We started with 40 trucks and a spreadsheet. Every load came from a chat; every price was negotiated by phone. The first 'platform' was a Google Form. What actually moved the margin was not the app but a corridor-density rule: we only take loads where we can guarantee a return trip within 48 hours.
Show full postPaid ads brought sellers who never listed a product. Bazaar visits brought sellers who listed everything. The thing that finally worked was boring: a weekly phone call from a real person for the first month, and a 'trust score' visible to buyers.
Show full postNot the TAM slide. I look for one number the founder knows by heart, one customer they can name, and whether they say 'we tested' or 'we think'. Most decks in Tashkent are strong on vision and silent on the last 30 days.
Show full postThe first three were brilliant and left within a year. The next seven were curious generalists we trained ourselves — six are still here. The interview that predicted it best: a messy real task, and watching which questions they ask.
Show full postWe thought the hard part would be the product. It was paperwork: a local partner, a free-zone licence and a bank that would open an account for a company with no Gulf history. What unlocked the first contract was a pilot priced below cost — and a case study we could show the second client.
Show full postLesson one: Latin and Cyrillic are two languages for a model, not one. Lesson two: a thousand clean dialogues beat a million scraped pages. Lesson three: users forgive a wrong answer, but not a wrong tone — we rewrote the system prompt more often than the code.
Show full post
