908: Small bets.
Working software.

The first thing to break was somewhere else. Audiko, the ringtone service built by Ukrainian company 908, had attracted attention when a Houston data-center accident knocked it offline. For almost a week, the young product disappeared. In a later account, co-founder Andrey Horsev recalled fearing the interruption would kill it. The service returned. The team eventually moved away from that provider. A little business selling a little piece of a song had acquired an expensive lesson in dependence.

01 / THE FIRST ENGINEThe ringtone that paid for the next idea

908 began as a web studio in Dnipro, then known as Dnepropetrovsk. Horsev and fellow founder Aleksey Ivankin built things for customers while trying to invent something they could own. Their obstacle was familiar to anyone billing by the project: a finished website brings a finished invoice. The next customer brings another beginning.

In 2012, Horsev put the transition at roughly seven years and ten unsuccessful startups. Audiko became an early commercial foothold. Its income helped support other products. That history makes 908’s investment pitch more interesting than its short website suggests: the company learned product development while paying for its own mistakes.

There was no single monetization formula. In an October 2012 essay, Horsev remembered Audiko initially making about $25 a day through AdSense. He disliked what the advertisements did to the interface. Revenue made the annoyance tolerable. Design taste had encountered a customer of its own: the business model.

For readers running an agency, the distinction is consequential. A paid project can finance a speculative one, but only if somebody decides how much money and developer time the experiment receives. Otherwise the experiment competes with a customer who has a deadline. 908’s long transition makes that trade-off visible. Owning the code does not automatically create a market for it.

02 / CAPITAL WITH HANDSA check with engineers attached

908 occupies an awkward, useful space between a software shop and an angel investor. Its website offers money at the working-prototype stage, backed by its own development operation in Ukraine. A founder arriving there is looking for help getting a product into use, as well as funding.

“Our goal is to grow successful products from ideas.”

908 / THE COMPANY’S STATED GOAL

The California expansion made that proposition explicit. In March 2012, 908 announced an office there and an investment in JustShareIt, a vehicle-sharing business. The founders described an intended investment range of $20,000 to $100,000. Those were the ambitions of that moment; they should be read as historical terms.

Poster offers a more tangible example. When 908 invested in 2013, the restaurant point-of-sale project was still taking shape. Its founders said the investor helped turn a prototype into a fuller product. An early test ran in a Dnipro tea house. Orders, waiters, and a working venue supplied feedback that an investor presentation could never provide.

03 / A DIFFERENT SORT OF ALERTThe problem came home

The route to Opendatabot began with a less cheerful kind of feedback. Ivankin has linked the idea to the bitter dispute around Klumba, a parenting website in which 908 had invested. The shareholders made conflicting allegations. His subsequent product question was concrete: could a business discover changes in official records quickly enough to respond?

An opening arrived in April 2016, when Ukraine released company-register data. The team rapidly built a bot that could notify people about changes. Opendatabot developed into a service for searching and monitoring public records, including companies and court information. A director’s name changing in a register became something a phone could tell you about.

The customers include individuals doing checks and businesses embedding data in their own systems. Banks, legal practices, retailers, logistics firms, and agricultural companies appear among the sectors Opendatabot serves. Its API brings records into existing business software, reducing the need for repeated trips through separate registry websites.

For a developer, the API is a way to automate checks on counterparties and monitor changes without building every registry connection separately. Opendatabot describes JSON responses and ready-made libraries. The purchase is partly convenience: the same underlying public information can become a recurring business task, performed inside software a company already uses.

The economics become clearer at product level. In January 2025, Ivankin described services starting at 400 hryvnias per month for an individual customer, with much larger corporate arrangements. He reported five million chatbot registrations. That is a cumulative audience measure, not five million people showing up every month.

04 / LESS FRICTIONDifferent products. The same useful habit.

Pipe.bot tackles another access problem: building a chatbot. Its interface starts with a mind map. The product advertises connections to messaging platforms, external APIs, documents, spreadsheets, and analytics. A team can map a conversation and connect it to other tools without treating every channel as an entirely separate project.

908’s employer profile places Pipe alongside Opendatabot and emphasizes readable code, developer experience, and removing unnecessary production steps. It also describes support for Architecture nights in Dnipro. The common thread is practical engineering: keeping the people who understand the product close to the work of changing it.

That emphasis gives a prospective colleague something specific to judge. Does an engineering organization make its software easier to understand and update? Those priorities become consequential when the company owns a product over years, rather than handing over a finished website and moving on.

05 / WHAT TRAVELSCopy the discipline

For a service founder, the transferable idea is to reserve money and time for experiments before the next invoice consumes both. For a startup, it is to ask an investor what happens after the transfer clears. Poster’s tea-house test makes a useful standard: can the product survive an ordinary customer’s ordinary working day?

The conditions matter. Experiments need spare cash and protected engineering time. A prototype needs users willing to try it. Public-data software needs accessible, usable records; a clever interface cannot manufacture an unavailable update. Read this way, 908’s story is a series of decisions about what to make repeatable. The ringtone, the restaurant order, and the registry alert each gave that decision a different test.