Writing
What I have learned building AI/ML products that help strangers transact.
- Turning conversational data into a north-star metric: classify contacts, score them cheaply, move them in the messenger.
- Every career trades craft for leverage, and the curve has a valley in the middle where the first-line manager stands. The two winning ends, the two currencies of leverage, and why AI is tilting the curve toward building.
- The published return envelope, the five-layer architecture, per-mechanism economics with minimum data bars, stage gates with kill criteria, and five platform scenarios for where the case clears the bar.
- Generating copy is free now; choosing the winner is the work. The five-stage pipeline, and the one loop you never close.
- The messenger is the conversion engine of a marketplace, not a chat feature. The AI/ML behind it: intent and inbox ranking, suggested replies, negotiation, the agreement signal, generative assistants, and the trust layer underneath.
- What personalised communications return, measured against a holdout; the four platform shapes; and the order to build the stack in.
- How modern buyer-engagement platforms use AI to decide — in real time — who gets a message, what it says, and when and where it lands: the trigger → modeling → content → orchestration → placement → feedback loop, the four personalization layers, and the results benchmarks. Published on HackerNoon.
- A tactical playbook for taking a consumer health & fitness subscription app from 0 to $1k to $10k monthly tracked revenue — discovery, MVP definition, build stack, store launch, and exit benchmarks, with the unit-economics maths. Published on HackerNoon.
- A practitioner's guide to the hard and soft skills a product manager needs — the three PM archetypes (technical, general, growth), product-building, monetisation, and the leadership skills that matter most at senior level. Published on Sostav.ru (in Russian).