Andrey Horsev learned the product business one experiment at a time

In 2012, Andrey Horsev was asked what he made of people calling him an upstart from Dnipropetrovsk who fancied himself a venture investor. His answer, translated from Russian, was two words: “That’s true.” It was an economical way to deprive an insult of its employment. There would be no ceremonial defence of his credentials.

The less economical part had been getting there. Horsev said the transition from web studio to product company had taken roughly seven years and a dozen failed startups. Money from client work went back into experiments. His partner Aleksey Ivankin had devised Tested, a local collective blog. Other attempts ranged from games to disk-encryption software. A football site lost its way when the tournament ended and the team tried to turn it into general news.

Even Audiko, the ringtone service that helped finance further products, had an awkward early chapter: a data-centre accident took it offline for almost a week after it first attracted substantial attention. These were the unglamorous costs behind the investor’s title. The polished biography came later; the invoices and abandoned ideas had arrived first.

A classroom left behind, a business learned

Horsev was born in the city now called Dnipro, Ukraine. His early business was web design, a trade with a particularly direct examination system: somebody has to want the thing you have made. It is a useful starting point for a career subsequently occupied with persuading people to use unfamiliar products.

He studied radiophysics at Dnipropetrovsk National University and left in his third year. In 2014, he explained that he wanted to start his own business and no longer saw the point of continuing that course. Both his parents were teachers. His own education took a less institutional route.

He also made a distinction that complicates the familiar programmer-to-founder story. He had done very little programming for money, he said. His work had been building the business and learning management, first through website services and then through products of the company’s own. The practical curriculum was customers, colleagues and the consequences of decisions. It offered rather fewer holidays than a university prospectus.

The ringtone was the useful little thing

A ringtone is a modest claim on civilisation. It asks for a few seconds of music, at the right moment, on the right phone. Audiko built around that small wish: let people turn music into something their phones could use. Horsev’s June 2011 launch note supplies a pleasingly concrete picture of the business.

A quarter of visitors were making ringtones from their own tracks. The iPhone version distributed through Cydia had reached 600,000 copies; the Android version was approaching 100,000. These are figures from that period, when making a phone sound personal was still enough to sustain an internet product.

The next step was Audiko Everywhere. Website owners could add the ringtone-making service with a small piece of JavaScript. An integration on the site of the band “…and My Friend the Truck” let listeners click a bell beside a track. The detail is more revealing than a grand declaration about distribution. Horsev’s team was moving the tool closer to the music, giving another website a reason to carry it. A bell did the explaining.

California, with the door open

In November 2011, Horsev spent a couple of weeks in Silicon Valley. He had not expected to get there that year; obtaining an American visa had previously seemed uncertain. Once there, he was struck by the appetite for talking about projects, the frequency of gatherings and the sense that people might help one another.

He noticed something else: a failed project could earn respect for the experience it produced. Someone had explored a route and could warn others where it led. After years of experiments, that was a professional culture he could recognise.

He wanted key members of the Dnipro team to experience the same setting. He and his long-standing partner planned to alternate time in the Valley while developing new projects. The journey was partly about shared context: the people building together should have a chance to see the same opportunities. California supplied an address, but also a room full of possible conversations.

The map kept changing. In September 2015, 908 moved into an office in central Wrocław, Poland. Two years later, Horsev announced that he had closed that office. The company retained a presence in Dnipro, while employees could also work in Thailand. Asked why he would not simply return to Ukraine, he said, translated from Russian, “I was born and grew up there, and I want to see the whole world.” The movements gave his working life an international geography long before it could be reduced to a location field on a profile. An office and a home city did not tell the entire story.

A marketplace has to earn its return visit

The internet is also made of less fashionable errands: selling children’s clothes, finding a buyer beyond one’s town, answering a message about a listing. Horsev launched Klubok in 2014, following the shareholder dispute surrounding the earlier Klumba project. By July 2016, he reported 700,000 registered users and roughly 100,000 visits a day, with a seasonal summer dip.

He described a marketplace that sometimes had to teach its customers how to use the service. Some sellers passed their clothes to more internet-confident acquaintances to list for them. That is a useful corrective to the assumption that every customer arrives knowing what all the buttons mean.

Advertising and promoted listings paid the bills. Mobile development began with messaging and gradually added functions. Horsev said the business was operating around break-even and reinvesting its earnings. His warning to would-be marketplace founders was to allow at least five years of their lives for the undertaking. The admission price was time. A marketplace can have its launch party in an evening; getting strangers to come back is a longer engagement.

The admission price was time.

ON BUILDING A MARKETPLACE

Before the bot had anything worth saying

On 29 March 2016, Horsev put a question to the DOU developer community: had anyone been experimenting with Telegram bots? He mentioned the expected arrival of Facebook’s messenger platform and invited people interested in the market to talk. The post catches a founder while a subject is still a question, before it has acquired a case study and a conference slide.

The following year, he was ready to be critical of the category. His essay asking why there were so few good bots considered conversational AI, scripted dialogues, forms, storefronts, search and payments. Simply pushing an existing activity into a chat window did not automatically make it useful.

His preferred case was personal notification: information a user actually needed, delivered when it changed. As he put it, translated from Russian: “A good bot’s scenario is a notification that a particular user configures for themselves.”

Opendatabot fitted that argument. It monitored public records so that users could learn about changes involving businesses. In July 2017, Horsev reported 52,000 users, including 15,600 subscribed to various services. The appeal was a practical one: let the software watch, rather than asking someone to keep checking. The bot could justify interrupting because it had a reason to speak.

Stickers, state records and a train to catch

The same year, his conversation with Roman Rybalchenko ranged across Audiko, Klubok, Stickerpipe and Opendatabot. Stickerpipe brought him into stickers and messaging; the discussion covered branded content, product testing and monetisation. This was a portfolio assembled around different sorts of everyday behaviour, from shopping to communicating.

He was also co-hosting the technology videocast California Today with Andrii Yasinetsky. Recording proved difficult when both travelled and their time zones kept changing. That is a recognisable indignity of an international working life: however interesting the subject, two people must still be awake at the same time.

By November 2020, Horsev told Mind that Opendatabot had 2.4 million users across its platforms. The service, launched in 2016, had added a mobile application in 2019. Its information came from more than 20 open state registers. The audience and the subject had grown beyond the early experiment in a messenger.

Another application of the technology appeared in RailwayBot. In February 2020, its founder Anton Biblia described using the pipe.bot platform to build mobile apps. Horsev reported that users had bought 50,000 train tickets through RailwayBot during 2019. A ticket found and bought is an unusually clear verdict on whether a digital convenience has served its purpose.

A wooden cube, and a familiar instinct

Horsev’s 2018 conference biography grouped him with Audiko, Klubok, Stickerpipe, Pipe.bot and Opendatabot, and named investments in PosterPOS and JustShareIt. Then it added a detail with rather less boardroom polish: in 2016 he had connected a wooden cube to the internet. A crowded professional introduction had found room for a piece of timber.

His public Stack Overflow profile offers a smaller, equally particular glimpse. Describing himself as an entrepreneur and hacker, he contributed an answer in March 2019 about opening a Bootstrap modal with Vue.js. Between the company names and the stage appearances sits an ordinary technical problem, with someone supplying an answer.

His LinkedIn profile lists Mountain View and an organiser role with Hackers & Founders Ukraine beginning in November 2011. A recommendation from Max Ischenko calls him straightforward and direct. Those details sit comfortably beside the terse reply to the upstart question, though a public recommendation is still one colleague’s view of a person.

The company website continues to list Horsev as 908’s CEO and describes its work as growing products from ideas, with investment available at the working-prototype stage. After ringtones, clothes, stickers and public records, the prototype remains an appropriate place to meet him: something has been made, somebody can try it, and the next conversation can begin.