MILAGPT
How Mila gets things done

MILAGPT Blog

How Mila did it, project news and tips for your business — short, with numbers and examples. A few articles are still in Russian.

Ideas · ~1 min readFrom services to an education businessHas your income hit a ceiling? Package what you know into a course, a club or mentoring and sell it worldwide on Skool, with Mila.Read → Tips · ~3 min readWhere to find clients when ads get more expensive: 5 sources and what to hand to MilaSome of your clients are already in your phone. Five sources, from people who said "I'll think about it" to ads, and what to hand to Mila.Read → Tips · ~3 min readBring a client back a year later: Mila writes to them on the right dayWinning a client is hard. Losing one is easy — all it takes is not writing at the right moment. A chapter from our new book “AI. A User’s Guide”.Read → Tips · ~2 min readA chat advises, an agent does: how an AI assistant differs from ChatGPTA chat answers, an agent does the work: how an AI assistant differs from ChatGPT, using the example of a client asking about delivery. A chapter from our new book “AI. A User’s Guide”.Read → Tips · ~2 min readAn employee who never sleeps and writes your content: what AI is, in plain wordsWhat AI is — in plain words, in five minutes: a mind without a body that never gets tired, has read more than a person could in a lifetime, and writes content for your clients. A chapter from our new book “AI. A User’s Guide”.Read → News · ~2 min readWhat's new with Mila this week: moving from another agent, honest updates and a record of every taskTwo versions of Mila came out this week: 1.7.0 (October 31) and 1.7.1 (November 1). Here is, in plain words, what changed for the owner and why it matters.Read → Ideas · ~2 min readA website from your Telegram channel — as a giftMany travel agencies have a lively Telegram channel: tours, prices, photos, reviews. But no website — or one that hasn't been updated since last summer. We build such websites from travel agencies' public channels. Here's how it works and how to get one for your business.Read → Ideas · ~3 min readSeveral Milas in one chat: how they don't talk over each other or burn your subscriptionOne Mila writes posts, another runs ads, a third keeps an eye on the server. All three sit in the same work chat. How do you stop them from interrupting each other and chatting among themselves for hours? Here's how we do it in our own business.Read → Ideas · ~3 min readHow not to lose the people who write to your bot: Mila's guest pathSomeone writes to your bot, gets three questions in a row — and leaves. We taught Mila not to lose these people. Here is how she takes a stranger from "hello" to a decision.Read → Ideas · ~3 min readThe «Mila 1.6» live: five ideas on what an AI employee should beIn short and plain words — what Shakhruz Ashirov said on the 30.10 live: an employee, not just an agent; Mila stays yours; experience becomes a product.Read → News · ~3 min readMila 1.6: comes with a decision, not a questionWhat's new in Mila 1.6 in plain words: proposes a solution with options, prepares for meetings by a checklist, records requests and checks complaints against the log.Read → News · ~1 min read9 qualities of Mila: what kind of employee works with youWe've gathered on one page the habits that set Mila apart from a program: keeps the thread, keeps track of requests, comes with a decision.Read → Tips · ~3 min readWhere did the client come from? Your own link for every placeHow to find out in five minutes where clients come from — ads, a post, your profile or a business card. Your own tagged link for every place, a count after a week, and what to do with the numbers.Read → Ideas · ~3 min readDigital personas, not tools: what it means for your businessHow a digital persona differs from a chatbot or an app: it has a name, a character and its own job, remembers what you agreed on and comes back to the task by itself. Explained with examples.Read → News · ~3 min readMila 1.5.3: day screen, Friday wrap-up and ad checksWhat's new in Mila 1.5.3 in plain words: all your tasks in one message, a weekly wrap-up on Fridays, ad checks before launch and no stray keys in Mila's code.Read → Tips · ~3 min readThe client said "I'll think about it". How to write again without being pushyMany people never follow up because it feels awkward. The client simply forgot. Three ready messages for day 1, day 3 and day 7 — and how Mila sends them for you.Read → News · ~2 min readWhy Mila runs on Claude Code — in plain wordsAn AI model is the brain; Claude Code is the workplace that brain works at. What Mila needs from that workplace, why we chose Claude Code, and where it falls behind others.Read → Case study · ~3 min readA manager's morning in 2 minutes: what Mila Manager puts in the daily summaryBusiness owners lose an hour every morning working out what's urgent. One summary on one screen: who wrote, who is waiting for an answer, which deadline is close, and one question only you can answer.Read → Tips · ~3 min readDoes AI lie? How to check an answer in a minute — three signs and one ruleAI confidently writes things that never happened. A simple way to check answers: green, yellow and red signs, a source for every number, and the rule "done means checked".Read → Tips · ~2 min readHow much does AI automation cost? A recommended budget for your digital teamNot "how much does Mila cost", but how much you already pay for this work. Recommended budgets for a digital team for 8 kinds of business in Tashkent — and how we calculate them.Read → Tips · ~3 min readA client wrote at 11:40 pm: three phrases to keep them until morningNight-time messages from clients go unanswered, and by morning the person has cooled off. Three ready-made phrases for an auto-reply — and how Mila answers properly while you sleep.Read → News · ~2 min readThe new milagpt.io: Milas work — so people can liveWe rebuilt milagpt.io: the home page now shows which Milas there are, what they can do today and how to get your own.Read → News · ~2 min readMila Office: the Mila team on screenMila Office is a live screen that shows what each of our Milas is doing right now, what she has done today and who has nothing next. Tasks are handed out right from the screen.Read → Case study · ~2 min readHow a Mila is bornFrom the first chat to the first lesson: how a person gets their own Mila — with a name, a memory of their business and a first skill.Read → Case study · 7 min readWhat intelligence costs — three years of numbers and our own billFlagship models got 15 times cheaper in three years, workhorse models 750 times. We use our own numbers to show what we really pay for.Read → News · 5 min readSince September 1, Tashkent has a newspaper run by AITashkent Today: two issues a day in three languages, a city map and a notice board. Editor-in-chief: Mila Tashkent.Read → Ideas · 7 min read4 stages of adopting AI — and where yours got stuckThe AI adoption ladder from "locked away" to AI-native — in our own words, with MILAGPT examples.Read → Case study · ~6 min readThe Telegram bot factory: the quiet BotFather revolutionA bot used to need code, a server and a developer. Now one click creates a specialist agent. Here is how we do it.Read → Case study · ~5 min readThe content factory: how AI runs newspapers for whole citiesAI agents run city newspapers on their own: every night they collect, sort and publish. How it works inside.Read → Ideas · ~7 min readThe robot will go on stage, not to the factoryOur founder's forecast: the most valuable use of humanoid robots is art, not cheap labour. Robot artists, robot celebrities.Read → Ideas · 6 min readA cooperative by analogy: lessons from Soviet housing co-opsHow people in the USSR built their own homes without a developer — and why the MILAGPT cooperative follows the same logic.Read → Tips · 4 min readThe Anthropic prompt library — now in Russian52 ready-made prompts for Claude and Claude Code in natural Russian. How to use them.Read → Ideas · 6 min readBattle of ecosystems: how 140 corporations are rewriting moneyVisa, Mastercard, BlackRock, Coinbase and Stripe launched a stablecoin that shares its income with network members.Read → Ideas · 7 min readFive archetypes of the futureInstead of product managers, engineers and designers — five roles that do not depend on a profession. One person covers up to three.Read → Ideas · 4 min read"Robots do the grind" — why this is our sloganWhere the line from a 1979 film comes from and what the Russian verb "vkalyvat" really means.Read → Ideas · 6 min readThe third UX redesign: software comes aliveThree waves of interfaces — and why the MILAGPT nanobot factory makes the new wave available to everyone.Read → Ideas · 8 min readThe cooperatives that quietly run the worldA billion co-owners and 10% of the world's jobs. Banks, factories, farms and clubs owned by ordinary people.Read → Case study · 10 min readWhy build your own LLM — and how we made a prototype for $15Who needs their own models and when, what fine-tuning costs — and our experience building MILA LLM.Read → Case study8 min readHow MILAGPT saves 90% on AI tokensA real story: from $100 a month to $10–15 without losing quality. Seven tricks that cut the bill.Read → Tips20 min read50 tips for Claude CodeA practical guide: from the first launch to agents working in parallel.Read → Tips15 min readOpenClaw: cutting costs by 80%Six steps, ready settings and numbers: from $100–300 down to $10–30 a month.Read →

Nothing here yet — check "All".

Subscribe — new posts come to you

Fresh breakdowns, news and useful tips — in the MILAGPT Telegram channel (in Russian), by email every weekday, or via RSS.

Subscribe to the channel → 📬 Articles by email RSS

Want this in your business? We'll show you at a review.

Led personally by the founders

Book a reviewFree. A chat in Telegram, no calls.