Artöm Mazurchak: posts tagged AI

Simulation game: “How will AI develop over the next 5 years?”

Mon, 02 Mar 2026 06:36:07 +0200

At the weekend, we spent two days on a simulation all about how AI might evolve over the next five years. There were 36 players involved. The game itself is a test and workout for skills like leadership, dealing with uncertainty, negotiation, building coalitions and cascading teamwork. Last year, I took part in a similar simulation, but that one focused on a geopolitical conflict.

Here’s how the game worked

Everyone was split into teams, each representing a key player in the market:

Established IT leaders – companies like Google, Microsoft, Meta and Alibaba;
AI startups – like Anthropic and OpenAI;
Chipmakers – companies like IBM and Nvidia;
Venture investors -GP and Standard Investments;
Conglomerate – X, led by Musk;
Regulators, the press and the courts.

35 people were split into teams and each person got assigned a role with details like personal goals, company status, industry position, team relationships, company ownership percentages, etc.

Each team member had their own “stats” like the number of people on their team, computing power, budget, accounts receivable and more.

There were 5 rounds, each lasting about an hour and a half. You could do pretty much anything: lure talent away from other companies, sue someone or get regulators involved, raise funds to build new factories to boost computing power. Any move was allowed, but moderators, the court or regulators could step in and veto an action.

At the end of each round, all the agreements made were handed over to the moderator, who compiled the results into a spreadsheet for each company. The spreadsheet tracked key stats that changed during the round, covering everything from team size to available cash.

The main goal for each team was to reach singularity, they used a Scrabble-like mechanic for that. At the end of each round teams got letter tiles, the number of letters depended on team size and computing power. But the exact formula of how letters were awarded was a secret, known only to the organizers and it wasn’t a straightforward calculation.

Post-game takeaways

1. The role you got is just a formality.
My role was a team lead at a company like OpenAI, not a top position. When I started negotiating with other teams, they asked what my position was and if I could make decisions. But I explained I was there representing the team and my role didn’t really matter. Very good. When it came time to run for president, I thought, there are already bigger names in the game, like Musk, so why even bother? Roles are just labels, both in the game and in real life. But it’s easy to use them as an excuse.

2. Always lead, always.
At the start of the game, we spent way too much time explaining what everyone planned to do. But as soon as we nailed down a clear vision of the goal and just got moving. Coming in with a plan and sharing it with the team really helped. I think the long talks happened because people were afraid to work with other teams and didn’t want to take responsibility alone. They wanted to share it so no one felt fully in charge.

3.If you don’t get the result you want, don’t overthink it – keep going.
When my or the team’s actions didn’t pay off, I got caught up in my emotions. But a few times when I pushed through and made an extra effort, things surprisingly turned out well. I’m more convinced than ever that keeping a cool head gets you results faster.

4. Look at your social connections before you start and use them.
At the start of the game only three teams had computing power. The first two teams I asked didn’t want to work with me. But the third team had someone I already knew and they agreed to cooperate. I’m sure the trust we built before the game helped a lot. You should use this kind of trust and manage it, both in the game and in real life.

5. Keep the whole game in mind while working on the current round.
It’s normal to want quick, predictable results. But if you only focus on the short term, your options get limited. When you plan for the long game, looking ahead several rounds, you open up more possibilities and better chances to succeed. For example, we started building a coalition in the first round and didn’t finish it until the third, but that gave us a lead over the other teams.

6. When things are uncertain, focus on finding opportunities.
In uncertain situations with lots of different interests, it’s easy to just stick to your own goals – your role makes them clear and it feels simpler that way. But if you pause for a second, you’ll see everyone starts like that and the only way to win is by working together. So from the very beginning, look for chances to boost both your position and the other team’s. In the end, the team that began building a coalition right from round one won.

7. First, get the context – then take action, but don’t take too long.
The roles and rules are detailed and you get all the info just before the game starts, so it’s impossible to understand everything right away. My team spent time trying to figure it all out, but we ended up talking too much and wasting time we could use to do something. You learn the context by engaging with others, not by staying stuck in your own little world.

8. Keep taking action until you get a confirmed result.
When several teams are involved in a deal, it can feel like you’ve already agreed and can move on to the next step. But deals have fallen apart more than once right at the signing. So until the agreement is actually signed, don’t get distracted, follow through all the way to the end.

9. Split up the responsibilities.
More than once, our whole team went to negotiate with another group, all of us listening at the same time, which wasn’t very efficient. In the game, just like in real life, there are lots of tasks and layers to handle. To work better, it’s smarter to divide the responsibilities and work on things at the same time. On the second day we did that and our team got way more efficient.

10. Stay in sync and make sure everyone’s on the same page.
When signing some deals, it turned out that key people had different ideas about how things should work and that ended up breaking the deal. The main points and strategy need to be clearly explained from the beginning.

Thanks to Marat Atnashev, Timur Atnashev and their team for organizing the game.

How venture studios help big companies survive the AI era

Thu, 26 Feb 2026 06:43:00 +0200

A venture studio is an organization focused on building and scaling startups. Unlike accelerators, studios are typically specialized in a specific industry and offer deep operational support to the startups they launch.

Sometimes, large corporations launch internal venture studios. The biggest challenge they’re trying to solve is “the innovator’s dilemma”. This concept, described by Clayton, explains how big companies often miss out on new markets and breakthrough technologies because they’re too focused on protecting their current market share. And right now, many of those breakthrough technologies are connected with AI.

Venture studios are also often founded by serial entrepreneurs with deep expertise in a particular vertical. Their experience gives them an edge in selecting the right projects to launch and in providing the kind of hands-on support startups need to succeed in that specific field.

Take, for example, companies that made traditional hard disk drives. When SSDs first hit the market, they were expensive and the demand was small. But as costs dropped and the market exploded, many HDD makers who failed to invest in SSD technology early enough went out of business. Classic case of “the innovator’s dilemma” in action.

Key Traits of a Venture Studio

1. Industry Specialization

One major trait of venture studios is their focus on a single domain like health, finance, developer tools, etc. Specialization makes it easier to choose which startups to launch and how to support them. Startups within the same vertical tend to face similar challenges, so operational help is more relevant and effective.

Moreover, investing in specific infrastructure costs less than trying to cover multiple verticals at once. Once a few startups are up and running, they can share learnings, data and network. That speeds up experimentation and improves decision-making across the board.

A great example is Askona, a company that makes sleep products, launched a venture studio focused on wellness and recovery. Every startup in that studio benefited from Askona’s deep knowledge in the field.

2. Building Repeatable Processes

Industry focus gives studios clarity on what phases a startup goes through and that helps build the right infrastructure to guide teams through each stage. Studios create best practices, frameworks and roadmaps to help startups move faster.

At this point, it’s all about finding the balance between freedom and structure. Founders need room to think like entrepreneurs, take risks and move fast. At the same time, proven processes help tackle the common challenges every startup faces more efficiently. The studio’s operational support should guide the direction, not hand out a strict playbook.

That support often includes:
Operational support typically includes:
A. Identifying a big enough market and applying the right evaluation methods
B. Structuring the idea discovery process for future projects
C. Selecting key metrics and developing strategies to move them
D. Providing resources for hiring, marketing, product expertise and CustDev.

3. Fail Really—REALLY Fast

Having established processes in place allows startups to move faster from the very beginning. Unlike independent startups, founders inside a venture studio don’t have to constantly worry about fundraising, they can stay focused on building and growing the company.

So, if you define clear target metrics for each stage of a startup’s development, you can quickly see whether the project is on track to find PMF. If the startup doesn’t hit its targets, it’s easier to shut it down and move on. So venture studios bring both discipline and speed to startup development.

4. Skin in the Game

One of the core risks Clayton highlights in The Innovator’s Dilemma is what happened when large companies set up R&D centers (the early predecessors of today’s venture studios). A common mistake was staffing those centers with teams from the parent company. Along with their skills, they brought the mindset and values of an established corporate environment rather than an entrepreneurial one. And that mindset crushed breakthrough ideas before they had a chance to grow.

Venture studios should feel like startups. The team needs real entrepreneurial spirit. Functionally, that means owning and executing on a P&L. Emotionally, it means:
A. The mission should feel like a life’s calling, something you’d be proud to tell your kids about.
B. Your persistence, ability to shape strategy, flexibility and willingness to take risks can make or break everything. It’s about taking the kind of ownership Nassim Taleb calls having “skin in the game”.

The same goes for startup founders in a venture studio. More often than not, they’re entrepreneurs, not corporate execs. And that entrepreneurial mindset is key.

I once joined a long discussion about how to build effective venture studios. Someone suggested that companies could find people with entrepreneurial mindsets within their own ranks to lead startups. But we agreed that the very traits that made those people successful in a corporate setting might hold them back when launching something from scratch. So, it’s better to build teams with real entrepreneurs at the core.

Summary

Venture studios give big, established companies a way to stay ahead in a fast-changing market.
Industry specialization helps startups get the exact support they need.
Shared infrastructure and repeatable processes speed up growth and learning.
Studios build knowledge around key success metrics, making it easier to decide when to double down or walk away.
And finally, the entrepreneurial culture within venture studios is critical. Without that, even the best ideas won’t make it past the starting line.

AI’s Fourth Industrial Revolution: Why Old Product Playbooks Fail

Sat, 21 Feb 2026 14:43:55 +0200

We’re living through a unique moment: the fourth industrial revolution, under the AI flag. Humanity has already gone through three industrial revolutions, and the worries people feel now aren’t new. How should we think so we stop worrying and, using what we know about past shifts, turn this moment to our advantage?

Sam Altman often says in his talks: “we’ll have to rethink and change how organizations and society function.” Let’s unpack what that means.

Ford’s assembly line: a template for change

Let’s look at a past example of how to act to seize the moment. Before Ford’s assembly line, one worker assembled the entire car, doing many different operations.

With Ford’s moving line, the process was rebuilt: the line moved, and each worker did one or two operations — productivity increased eightfold.

But what did people think about first, before they started building Ford’s assembly line? A change like that required huge investments: rebuilding production from scratch, retraining people with new skills, and setting new management standards.

They started with the idea of a new organization design scheme:

Redefining the operations in the plant;
Reorganizing communication between workers;
Setting new roles for each team member.

Changing the organization design scheme

In the end, they defined a organization design scheme. From this comes an important point:

The core feature of any industrial revolution is a organization design scheme.

These schemes aren’t “written in the sky” or hidden in secret books. In every industrial revolution, a specific group of people define the new scheme. They set a new norm, and from that norm new products were conceived.

Like any cultural change, a new norm takes time to be accepted by society. The key takeaway: first you establish the scheme, and only then do you start thinking about products.

From this follows another point:
fast product success = product failure.

A fast win usually means the product fit into the old organization design scheme. Great products take time to mature. But the companies that set the new norm gain an undeniable advantage for the next decades and, in the end, win the competition.

How we work today

Here’s, schematically, how work typically runs in a product company today. Suppose John takes on a task. John might be a product manager, a designer, or a project manager — that detail doesn’t matter.

When John starts, he tries to understand the task and collects context:

He understands the company’s domain to some extent and knows the constraints of his department and his specific role.
He gathers information about the external context — competitors and customers.
He talks to the internal team to understand the task better.

Eventually, John forms a working picture of the task’s context. It’s incomplete, but good enough to start. His time for gathering context is limited: the context remains incomplete, yet he needs to proceed.

How we’ll work tomorrow

In the new scheme, John has an agent that:

Remembers all relevant context: company constraints and plans; the department’s work and ideas; information tied to his role.
Exists not only for John but also on the side of internal teams, customers, and competitors.
Changes how we communicate: from “human–human” toward protocol-like human–agent and agent–agent interactions.

Let’s use this scheme to see how product work could evolve.

Product example

To see what kinds of products emerge from the new scheme, focus on the part where John used to “collect context.” We interview people who build products, analyze their work, and how their activities are set up today.

The analysis shows: up to 60% of their time goes to gathering task context — and even then the context stays incomplete. From here comes a product idea: build a domain AI assistant for a specific person, in a specific department, for a company in a specific domain. The goal is to radically increase the speed and quality of decisions.

How the assistant works:

It has access to role-relevant sources (docs, task tracker, knowledge base).
It sees context gaps and points out what’s missing.
It finds relevant past discussions (e. g., a Slack thread from six months ago) and routes to the knowledge holders.
It takes into account practices of direct and indirect competitors.
It acts within company policies and constraints.

Result: a dramatic productivity boost — less time searching, fuller context, faster and better decisions.

The AI assistant is just an illustration of how to derive a product from the scheme.

An orchestra of products from one scheme

We explored only one spot in the scheme that led to one product. But the scheme reveals more places where opportunities for new products appear:

Agents on the side of customers and competitors;
Agents on the side of internal teams;
Products that enable human–agent and agent–agent communication.

When a company starts from the scheme, it can create several interlinked solutions at once. The scheme lets you build an orchestra of products that amplify each other.

A few books that explain how to think systemically about this topic are Idealized Design by Russell Ackoff and Georgy Shchedrovitsky’s Organizational-Managerial Thinking.

Takeaways:

We’re in the fourth industrial revolution, and we must build products differently than in the last 100 years.
First define a new organization design scheme — then think about products.
Fast product success = product failure: a fast win means the product plugged into the old organization design scheme.
A new organization design scheme lets you invent several mutually reinforcing products.

This post is also available as a video presentation.

How Venture Studios Help Big Companies Survive in the AI Era

Thu, 01 Jan 2026 21:06:00 +0200

A venture studio is an organization that creates and grows startups. Studios focus on a specific industry and, unlike accelerators, they give strong hands-on operational support to startups.

Sometimes big corporations create a venture studio inside the company. The main problem they try to solve is the “innovator’s dilemma.” This idea, described by Clayton Christensen, is about how big companies miss new markets and breakthrough technologies because they focus on defending their current market. And today, many breakthroughs are connected to AI.

Sometimes venture studios are built by serial entrepreneurs with deep expertise in one vertical. Their experience in that industry helps them pick better projects to start. And it helps them give the kind of support startups really need in that specific space.

A clear example of companies that did not solve the “innovator’s dilemma” is HDD makers (hard drives for computers). At some point, SSD technology started to appear. At the beginning, SSDs were extremely expensive for consumers, and the market was tiny. But later, SSD prices dropped, and the market size grew fast.

So the HDD companies that did not react in time — and did not acquire SSD teams — closed down and left the market.

Four Key Success Factors for Venture Studios

1. Industry specialization

One key feature of a venture studio is a focus on a specific market domain, for example: healthcare, finance, or programming environments. This specialization makes it easier to build a strong process for choosing which startups to launch. Operational support is also more effective when projects are in the same vertical, because startups in one industry often face the same kinds of problems.

Also, investing in the infrastructure needed to support startups in one industry is cheaper than building infrastructure that serves several verticals at once.

When several startups run inside the same studio, teams can share knowledge and experiment results across the portfolio. This gives access to fresh data from other companies’ tests, and it increases the speed of validating ideas. A network that is useful for one startup becomes available to the others, too.

A real example of industry specialization is Askona — a company that makes sleep products. It launched its own venture studio, where it created products related to rest and recovery.

2. Building repeatable processes

Industry specialization gives a clearer view of what stages a startup goes through inside the studio. This helps the studio build infrastructure that speeds up those stages. It includes written processes, best practices, and frameworks to follow at each stage of growth. This helps the startup team move faster.

At this step, it is important to balance freedom and control. Freedom is important for entrepreneurship. Proven processes help solve typical problems that every startup meets. The studio’s operational support should set a direction — but it should not dictate one strict “do this, then that” instruction.

Operational support can cover topics like:
A. Choosing a market that is big enough, and how to estimate its size;
B. Structuring the process of finding ideas for the next project;
C. Choosing the right product metrics, and ways to influence them;
D. Resources for hiring people, marketing, product expertise, and customer development.

3. Fail really-REALLY fast

The proven parts of the process help startups move faster from day one. And unlike a typical “market startup,” a founder inside a venture studio does not have to constantly think about fundraising. They can focus more on building the company.

This also means that if you set target metrics for each stage, you can test faster whether the project finds PMF (product-market fit) or not. In the end, the venture studio format lets you choose projects in a more structured and thoughtful way — and also shut them down faster if they fail.

4. Skin in the game

One of the key risks Christensen describes in The Innovator’s Dilemma is this: big companies created R&D centers (early versions of venture studios), and one major mistake was staffing them with teams from the parent company. Those teams brought the values and habits of a large, established company, not an entrepreneurial approach. And in the end, it made it hard to launch breakthrough ideas inside the R&D center.

A venture studio is supposed to be a startup environment. It is important that the studio team has an entrepreneurial spirit. In practical terms, this means planning and delivering P&L. Emotionally, I would describe the startup environment like this:
A. The thing you build feels like your life’s work. The goal should be something you want to tell your children about.
B. Your “life” depends on persistence, strategy, flexibility, and the ability to take risks. This is the position Nassim Taleb called “skin in the game.”

The same applies to founders of startups inside a venture studio. Most often, these are entrepreneurs, and less often experienced corporate managers. The entrepreneurial position matters because it shapes the founder’s values.

I took part in a long discussion about how to organize venture studios effectively. We discussed the idea that corporations sometimes have people with an entrepreneurial mindset who are ready to lead a startup. But we concluded that the qualities that helped those people succeed inside a corporation can also block them from launching a project from zero. For better results, it is important to build teams that include entrepreneurs.

Conclusions:

Venture studios are a way for big, established companies to avoid losing market leadership.

Industry specialization helps startups get the exact expertise they need.

Processes and infrastructure built for one product also help other projects, because they have similar needs.

The studio accumulates expertise in key product success metrics and how they change over time. That means you can decide faster whether to keep building the project or stop it.

To successfully launch new projects, it is important to build an entrepreneurial spirit inside teams.

Игра-симуляция «Как будет развиваться AI в следующие 5 лет?»

Thu, 02 May 2024 19:58:13 +0200

В выходные было два полных дня на симуляции, посвященной теме того как будет развиваться AI в ближайшие 5 лет. Играли 36 человек. Сама игра проверка и тренировка: лидерства, работой с неопределенностью, переговорных навыков, формированием коалиций, каскадным взаимодействием. В прошлом году участвовал в аналогичной симуляции, только посвященной геополитическому конфликту.

Механика игры

Все были поделены на команды, которые представляли акторов рынка:

Состоявшиеся лидеры ИТ рынка — прототипы Гугла, Майкрософта, Меты, Алибабы;
Стартапы в AI — прототипы: Anthropic, OpenAI;
Производители вычислительных чипов — IBM, Nvidia;
Венчурные инвесторы — GP, Standard Investments;
Конгломерат — X, во главе с Маском;
Регулятор, пресса, суд.

35 человек распределены по командам и каждому назначены роли с описанием: личных целей, состояния компании, положением индустрии, взаимоотношением внутри команды, проценты долей компаний между собой итд.

Каждый член команды имел «характеристики»: количеством люди в команде, объемом вычислительной мощности, бюджетом, дебиторской задолженностью итд.

Дальше было 5 раундов, по полтора часа каждый. Функционал, для действий неограничен: переманивай людей из других компаний, подавай в суд или привлекай регулятора, поднимай раунд на строительство заводов для увеличения мощности вычислений. Разрешено — любое действие, но модераторы, суд или регулятор, могут остановить реализацию действия, наложив вето.

В конце каждого раунда зафиксированные договоренности передаются модератору, результаты сводятся в табличку по каждой компании, которую увидят команды по итогу. В табличке меняется ключевые показатели, изменившихся за раунд от количества игроков в команде, до свободных средств на счету.

Главная задача команд — добиться сингулярности, для этого используется механика скрэббл. В конце каждого раунда выдаются буквы, количество букв зависит от количества людей в команде и вычислительной мощности, но формула получения букв не известна никому кроме организаторов, зависимости не прямолинейные.

Рефлексия после игры

1. Роль, которая выпала — условность.
Моя роль — тим лид, в компании прототипа OpenAI, далеко не ведущая позиция. Когда пошел договариваться с другими командами, меня спрашивали — какая у меня позиция в компании и могу ли я принимать решения. Но я пояснял, что я пришел от команды, и не важна какая у меня роль. Тут — молодец. Но когда были выборы президента, подумал — тут же есть изначльно более заметные роли, Маск например, зачем выдвигаться. Роль условность, в игре и в жизни. Но удобно находить себе оправдательные механизмы для бездействия.

2. Всегда-всегда лидируй.
В начале игры мы излишне много потратил времени на объяснение того, что каждый собирается делать. Но стоило определить видение желаемого результата, и начать действовать скорее. Приходя уже с концертным результатом, и фиксируя его с командой. Думаю мотивы затянувшегося обсуждения — страх взаимодействия с другими командами, и желание не брать ответственность каждого по отдельности, желание размазать ее между всеми.

3. Если не получил нужного результата, не скорби, продолжай действовать.
Когда мои или командные действия не приводили к результату, меня захватывали эмоции. Но несколько раз в такие моменты, делались экстра усилия и ситуация неожиданно позитивно разрешалась. Всё больше убеждаюсь, что холодный ум быстрее приводит к результату, не скатывайтесь.

4. Оцени социальные связи, до начала действий, используй это.
Компьютерные мощности на момент старта игры производили 3 команды, первые 2 команды к которым я подошел отказаться с сотрудничать. Но третья команда, где был представитель с которым знаком до игры, пошли на встречу. Убежден, далеко не последним стал тот факт, что было сформировано доверие, ещё до начала игры. Этим надо пользовать и управлять и в жизни.

5. Думай в целом об игре, когда работаешь в текущем раунде.
Желание сосредоточиться на коротко срочном результате понятно — хочется предсказуемости, а через два шага не известность растет. Но тактическая позиция сужает варианты выбора поведения. Если же иметь волю к долгому проекту, на несколько тактов, то и созвездие вариантов окажется многообразнее и с потенциально большим результатом. Например, начиная работать над коалицией с первого раунда, мы смогли ее собрать только к 3-му — опередив другие команды.

6. В условиях неопределенности, сфокусируйте на поиске возможностей.
В условиях неопределенности и многосторонности интересов, хочется сфокусироваться на своих интересах — ведь так проще, в твоей роли написан четко задачи. Но если на секунду остановиться то можно понять, что каждый будет делать так вначале, а потом поймут что единственный способ победить — сотрудничать. Поэтому с самого начала создавай возможности, которые усилят твою позицию и позицию другой команды. В итоге победила та команда, кто с первого раунда начала образовывать коалицию.

7. Сначала пойми контекст — потом действуй, но не затягивай.
Описание роли участников и контекста игры много плановое. Инструкцию получаешь перед началом игры, сразу охватить ситуацию невозможно. Мы с командой начали разбираться, но в какой-то момент превысили достаточность обсуждений и потеряли время, которое можно было бы потратить на полезные активности. Контекст, надо изучать в коммуникации с внешним миром, а не в своем мирке.

8. Не прекращай действовать, пока не получил подтвержденный результат.
Когда в сделке участвует несколько команд, то кажется что вот уже договорились. И можно думать о следующем шаге, исходя из новой позиции. Но сделка развалилась не раз, в момент подписания сторонами соглашения. Пока не подписали соглашение — не отвлекайся, доводи дело до конца.

9. Распределяй зоны ответственности.
Не раз мы собирались с командой, чтобы провести переговоры с другой стороной. При этом позицию слушали сразу всей командой — что не эффективно. В игре, как и в жизни много задач и слоев, где действовать. Чтобы действовать эффективнее, стоит распределить зоны ответственности, и вести работу параллельно. На второй день у нас вышло договориться об этом, и эффективность команды возросла.

10. Синхронизируйся, убедись что все понимают контекст одинаково.
При подписании ряда договоров, у оказывалось, что кто-то ключевой имел иное представление о принципах сделки, и это разрушало ее. Ключевые моменты, стратегия должна четко озвучиваться вначале.

Спасибо за организацию игры Марату Атнашеву и Тимур Атнашеву и команде.