! We have a wonderful office on the fourth floor, overlooking the river, in the city center.
- Are you planning to recruit a big team?
Q: - Now we are 6 people. By the end of the year there will be 10. And then everything depends very much on metrics. To achieve the goals that we set for ourselves for a year, we will need about 10-15 people.
- I’m not very much in the subject line, but I lived in Minsk for a couple of years and I heard it with the edge of my ear, supposedly all the guys there were taken to Wargaming, EPAM and abroad. Have you encountered such a problem?
Q: - Yes, but the trick is that there are still very few food companies - especially startups. And startups that are more or less adequate do something and do not look like half-dead zombies on the fingers of one hand. I do not think that five companies is a big competition.
- Do you already feel that you are turning into a business?
Q: - We can say that we have already groped something, but I cannot say that this is what will lead us to global success.
Before the product appeared, only developers developed applications for Alex, wrote code, hired companies. We have essentially created a new market segment - now people without a technical background are also doing them. And now this segment is growing, and we are growing now, and the smart speaker market as a whole is growing, the number of applications from users is growing. We are already making a little bit, while it is not yet profitable - it’s far from him, but already somewhere on the horizon.
I think, as long as all people fit into one room, this is more a product than a business. That's when you pass for 15-20 people, then you will need to think about the processes and much more.
- When you received investments, what was more - euphoria or fear that you could not cope?
Q: - It must have been a feeling of mailstone. Most startups do not live the first six months. And when you lived, made a thing that is interesting to someone, people use it, and, most importantly, really love it and are ready to pay - then you understand a bit more what to do next.
Well, that is, there was no such feeling that you were dancing on your head and throwing slippers. Just understanding that the next step is now beginning, and in it there are your goals, your KPIs. YC helped a lot with this, said what the product should look like in a year, and what needs to be done to do this. And so we do.
- Tell us, what did you do in Belarus before leaving the valley?
Q: - I wrote the code for a long time myself, but not very well. Then I realized that I was better given communications, all sorts of business things, marketing. I started to do this and accidentally fell into the realm of dialogue interfaces - bots in the Telegram, in other messengers, engines, frameworks for bots. Spin around it. In 2015, we made several products with the guys and even raised some money.
Together with Max, we started to make one product. Although he did not succeed, but we got the experience. Then together they built an agency of 7 people. And by the summer of the 17th year they went back to groceries.
- Max, and you?
Maksim: - Programmed began in the first year. I had some kind of setup in my head, that while I was getting ready for a university - I was teaching physics, mathematics and all sorts of things, but I don’t take up programming in principle. As soon as I saw myself in the lists of BSUIR, I immediately took it, started C ++, all sorts of other things. The first few months was stupid, it was very hard. Then I learned that there is a web, I started sticking to the frontend. Saw Ruby, wrote a couple of months on it. In general, toss, spin, spin and stop at the back end.
Set a goal to find a job. For several months, I raised Ruby, SQL, made a couple of test cases. They took me to one company, and I worked there for half a year. There were about five of us, and we just made different products at an outsourcing. Then I went to another company, I worked there for about a year and a half. And then just started bots platform in the telegram. I don't remember why, but I was so inspired by it all. For a couple of days some saytik has collapsed (it may still be working). We began to make bots to order.
And in the wake of all this, we met Vasya and started making products.
Phill : - That is, are you a stack larger than the backender?
M: - Well, backender is for me some kind of a rocker or javist. And since I am writing backend in Ruby, there is a place to go for a walk. 30-40% of the time had to write to the front, so I can say I am full. Eight or nine months, until we took the first developer, I wrote the whole Storyline myself - both the front, and backing and devops and everything else.
F: You have no desire to stop writing code and become a pure lead?
M: - I like to write code and manage developers, I also like it, but I would not do full time. I have a desire to deal with the product. Test all hypotheses, watch analytics, write roadmaps, communicate with users and so on. This is super interesting. Perhaps in the future, when we get more guys, and their expertise will be stronger, I can pass on everything to them. But I don’t know yet when this will come true.
How we got into the valley and failed the first idea
Our friend who once participated in Boost VC came to Minsk. We met, showed him the project, which was done for one cafe. He said: “Cool stuff! Didn't you think it would be turned into a separate product for cafes? ”
We are,“ let's try. ” It turned out he knows Adam Draper well from Boost VC, and says: “Apply there and if you can pass, take a trip to the valley. It will be more interesting than what you are doing with the agency. ” We are okay, filled out the application did some landing. First, the selection took place, then the Skype interview with Adam. We made an offer (BoostVC usually gives 25-50 thousand for 7% of the company). We accepted it and went to the valley with a product for restaurants.
Adam Draper
McDonald's and all sorts of fast food are self-service terminals. They are quite expensive. Most medium-sized cafes do not have the ability to deliver such, but do not want to hire a lot of cashiers. We figured out how to replace such terminals in an easy way - to order food through bots.
People through the messenger will be able to make orders, or - if it is a cafe, next to work - find out if there is a queue now, so as not to waste time and not go down in vain. That is, you place an order, you come at a convenient time, and you get a discount for it. And for owners and managers, the main point is automation.
Here we come, try to negotiate with these people. But it turned out that small cafes do not need automation - they need customers. And big companies do not care that the self-service terminal is worth a dozen. If you are nobody and call you in any way - almost without a chance that a big client will pay attention to you.
I found one dude who makes tablets for orders in restaurants. He went through YC, and then worked for three years on a deal with Wendy's. We realized that nothing will work at all, we do not understand anything in restaurants and will not live in this mode for three years.
Began to think what the sharim. When we made bots to order, I always drew such Mindmaps - diagrams, how the dialogue will go, which lines the user will say, which bot. Huge chat with arrows that we showed to customers for approval. So they thought how cool it would be to transfer this diagram to a real bot with one button. And so that it was not necessary to code, pull the developers, if the client sends changes.
So we decided to switch to Storyline.
F: Max, did the Storyline idea seem interesting to you from the point of view of implementation?
M: - I, in fact, am not a fanatic especially to program for the sake of programming. I liked the idea itself. Yes, it was fun to write an interactive editor of applications, it was interesting to cut new features, maintain old ones, fix bugs, grow metrics. And now the front-endder has come to us, and it is very interesting to optimize the canvass together - so that it can be rendered faster, opened up with users, and does not lag.
F: And before that, you didn’t work with canvas and 2D graphics?
M: - Yes, it did not work. But this is what I call it - canvas. In fact, we use d3.js and SVG, because writing on the canvas would be even more difficult. On the Canvas, we would need to say "Take this point with such coordinates, draw a line to some coordinates, with such and such thickness." And it would take a lot of time to draw a block with fill, with shadows and so that you can drag-and-drop. Therefore, we chose SVG, and there are no particular problems with performance.
F: - Do you spend a lot of effort on prototyping?
M: - I don’t know if it’s good or bad, but I’m doing it pretty quickly. Why we moved so quickly - we literally walked away from the idea to the minimum implementation in a day or a half.
Of course, now I am stepping on a rake because of this, so I have to move on to a more standard process. Still, we no longer have hundreds of users, but tens of thousands. They find some specific bugs, use cases that we did not provide. Therefore, we have to slow down, add testing - manual, automated.
But it is very cool when you can come up with a couple of hours, immediately make and immediately receive a feedback. In my opinion this is not enough in the minds of many developers. Too many sawing to saw, for weeks and months. And for me it's very cool - to come up, quickly do and show.
- When the prototype turned out, do you finish it or start rewriting from scratch?
M: - When we started, we had a tense situation - users started coming right away, it was necessary to show that there is traffic. Then we did not rewrite everything, but decided to use the prototype further and continued to cut features on top of it.
Initially, there was such a concept: you write the text that says Alexa, put two slashes and write the text that the user says. Then it was necessary to pervert, to make all sorts of transitions, variables, and in fact there was a bare text that was hard to debug. And our users are not technical. They opened the bracket, forgot to close it, and everything broke. But then we went to YC, we were very drunk - “Let's go! Let's!". Users grew, bugs grew. It seems we are so stuck with this prototype of the month for three.
How we launched a prototype and replaced it with a product on the go.
We made the first super primitive version of the product and began to show it to everyone. By this time we have already gathered the community, made a telegram channel. They asked, “look at such a thing, what do you think?” A lot of dudes wrote to us. With all we phoned, asked around and realized that what they want is much more complicated than what we do.
We were helped by the dude from Slack - Amir Shevat, now he is VP by Developer Experience on Twitch. He also fumbles in bots, even wrote one of the most popular books on the topic "Designing Bots". We came to his office in Slack, showed our "prototype" - at that time it was a picture in Sketch.
Amir Shevat
He says “well, cool, but how do you want to develop it further?” We say, let's go on messengers - Telegram, Facebook. But they also thought that it was cool for voice pieces, such as Alexa, Cortana, Google Assistant, because there are no controls, no menus. There is only what the user says, and that the device. Amir says, “cool, try making the first version. It will be interesting". And on the same day, we saw that Amazon had just opened api for Alexa, which allowed it to programmatically create skills (bypassing the Amazon developer console).
We started to make a product, but quickly came up against the problem. Super hard to do in the application canvas - a piece where you can blocks, something to write.
Approaching the demo day in Boost VC. A month remained before him, and the work on the canvas was about the same. We sat at night in the room where we lived, and thought - “Damn, something we do not have time at all. We will not have time to start the campaign. ”
And here we found a service for Coggle mindmaps, plugged it in a couple of hours with an iframe, quickly screwed in a couple of buttons and filled it. It turns out, when you fought in Storyline, you saw Coggle (but we had its logo and controls just taped the background color). Users made diagrams in it, and we took them JSON and converted them into a skill for Alexa.
So we collected this thing, took a little video on the phone, how it works and launched it in several Facebook groups related to Alexa. The next day we already had more than a hundred likes, comments. Everyone asked to give access. From there came a couple of hundred first users.
Since Coggle was essentially a text, and there it was impossible to create objects, pictures, all sorts of things - there was a big task to migrate all user objects from a mind map into structures. We did not even have our own database then. We kept only the users, all the maindmaps were stored in Coggle, and he gave them to us by apishke.
It was necessary to capture all the storage with us, transfer all projects to our base. And this was when the users have already begun to publish their projects. So, all these stories about startups, when a prototype is fired and you have to replace it with a product on the go, this is about us.
F: - If my friends and I are going to do a project, we will choose technologies that we know, and not those that could be better suited. Did you have it too?
M: - Yes, in principle, so. But we are very lucky, because JS and Ruby are such a great stack for a startup. In Ruby, everything can be washed down very quickly, and I would not say that everything becomes bad afterwards. Unless, of course, you reach the level of Twitter - so Twitter was rewritten from Ruby to Scala.
While you're a startup, Ruby on Rails is enough for you. But if I wrote in Java, probably, not everything would be so rosy.
F: The problem here is that you can do well only with what you can do. Do you have a clean JS?
M: - We have the usual JavaScript and React with Redux
F: - Why did you choose them?
M: - When we started writing Storyline, I didn't use React. Before that I wrote on Angular 1.4, 1.5. Slightly poked the second Angular, which is on TypeScript. When we started Storyline, React had already won the war, so it was more promising to write on it (and easier, as it seems to me now). So I just googled the React Tutorial, put it in a couple of days, started the default generator from Dan Abramov, and Redux was already there. Now I do not regret, although I did not have much experience with other libraries.
F: - Can you say a few words about architecture?
M: - Everything is quite simple, we have a Single-page application on React. Rails are used for two things. The first is an apish for a client application. The second is hosting for the skills that users create via Storyline.
The front consists of two large components. The first is the Dashboard and everything inside the project, and the second is the editor of the skill. It is already written interspersed with d3.js. Blocks are drawn there - it’s necessary that all this be fast, super interactive and beautiful, so that you can draw all sorts of arrows, animations and so on. It was super important for us, because we essentially have it for visual programming.
And Alexa, you can say, works like a normal bot in Telegram or Facebook. You feed her some endpoint, some servachok. As soon as the user starts talking to Alex, she sends you rekvesta. You work as a proxy - you receive a request, you respond to it. Alex is formatting your answer and pronouncing it. Nothing supernatural - the usual monolithic rail application, everything works on Amazon Web Services. We try to do everything as standard - React, Redux, Ruby on Rails - it's easier to find developers.
F: - What part now seems the most difficult?
M: - Both technically and conceptually difficult, the implementation of cross-platform seems to be. Now we only support Alex, then we will be Google home, Cortana, and other things. To be honest, we are still not technically prepared for this. I think there will be big problems there, and it will be a huge challenge.
F: - Can you tell how many lines of code there are in the largest file on the project?
M: - There are probably 700 lines on the frontend - there is one file there. On the backend, probably smaller (I just write better in Ruby). There may be 250-300.
How programs are arranged in accelerators and how life goes on in Silicon Valley
Tim Draper
We lived in San Mateo, California, at Draper University. It was built by the father of our investor, Adam Tim Draper. He is there one of the most famous dudes. He is even called the founding father of the valley. He bought the hotel and made him a university, where his program runs twice a year. There we lived, and other companies. And for those involved in Boost VC, they provide free housing for four months.
This is a very weighty cash bun, because the prices in San Francisco ... It also gives the office space in its space, literally 10 meters from the house - just the way to go. So a typical day looked like this: get up, go to Boost, go out at night, cross the road and go back to sleep. Approximately 97% of our days looked like this.
There is no formal program in YC or Boost VC. You do not sit and do not listen to lectures. But every week joint dinners, where they invite famous guys. There was a Quora fighter (which turns out to be a CTO on Facebook), guys from AirBnb, Amplitude, Zapier, Dropbox. A huge table is laid, everyone is sitting, eating, listening, talking.
Rallies are held every two weeks. There all are divided into groups, according to some magical parameters - industry, technology and so on. In total, we had about 40 companies, and 14 of them were in our group. Each group has its own partners. We had Michael Saybel (CEO YC and Twitch founder) and Gustaf Alztromer (head of growth in Airbnb).
We were going and saying what we had done in the past two weeks, what progress on the metrics, what we would do in the next two weeks. The third question has always changed: What are the problems, what is stopping you from growing, how are we going to become a company for $ 1 billion.
And if by the next time you didn’t do what you had planned, it’s a shame to talk about it before the whole group and partners. It strongly motivates to move forward, puts you on the rails so that you grow. If you grow by 50% for several months in a row, it begins to seem that next month it will be impossible to grow. And now you start looking for all sorts of tricks, how to do it in any way, you get used to it and grow up.
- In the valley is a cool community, everyone is helped. Is it possible to bring your code to someone from the development tops, show, ask for advice?
M: - I think it will look a bit strange. The same guys from Slack can ask for some advice, for which there is no right answer. You can sit, philosophize. They can only give advice from their bell tower.
But with programming, I don’t know. Still, you can google, protest, find on github. I don't think any super cool dude would be against it, but it doesn't really make sense.
Q: - It makes sense in specific areas. If you do something on machine learning, and there is a dude who has been doing this for 20 years, then yes, it will be super useful.
F: - Max, what do you think, what is it like to be developers in a startup compared to other areas?
M: - I only know what it is to be a founder, as a developer, I hadn’t worked in startups before. Now I don’t consider myself to be a developer, because I have to write less and less code and dive more into a bunch of different processes - almost before buying water in the office.
But still the developer in a startup is very cool. You do not just saw features that the customer told you, but you can influence and invent them yourself. But it is a very big stress. Everyone wants fast, and you yourself too. In an outsourcing, you can say “I spent 10 hours on a task”, although in fact I spent 4, and everything will be okay. You can just sit and sort through technologies, rewrite one class many times, cover everything with tests.
And then you yourself control your time. There is always the interest to do faster, better. In a startup, you need to balance between a bunch of different things.
F: Did you have to somehow confirm your skill in order to receive investments?
M: - It may sound sad and insulting, but it seems to me that the role of technology was not particularly important. I think that 95% of startups are the same, if we don’t talk about complex technology companies that have technology in their essence.
And all the other technologies fade into the background. Write some piece or application - a lot of mind is not necessary. There are a lot of frameworks, tutorials, it's all pretty understandable and simple. And to find a suitable idea and a suitable audience, to have time to make a product, to understand who needs it - at times, many times more difficult.
It seems to me that the role of technology in startups is overrated in the minds of people. Stack, quickly or slowly you write code - it's all not so important. Much more important is your strategy, how do you find solutions.
- What are they looking at more? At the very idea or your ability to grasp it and bring it to the end?
Q: - YC primarily looks at the founders - how many units of work you can perform per unit of time. The main thing is how fast you can do, test, understand what does not work where and continue.
The second thing they look at is the market. If you do it very quickly, but in a bad market you will not swim far away. In very large markets, as a rule, there is high competition, so they try to find small but rapidly growing ones. We have, I believe, was the perfect timing. We made the product as soon as it became possible to make it. Timing is one of the most important pieces.
And the third is the product itself. It is important for them that people love him. It is better to make a product that hundreds love, than a product that thousands like. If you just like it, it will be difficult to convert. Much better to find a small community. We grew from zero to 14 thousand users organically simply because some people told us about others.
Just last week in our group on Facebook wrote a man with his real estate business. Now a popular phenomenon is when people look at home not with the realtor, but themselves. He wanted to make a skill for Alexa, put it in all the houses according to the device, so that clients could ask her about everything themselves. He appealed to the agency to develop such a skill, and there they asked him for $ 45,000.
He was upset and went to look for how to do such a thing himself. I found our product and made this thing - for which the agency requested 45 thousand - in just a few days. And so he bought our paid version, although he didn’t even need the features that were there (a good reason for us to think about new segments). He just loved the product and supported us.
What will happen with voice assistants in the coming years
Everyone has seen how buildings are usually built. They enclose the territory, and there they do something for a long, long time behind the fence. Then suddenly, in a matter of weeks, the skeleton of the whole skyscraper grows. From the side it seems that the construction has finally accelerated and is about to be ready. But when the skeleton is built, the work seems to slow down again. The construction has been settled for a long time, until the finished skyscraper is made.
With technology markets as well. Good finished products appear at the wrong time hype and hype. They appear later when the noise subsides. And then they become monopolists and change the market.
Now the voice technology is the skeleton of a skyscraper. There is a good recognition technology. Something appears on the market. There are 50 million speakers sold in the states. Google actions can be run on any Android smartphone, which is 86.2% of the smartphone market. Microsoft has a world of 500 million devices on Windows, and the use of voice technology is growing there by 50% from month to month.
But this is not a skyscraper. It will appear when there are new cool use cases for voice assistants (for example, ordering goods and services by voice), when people have a habit of using them. It will be in 18-24 months. That's when more interesting things start happening on the market.
And in 3-5 years it will be like this: you wake up, you have a device near the bed. You say, "Alex, good morning." It includes your favorite song, brews you coffee, opens the windows. You go to the kitchen, launch a podcast or news with your voice. You sit in the car, drive to work, listen to the podcast there. The assistant turns into a seamless operating system.
Everyone says that around the voice technology is HYIP, but it has not even started yet. HYIP will start in a year and a half, and then it will be really interesting. Next year, BMW and Lexus will come out with a built-in Alex. In America, Amazon integrates Alex into a set-top box, that is, it is already becoming a TV remote. Under Alex there are already about 40 thousand applications, and a couple of weeks ago, Amazon did a thing that allows you to run them without saying the name. So you say, “Alex, order flowers, and she herself selects applications that can do this.” I think this will be the number one thing in the future. First, you will not need to pronounce the name of the application, then the name of the assistant. He will be in the mode of "always listening" and contextually understand, if you refer to it.
- You were interested in the topic of linguistics - to teach a computer to speak, all that?
Q: - Yes, it is very cool. I remember in Startrek there was a computer with which they spoke. It was fun, but it seemed so futuristic. And this is actually happening.
There is a story. I'm going from work to training, I sit in Uber, and then the driver is talking to someone. I look - in the car there is nobody. It turned out he spoke with Alice in Yandex Navigator. And here behind him was some dude. The driver began to swear, and Alice thought that he was swearing at her - telling him, “how are you talking to me?”. The driver is such, "shut up there, generally not up to you." It looked so real. You think that the future is somewhere far away, but it is here, it has already come.
Of course, the extent to which devices and helpers can respond contextually is far from ideal. But speech recognition is getting better. If now Alex correctly understands 8 words out of 10 (according to my feelings), then in a few years he will understand all 10. This will greatly affect how people will talk with their device. Siri is super stupid, doesn’t understand anything at all, and this is very outraged. Voice assistant such a product - in order for it to work well, the technology itself must be good.
- You would not want to tie some kind of training, so that Alex understands not literally what you say to her, but use the context?
Q: - Partly it works in Storyline now. For example, there are intentions YES / NO, and a bunch of synonyms on the Amazon side are automatically added to them. If you set Yes, Alex will understand Ok, Fine, True, and so on. All this is done at the level of Alexa, and we use it as an interface.
“When will cars ever talk better than we do?”
Q: - This is a question about artificial intelligence in general. Turing test is now quite possible to pass, specifically to train the system for a narrow task. But speaking better than a person means understanding the context, emotions, intonation, combining everything to give an answer. I don't know, it's very hard for me to imagine.
I may be completely wrong, but it seems to me not earlier than in 10-15 years.