Vibe Coding: Know What You Want

Vladimir starts from the very beginning: a family of Soviet engineers, a father who graduated from MEPhI, a grandfather who ran institutes, and an ordinary neighbourhood school, not an elite one. Childhood already holds the pattern behind the whole career: he reaches 80 per cent of the result with 20 per cent of the effort, the remaining 20 needs another 80 per cent of the time — and the interest runs out. That is how skiing ended (he moved to a snowboard and buys no gear at all, riding a couple of days on each) and the piano (he never started on solfège). In 2009, the first year of the unified state exam, he went to the aviation institute — the boy at the next desk was going there and the tram was convenient — and picked the hardest faculty, applied mathematics; functional analysis he sat through with completely glassy eyes. What the institute really gave him: the words I do not know are banned outright; Alexander softens it on the spot — allowed only as a promise to come back tomorrow with the answer.

The mobile wave caught him in his third or fourth year: he turned a lab assignment into an app, and through a vice-rector joined the people writing the institute's own app with timetable and news, a wow by the standards of 2010–2011. He arrived at Sunlight as its first mobile developer, with the word manager in his record from day one: in two and a half years the team shipped three apps — the customer one, one for shop staff (instead of corporate hardware, the phone already in their pocket) and content apps — and he invented the analyst, team lead and manager roles himself, because nobody in the company understood mobile development. Then came My Perekrestok and the argument whether a retailer needs one app or two; the main X5 lesson was delegation: past twenty people you stop climbing into every repository. At Lenta in 2019 they launched express delivery in three months, though everyone thought it mad, and in 2020 stood it up in Cherepovets in a week as the town went into quarantine.

After Utkonos, where he became CTO and warned the team honestly that he was more of a manager (six months later they admitted their shock: in 2021 a CTO was still expected to write code), Vladimir chose not to wait for a chair to free up but to restart the game as a new character, the entrepreneur. He calls March 2022 extreme: it began with an app long written in Python that never flew, like 99 per cent of such apps, and the practice slid into consulting, where the employment contract simply became a contractor agreement. He put the main mistake into words only a year later: he left with the mindset of an employee selling his hours. What he gained: an understanding of where salary money comes from, and he now treats shareholders' money as his own. He returned to employment with a child on the way: risk appetite, he says, is set by the cushion — how long you last if the financial tap is shut.

Earlier still, his record picked up three months at Ozon, where a cartoon-cat picture saying it is fun and scary became his motto. His advice to friends has changed: until 2024 he told them to move, now he tells them to calm down, because it is much the same everywhere, with no guarantee you land where you expect; he flags this at once as a personal value judgement. Alexander agrees that decisions are now made without rose-tinted glasses: the era of cheap money is over. In April 2024 Vladimir returned to fintech — he came in as CTO of an instalment-payments service with a fully outsourced team and built in-house development within six months, and now, as operating director, he sits deep in sales, risk and marketing, because instalments are a business about risk.

He took up vibe coding chasing a digital asset of his own rather than a rented coffee kiosk where a cup costing 50 roubles sells for 300 until the lease runs out. He tried it on holiday in May and got hooked, comparing the feeling to a slot machine: you feed in coins, the outcome is unpredictable and you keep pulling — and paid 200 dollars for a subscription, probably one of the first times he had ever done so; Alexander recalls paying the same 200 dollars to work through the DORA methodology with a model, having assembled 80 per cent of the calculation logic by hand. The demo is deliberately dull — a to-do list, so the whole pipeline fits into ten minutes. First ChatGPT: I want a simple app plus a short spec, then a request for a database schema; the model asks about the backend, he answers Supabase, asks for a script that creates the database in one query, eyeballs the tables and runs it.

The whole result goes into Lovable with a request to integrate Supabase: the tool asks clarifying questions as a quiz, writes a plan — and five minutes later, which he calls slow, hands back an app with authentication, persistence, editing and filters nobody asked for. Search is added live on air, the model itself proposing debounce and case-insensitive matching, then a dark theme, and at the end it draws a flowchart of the architecture and names the bottlenecks. The lesson: you have to know what you want — those who come to him saying it did not work are missing the groundwork, and the breadth of a C-student now beats narrow specialisation. Asked whether this can go to production, Vladimir answers that today it looks like a monkey with a grenade, yet progress is unstoppable, just as the assembler crowd gave way to those who built C and Java. Alexander replies that responsibility stays with the author: in his company an internal agent assembles services, but engineers sign off production readiness; Vladimir agrees, arguing only with the tools' remaining doubters. The closing stretch is personal: thirty kilograms lost in a year, a ban on working after eight in the evening, and the advice to attend gatherings outside your circle.