Field notes on building products with AI.
Practical articles about Telegram Mini Apps, AI agents, MVP development and business websites: costs, timelines, common mistakes and lessons from real projects.
AI native: what it means and why the word artificial gets in the way
AI native means AI is built into the work rather than living as a separate project. We look at the word itself, at what artificial means in English and in Russian, why vibe coding replaces neither the programmer nor the employee, and where an AI-native company starts.
ReadAI adoption in a company: where to start when nobody knows where to start
AI adoption means bringing AI into the daily work of a company. Step by step: who should build the first agent, which process to pick, what stays manual (registrations, tokens, access), what it costs and which mistakes we see most often.
Readminimir on Zenodo: a preregistered artificial life experiment with a DOI, and what protocol B2 changes
minimir, my artificial life experiment, now has a citable DOI on Zenodo, the open research archive run by CERN. What protocol B showed, how it was run, and how B2 is designed.
ReadBuild log, 27 October: 26 clicks, a subdomain Google never crawled, and a registry with 24 lines
Weekly build log: Google Search Console numbers for the site (26 clicks, 1,720 impressions, 132 impressions in AI features), why the minimir subdomain was never crawled, what changed in the project registry and on GitHub, and what comes next.
ReadFelix: a Telegram bot and Mini App for event requests
How I built Felix for venue and equipment enquiries: guided questions, catalogue matching, indicative pricing and manager handoff across a Telegram bot and Mini App.
ReadInside my @ilinmaksbot: project questions, an estimate and a brief
How I built @ilinmaksbot: project questions, an initial estimate and a brief sent to me in Telegram. A three-language portfolio and the limits of automation.
ReadAI Engineer vs Forward Deployed Engineer: tasks and ownership
AI Engineer vs Forward Deployed Engineer: product work, customer implementation, shared skills and ownership. A practical Telegram enquiry example.
ReadI put my own face on the front page. How I launched Likeness House
How I built Likeness House: a bilingual Node.js website with enquiry forms, client accounts and an admin panel. A case study from my own product.
ReadHow I built Dantiva and AONETHA: one sentence in, a video out, and the first paying customers
Dantiva turns one sentence into a video, an image or a text, with Veo and Nano Banana inside. Dantiva serves the Russian-speaking market through the web, a Telegram bot and a mini app; AONETHA is the international version with Stripe. First purchase on September 17, first subscription on September 21.
ReadA tiny world with no rewards. What minimir is testing, and why a first null result still counts
minimir is an artificial life experiment: simple creatures, tiny neural networks, no rewards. It asks whether wanting appears when a creature can sense itself. The first measurement shows a signal lost in noise. What that means and what comes next.
ReadA city where nothing hurries you. Zagreb Pomalo
A browser game about the real centre of Zagreb. A day lasts 48 minutes, a hostel bunk is 18 euros a night, a room of your own is 450, and that arithmetic is the plot.
ReadMy Lovable site looked finished. Google could read one language out of four
What the first Lovable version of a robot catalog got right, what Google could not read, and the three checks to run before you launch a site that lives on search.
ReadDantiva got its first paying customer. Then the product changed again
A founder update on Dantiva: the first confirmed purchase, the first monthly subscription, a rebuilt light workspace, a Gemini LLM studio, GPT Image 2.5, and two Croatian innovation applications.
ReadThe model is the easy part. What building Dantiva taught me
Build notes on Dantiva, an AI photo-to-video studio on Google's models: three front doors, a reserve and refund ledger, a rename I refused to finish, three pricing rounds, and the wall the product hit.
ReadI'm not an engineer. I fine-tuned my own language model on a MacBook Air
How I fine-tuned Qwen3 8B with LoRA on a MacBook Air with 16 GB, four attempts, 576 examples, where the metrics lied, and what the model actually learned.
ReadHow much does it cost to implement an AI agent in 2026
What drives AI agent cost in 2026: process, integrations, data, evaluation, hosting and model usage, and how freelance and in-house compare to a scoped project.
ReadAI agent or plain automation: where the real ROI is
When simple automation is enough and when an AI agent pays off: ROI in hours, response time and errors, and why Gartner expects 40% of agent projects to fail.
ReadThe AI job market in 2026: who gets hired, what they earn, and which roles are fading
AI hiring in 2026: demand for AI Engineers and FDEs, US salary benchmarks, European job data and junior hiring. Sources and data periods explained.
ReadThe EU AI Act for small businesses in Croatia: what to do before 2027
What the EU AI Act means for a small business in Croatia: risk tiers, deadlines to 2028, duties for chatbots, CV screening and generated content, a checklist.
ReadWhat a Forward Deployed Engineer is and why a small business should care
What a Forward Deployed Engineer does for a customer and how a small business can arrange AI implementation: process, data, evaluation and handover.
ReadPrompt engineer is not a job anymore: which AI skills the market actually buys
Why the prompt engineer title dropped out of hiring plans, what it pays in the US and Russia, which skills the market buys instead and who to hire.
ReadA RAG chatbot on your company knowledge base: what it is and when it pays off
RAG in plain language: which businesses benefit from a bot that answers from their documents, what to prepare, where projects fail, why a pilot takes two weeks.
ReadHow I build websites with Lovable and get them ready to launch
What Lovable is, when it works well for websites and MVPs, and why a strong first draft still needs product thinking, SEO, testing and launch work.
ReadParser, bot, service or agent: what to automate a process with
How to pick the right AI automation tool: when a parser is enough, when you need a Telegram bot, when an internal service, and when an AI agent. Timelines, budgets and the usual mistakes.
ReadAI agents for business: what I learned building my own team
A practical guide to AI agents for business: how agents differ from chatbots, which processes to automate and how to run a safe first pilot.
ReadHow AI search engines decide what to cite
What makes ChatGPT, Perplexity and AI Overviews pick a source, and the page changes that get a small site into those answers.
ReadWho to Hire for a Website, Bot or MVP in Croatia in 2026
Zagreb agency, freelancer, offshore team or a solo product builder: real prices, timelines and the questions to settle before you sign.
ReadWhat You Can Actually Ship in Two Weeks with AI Tools
A realistic two-week scope with AI tools: a landing page with a lead bot, an MVP with auth and one core flow, or an AI agent for one process.
ReadSolo Builder or Agency: Who Should Build Your MVP?
A practical checklist for picking between one strong builder and an agency for your MVP, site or bot, with 2025-2026 rate data and questions to ask.
ReadTelegram Bot vs Telegram Mini App: What to Build
When a Telegram bot is enough and when you need a mini app: platform limits, build time, real project examples, and how to upgrade later.
ReadWhat a Telegram mini app costs in 2026
Price bands for Telegram mini apps and bots in 2026, what pushes the number up, and how to scope a first version that ships in a few weeks.
Read