I'm rediscovering cloud gaming. The first platform I'm testing right now is Fog Play from MTS.

I picked two games for the test: Death Stranding and 007 First Light. There's not much to say about the first: Kojima is still a genius, and the game is excellent even almost seven years after its release.

But let's spend more time on the second. 007 First Light is a spy action game from the makers of Hitman. Bond is only just joining MI6 here; he still has to earn his famous 007 number.

I've spent three hours with the game. It throws you into the action straight away, before the tutorial. And this is a full 50–60-minute opening sequence, with cinematic transitions that make you feel as if you've stepped into a Bond film.

I especially liked the attention paid to the locations. Even the minor ones are packed with details that add to the atmosphere. The world feels believable.

The tutorial is unusual too. Some things repeat, but not long enough to become tiresome. The scenes are staged so well that you could imagine them in a film about a future agent's training.

In the story, the 00 agent programme is being relaunched. That also makes room for some up-to-date technology, including a quantum supercomputer.

And, as you'd expect from Bond, there's style in the details. At one point you get to choose an OMEGA watch: its colour and strap. That's both customisation and a brand tie-in. Even in Her Majesty's service, you have to decide which strap suits your mood today 😎

It's still too early to judge the cloud gaming service itself, but I've already played for more than eight hours over Wi‑Fi, and I plan to keep going.

A new Winamp is promised for the first half of 2027, with its own subscription music streaming service built on Deezer's catalogue and technology. It will bring streaming, local files, internet radio and podcasts together in one player.

What I'm most curious about is how the old-school crowd will react. In the 2000s, they played pirated music through Winamp; now they'll find a subscription inside their old friend. Will that make our old-school stomachs turn? Or have we simply got used to the idea that music will never be free?

Drop the tracks you played on repeat in Winamp in the Telegram comments! Meanwhile, I'm off to rewatch TPB AFK.

We checked in again on Great on Frame, and wow: it now lists more than 130 games and apps, both VR and conventional ones.

Here's an important clarification from Valve: the compatibility check is for games running directly on Steam Frame, without a gaming PC. They run on the headset itself.

To be fair, the list contains plenty of games you'd expect: SUPERHOT VR, Beat Saber, Arizona Sunshine 2, the remake of the first Arizona Sunshine, Pistol Whip, Moss... Quite the VR all-star team. Among conventional games there's Hades, Portal 2, Balatro and Hollow Knight: Silksong. There is certainly something to play.

The device has already attracted reviews, too. Some praise how smoothly supported VR games run: PC Gamer, for example, liked playing Eleven Table Tennis on the standalone headset at 120 Hz.

But honestly, there was no big surprise. I still don't see a new system-seller for Steam Frame: no high-profile exclusive that makes you want to buy the headset immediately. A standalone port of Alyx is welcome, of course, but the game itself is already familiar.

And that's a shame: a strong selection of old favourites is unlikely to spark a new wave of interest in VR.

My main hope lies with Steam's conventional game library. The headset has 16 GB of shared memory for the CPU and GPU — twice as much as Quest 3. That's useful headroom, although which games run will also depend on performance and ARM compatibility.

If Valve keeps improving all this, perhaps the hardware will comfortably run a sizeable slice of the regular Steam library. In other words, the library could drive sales rather than exclusives.

My conclusion: Steam Frame is a Steam Deck with a giant “screen” perched on your face 😌

The most striking thing at IGROPROM was seeing games about Russia's “special military operation” in Ukraine take off, despite the subject seeming difficult or taboo. Projects like these shouldn't be written off in advance for fear of being blocked or finding no audience. I'm glad there's still room in PC games for sincere creative work on difficult subjects, despite all the censorship and agendas around us.

The festival itself felt welcoming. I liked the spacious, attractive and comfortable stands. Indie teams were treated like people ❤️

The game selection was good too: a lovely showcase.

I barely had time to play anything, buried as I was in work, sessions and the pitch event. Besides the games I've already covered, I'll single out Project Sever. I got to play a little, then quickly crashed my submarine.

Festivals like this let you try games before release and immerse yourself in fan and geek culture. That's a pleasure in itself.

You can also watch cosplay and even a lightsaber duel. Middle-aged blokes dress as Genshin Impact girls, monkeys enjoy a symphony — everyone has their own cultural programme.

For developers, IGROPROM is also a chance to compare notes: where other teams find support, how much it costs them to build Steam wishlists, and how they deal with cross-border payments. In a market this difficult, it helps to know what has worked for others and what you might try on your own project.

So that was IGROPROM... 🐯 And if you'd like to invite a great expert to your event, give me a shout!

Meta* has unveiled Muse Charm, a pocket-sized device with a screen for an avatar of its Muse AI agent. Touch the fingerprint scanner and you can start talking. It's promised for December.

But haven't we seen this before with Rabbit r1? That little box also promised to carry out your requests. At launch, it was slow, buggy, short on features and full of unfinished integrations. In WIRED's review, the writer repeatedly pulled out a phone to finish tasks the assistant couldn't handle. So you had to carry two gadgets and do both their jobs yourself.

Meta now has a new Muse agent aimed squarely at everyday tasks: sorting through email, planning trips, helping with shopping. The company already promises these features in an app; now the agent has a physical home too.

Given Meta's experience with smart glasses, it probably understands better than most what people ask a personal agent to do. Can Muse handle those requests well? Another rhetorical question...

For me, though, the bigger obstacle is the hardware. These days, carrying one more gadget without a compelling reason is a real luxury. Glasses are one thing: besides sitting on your face, they do have other uses. But a trinket with no obvious benefit brings back the question: “Why, Mark, WHY?!” We already have a whole fleet of devices to carry and charge — phone, watch, earbuds. Now they want to add a Tamagotchi, a little creature in your pocket demanding attention 👊

Why not put an agent like this in a wireless-earbud case? Or in the earbuds themselves? At least people already carry and charge those.

This makes it the second device from Meta Connect that, to me, looks like an attempt to improve on someone else's unsuccessful idea. The new VR glasses have Vision Pro behind them; this Tamagotchi has Rabbit.

Meta is designated an extremist organization and banned in Russia.

We recently wrote about retail pilot ideas as more people turn to LLMs to find products. But not every marketplace welcomes that future.

Amazon has blocked access for Muse, Meta's* general-purpose AI agent, which can shop on a user's behalf. Amazon says the agent did not identify itself when visiting the site, while its access to customer accounts posed data-security risks. The company first asked Meta to exclude Amazon from the service, then blocked the agent itself.

The marketplace has previously made life difficult for AI bots and agents from Perplexity, OpenAI, Anthropic and even Google.

I also see a straightforward fight for the customer. When someone else's agent chooses the products, it becomes harder to influence that choice through your own storefront, ads and recommendations. The customer may still be yours, but someone else is doing the talking.

Not every online retailer is so closed off. Walmart and Target, for example, are developing integrations with outside AI services, including ChatGPT. But those are agreed partnerships: an integration does not mean any bot can enter the store.

Perhaps it's time for an AI-agent openness index for online retail: where agents can search products and compare prices, where they can build a cart and check out, and where they get turned away at the door.

Frankly, given the current state of our “sovereign” LLMs, I'm not sure whether opening marketplaces to such agents is a blessing or a curse. Helping someone choose a purchase and placing the order on their behalf carry very different levels of responsibility.

And here's another retail pilot idea: build a general-purpose LLM agent of our own, shopping included. An assistant you can entrust with everyday tasks, as well as finding and buying products across stores.

The first thing to test? Which stores will actually let it in ⛓

Meta is designated an extremist organization and banned in Russia.

Meta* has new VR glasses. First, three facts:

• Around 100 grams on your face. The processor and battery live in a separate unit connected by cable. That's a serious reduction in headset weight, though you still have to carry the whole setup.
• Micro-OLED, with a catch. Sharp images and deep blacks, but a horizontal field of view of roughly 70 degrees versus 110 on Quest 3. The image looks better; the window into VR is noticeably narrower.
• $1,300. Sales are promised for spring 2027.

Industry reviews are already calling them “the Apple Vision Pro we should have got.” But Vision Pro has effectively failed as a mass-market consumer product: sales never really took off. Its bulk was only one reason. To me, the bigger issue was the lack of compelling reasons to use it every day.

Watching a film on a giant virtual screen or spreading work windows around you is fun. But between “fun to try” and “I need this daily” lies a whole shelf of forgotten gadgets.

Which brings us to the obvious question: why make another one?

As an engineering exercise on the way to Project Orion, fair play. Testing compact optics, controls and the lightweight-glasses form factor is useful work. As a consumer device, though, the use cases still feel too narrow for the price.

My prediction: nine chances out of ten it won't become a mass-market device and will eventually fade away like Quest Pro. For now, you can safely ignore it 👀

Meta is designated an extremist organization and banned in Russia.

The main intrigue for me has been removed: Valve has ported Half-Life: Alyx to ARM. Now you can play it directly on Steam Frame, offline, without a gaming computer.

But if you think that now a bunch of developers with games of the same level are rushing to make ports... Then I beg you. There are still much more reasons “why not” than we would like.

Why not No. 1 - Valve has its own Source 2

Your own game, your own engine, your own hardware. You can modify the entire bundle for a specific task. And even in this situation, Valve needed a year of work on the Vulkan and SteamVR drivers. Alyx itself also seriously updated the engine. 16 GB of memory also helped: the game content was saved unchanged. In general, even for the authors themselves this turned out to be a decent amount of work.

Why not No. 2 - eye tracking is not yet available everywhere

It is the combination of eye tracking and foveal, also known as foveal, rendering that Valve calls the key to the success of the port. Wherever you look, there is a high-resolution picture. Resources can be saved at the edge.

But Quest 2, Quest 3 and Quest 3S do not have eye tracking. There is foveal rendering with a fixed area of ​​high quality, but these helmets are not able to move it following the eyes.

And here the question is the size of the market. If, without eye tracking, a heavy port does not achieve the required performance, mass-produced Quests fall out of its potential installation base. And there are so many autonomous helmets with eye tracking that the entire installation base can be included in the credits.

I wouldn’t have counted on a large installation base for Frame alone at the start. It turns out that you need to invest in an expensive port, but a large market in which these investments can be recouped is not yet visible.

Why not No. 3 - the rest have different engines under the hood

Valve - Source 2 is doing a hell of a job optimizing for eSports, where you need high FPS, minimal latency and the ability to play on a wide variety of hardware. During the transition of Dota 2 to Source 2, the company specifically emphasized the improvement in performance even on older laptops. And in the published Dota 2 minimums, the GeForce 8600, a video card from 2007, is still listed. Almost twenty year old iron, for a moment 😱

Most developers use Unity and Unreal - universal engines with tons of capabilities for beautiful 3D games. But a beautiful PC VR game still needs to fit into the capabilities of a mobile chip. And I wouldn’t expect that everything will turn out the same way as Alyx on Valve’s native engine.

To be fair, Valve says that most of the driver and SteamVR improvements will benefit other games as well. This is good news, but I don’t expect a massive flow of graphical games for offline mode just yet.

At the same time, the device itself is very good. I wonder how all this will live in the Valve ecosystem 🧸

Our series about AI gradually taking over the internet has a new episode. This time, it is about online retail.

The Russian SME Corporation shared some intriguing figures. Do we trust them? No. Although the post did not link to the research, it is easy to find online. What really bothered me was the bold headline, “74% of Russians have made a purchase based on an AI recommendation.” The sample of 1,500 respondents included only people who had already used AI-powered search. Still, the results do support a few conclusions.

Among the surveyed AI-search users, 74% had already bought products or ordered services following an AI recommendation. The most common choices were consumer technology and electronics at 69%, medical services at 50%, clothing and footwear at 46%, and groceries at 44%.

For now, the model looks familiar: AI recommends something and then sends the buyer to a marketplace. Some 32% of users always follow links from AI answers to retailers’ websites, while another 59% do so occasionally. Interestingly, 71% are willing to try agentic shopping, where AI not only finds a product but also completes the entire purchase journey. Yet only 12% are ready to let an agent make a purchase without their final confirmation.

I have already tried an experiment like that. I connected an agent to Ozon through Playwright and, for fun, asked it to buy a pack of toilet paper. The agent succeeded, the order arrived the next day, and my 🍑 was saved. Then Ozon blocked my account for a month 😂

On the one hand, you can understand the marketplaces’ position. Bots create load, scrape data, and make it harder to control the customer journey. On the other hand, this creates an interesting paradox: AI is already bringing them customers, while they are shutting AI out more aggressively.

In the longer term, I think AI systems could take over the entire front end of online retail. A person will give a task to a single agent, which will search for offers, compare prices, check reviews, and place the order. Shoppers may simply no longer need the separate storefronts of Ozon, Wildberries, and Yandex Market.

Marketplaces would then become infrastructure: fulfilment, payments, and so on. In other words, they would become the “Russian Post” of the future — technology-driven, fast, and invisible to the end customer. Those that survive, that is 🔪

For Ozon shareholders, this raises an uncomfortable question: can the company remain a growth business if it eventually loses its direct relationship with the customer? This is not investment advice — just a bit of foresight to think about.

This is where the futurism ends and the practical work begins. Traffic from LLMs is still limited, but that is clearly where things are heading. Online retailers should not throw every gate open to agents. They should test selected categories and products, ideally alongside GEO — generative engine optimization for brands and content appearing in AI-generated answers.

I know several strong teams working on GEO. Together, we can build a pilot, test the hypotheses, and work out how to prepare a business for a world in which the customer talks not to your storefront, but to their own AI agent.

And if you have managed to buy something through an agent too, tell me immediately. I am ready to ask for your autograph 😍

Scientists have mapped the Drosophila nervous system - more than 166 thousand neurons and about 125 million connections. And enthusiasts have found a use for it: teaching a brain model to play Bitsaber.

For now, however, she repeats the movements from a specific replay. He is still learning to react to blocks independently.

Imagine the existential horror of finding out you're just a model fly in a simulation playing bitsaber.

In general, there is fundamental science behind this: researchers are figuring out how the nervous system is connected to behavior. But personally, my professional deformation immediately turns on: where can this be applied?

Could living insects ever be “programmed” to fight pests? Pure fantasy for now. Or learn from a fly how to process visual signals and test this principle on a small robot? Already a hypothesis for the pilot.

By the way, if you have invested a lot of effort and money into development, and it is not yet clear who needs the resulting technology and why, write. We’ll help you find application scenarios, formulate hypotheses for pilots, and figure out what to check first. And with tasks outside our profile, we will refer you to specialized Skolkovo experts :)

Attended VK Play Day - a meeting of developers with the VK Play and VK Tech teams. I recorded for you a fragment of a speech by the head of VK Play, Alexander Mikheev, in which he talks about how games will go beyond the boundaries of the site itself.

Separately, the endless stream of neuro-games got its share. And it's not just about the loss of user trust. All games, including small demos, must be moderated. When prototypes assembled in five prompts arrive on the site, people have to check them - and there are no more moderators and testers. The result is a DDoS neuroslope, which causes normal developers to wait longer for their turn.

But here’s my thought: perhaps the initial screening should be given to the community? Make a sandbox where players will try these demos for bonuses or just out of curiosity and clear the neuroslop. Something alive has surfaced - we are sending it for a full check. Let the rest be composted there. The main thing is that this sandbox does not turn out to be a bottomless cat litter box.

For me, the most interesting thing here is how the developer’s task is changing. Already at the idea stage, it’s worth understanding where a person will encounter your game, in what situation he’ll start it, and why he’ll stay. The TV in the company of friends and the Zen feed have completely different scenarios.

But the most interesting thing was, of course, in the networking area. There they discussed the most disliked issue for many creators - return on investment and multiple growth. Let me explain: return is when you earned more money than you took from the investor and the state. For example: what game, niche or even business model (hello vibecoding studios) could grow to at least 1 billion rubles in annual revenue? Where is the next growth point and where should the money be spent? So, if you are asking the same questions, you are open to dialogue 👀

Well, almost…

GoPro has signed a merger agreement with Starman Optical and announced plans to expand into equipment for AI data centres. So let’s send this legend into a well-earned retirement with a farewell post.

We may be witnessing the end of a remarkable story that began more than two decades ago. In 2002, an enthusiast named Nick Woodman (not to be confused with another famous Woodman) identified a pain point shared by tens of thousands of people and created an entirely new category of device. GoPro went on to become a generic term for action cameras.

There is no shortage of detailed post-mortems explaining how GoPro lost its footing. They have been appearing since 2016, when the company suffered its first major failure: its ill-fated attempt to take on DJI in drones.

My own take is simple: this was not primarily about tactical missteps; it was a strategic failure. While GoPro was trying to catch up with Chinese competitors in consumer drones, Chinese companies were making inroads into the action-camera market through genuine innovation — Insta360 being the obvious example.

A failed new-product strategy, an attempt to coast on past success and a chase after an arrow already loosed from an Asian bow led GoPro to this predictable outcome.

Simply chasing trends that have already taken off or copying other companies’ products is always a bad idea, even behind trade barriers and with government protection. Unfortunately, I see this regularly in the Russian market: there are plenty of strategic mistakes built around the logic of “import substitution,” even when it is dressed up as “leapfrogging.” Get in touch if you need a strategy and want your company to avoid GoPro’s rather prosaic ending ☺️

Meta* has launched Pocket in the US, an app with an endless feed of interactive content that users create from prompts.

At its core, Pocket is TikTok with mini-games instead of short videos. One notable feature is “remixes”: with the creator’s permission, users can clone a game they like from the feed, add to its prompt and publish their own version.

At first glance, this may look like a pocket-sized game development tool. But Pocket is less a new game builder than an emotion builder — and a major product experiment for Meta*. Social networks used to ask us to watch, read and listen. Now Zuckerberg is testing whether we are ready to scroll, play and immediately remix other people’s creations.

My guess is that millennials will give yet another entertainment app from Meta a pass. But the bet on a younger audience raised on TikTok, Roblox and neural networks could pay off. If the experiment succeeds, Pocket may not remain a separate sandbox for long. I wouldn’t be surprised if mini-games and interactive posts eventually appeared in the main Facebook* and Instagram* feeds, although the company currently denies that this is the plan.

So we should get ready for an era in which endless games hook us alongside endless videos: created in a minute, played for ten seconds, then forgotten — or remixed.

Meta is designated an extremist organization and banned in Russia. Facebook and Instagram, which it owns, are also banned.

If your studio has been releasing games for several years and earns steady revenue, support measures may sound like something intended for startups, grant applicants, and accelerator participants. Yet for an established business, tax incentives often deliver the greatest benefit: with substantial revenue and a large team, the savings can be significant.

There is no universal solution. Does a particular incentive fit your revenue mix? Can it be combined with Skolkovo participant status? Does the region in which the company is registered matter? As soon as you begin to investigate, choosing the right support measures becomes a quest of its own.

There are many support mechanisms, but selecting the right ones and understanding every requirement without assistance is difficult. The conditions also change: some incentives available a year ago may be restricted in 2026, and certain programs or tax regimes cannot be combined.

The scale of this system is clear from a new study prepared with colleagues from APRIORI, the Internet Development Institute (IRI), and EBR Law Firm. It covers almost every current support measure for the game industry—from tax incentives and grants to acceleration, export, and infrastructure programs.

If you want to understand which measures suit your project or company, leave a comment or send me a DM. We can examine your situation and help identify the right route, from tax incentives and Skolkovo participant status to grants and other support instruments 📈

WATT Studio has released an orchestral trailer for the action-adventure game TSAREVNA. Age of Tales. At its centre is “The Tsarevna’s Theme” by composer Alexander Billion, performed by the Opensound orchestra under conductor Stanislav Malyshev. The piece combines folk motifs, a choir, and classical symphonic sound with a modern experimental arrangement.

It is wonderful to see the creators investing so wholeheartedly in the game’s promotion. Just listen to this soundtrack performed by a live orchestra. It gives me goosebumps, and I want to believe that all the music in the game will be equally impressive.

As a reminder, TSAREVNA. Age of Tales is a dark Slavic fairy tale in which ballet movements form the foundation of the combat system. Driven by revenge, the Tsarevna travels through worlds inspired by Slavic fairy tales and mythology—from the heavenly realm of Iriy to the depths of the Underworld. Ancient deities, creatures of Darkness, and numerous airborne ballet combos await her along the way.

The game is scheduled for release in 2027. In the meantime, add TSAREVNA. Age of Tales to your Steam wishlist ✔️

The interview explores how technologies developed in the games industry are being applied to film, animation, architecture, education, and manufacturing. It also examines the development of the creative economy, cross-industry technology transfer, support for small studios, and digital products built at the intersection of content, software, and new business models.

Core expertise: digital creative industries, gaming industry, game development, VR/XR, game engines, digital simulations, gamification, virtual production, technology ecosystems, creative economy.

Read the original “Creative Capital” interview on the VEB.RF Urban Laboratories website.

Note: the page may need to be opened in Yandex Browser because the certificates used by the site are supported in Russia.

According to Skarred Ghost, Lenovo is closing its US XR division and changing direction. Instead of enterprise VR/AR solutions, the company plans to focus on consumer AI wearables under the Motorola brand.

It is ironic because Lenovo produced the Rift S, perhaps the last mass-market wired Oculus headset. ThinkReality VRX, Legion Glasses, and other devices followed, but there was no longer a sense that Lenovo intended to compete seriously for the XR market.

The news is not a sensation in itself. But it is far from the first signal: VR winter is coming.

The clearest harbingers of a possible apocalypse, in my view, were the mass closure of VR game studios, cancelled games, and Meta’s* circus act as it abruptly reversed course and took an axe to its VR division. Those were the first three horsemen; since then, a whole herd of omens has appeared. One trend in particular has worried me: former VR exclusives receiving “flat-screen” PC versions. First came the news that Moss would launch on conventional PCs. Then we heard about a port of Arizona Sunshine—a game that, for many people, symbolized VR’s heyday.

Now Lenovo is effectively admitting that AI wearables are a more attractive investment than XR. This is not yet the “death of VR,” but the feeling of late autumn across the industry is growing stronger 🍂

Meta is designated an extremist organization in Russia.

Palmer Luckey is teasing us again. The founder of Oculus and Anduril appeared wearing an unusual pair of glasses and immediately declared that these were not merely “smart” glasses, but full AR with a wide field of view and colour stereoscopic imagery.

Much remains unclear. Are they standalone? Do they use an external compute pack? Is this a finished device, a prototype, or perhaps a clip-on attachment for ordinary aviator frames?

Still, the device looks genuinely interesting. After the bulky EagleEye system for the US military, such a compact form factor almost feels like a challenge to current AR trends. And honestly, I find the idea of an AR attachment for conventional glasses very appealing.

The key question is simple: does it make you think, “That looks good—I’d buy it”? If so, Anduril’s prospective military customers may well react the same way. Perhaps this familiar, compact form factor is exactly what the military needs, without making users feel as though they are wearing a helmet from Halo 🤩

The company is the first to name a specific date for phasing out physical releases. Sony fanboys are, to put it mildly, unhappy about the decision. But why did the Japanese company take this step?

First, it makes economic sense: there is no need for manufacturing, logistics, or warehouses; no retail margin to share; the secondary market disappears; and the platform controls prices and discounts. Second, the decline of disc releases is a well-established global trend. Disc sales have been falling for years while digital versions dominate. About four out of five PlayStation games are already sold digitally. Microsoft may also launch Disc-to-Digital this week, allowing users to convert disc-based games into digital copies. But physical media is not giving up yet: it remains important to Nintendo, collectors, and... Russian players.

For Russian gamers, the consequences of this shift are much more serious—a painful blow below the belt, you might say. Of course, our Fabergé eggs are made of steel and we can endure a lot, but the fact remains: the PlayStation Store has been unavailable in Russia since 2022, while discs remain the simplest way to buy a new game in 2026 without a foreign account or intermediaries. After 2028, buying new PlayStation games will become a headache by default. It will be painful, expensive, and... perhaps not permanent.

The biggest risk in the transition to digital is the loss of control. Buy a God of War disc and it is yours: give it away, trade it, or sell it. Buy a digital copy and you become hostage to a platform... A FOREIGN one!

Jokes aside, do you not think the conditions for a successful Russian console launch are becoming increasingly favorable?

On June 26, KRI, Russia’s game developers conference, will take place in Moscow. I invite everyone to a panel discussion that I will moderate. The title says it all. This session is for those who want to understand not an abstract “future of Russian game development,” but the real state of the market after its recent restructuring 📊

Even if you are not planning to attend our session, I recommend exploring KRI itself. It is not a festival built around cosplay and merchandise, but a major industry event for everyone involved in games: developers, publishers, investors, educational projects, service providers, media, and content creators. The discussion goes beyond games themselves to everything behind their creation: business models, hiring, investment, infrastructure, technology, and careers. Simply put, if you want to understand how the game industry works from the inside today, KRI is well worth a visit.

A little more about the panel. In one hour, we will bring together several perspectives: city support and the creative economy, platform infrastructure, studio experience, independent development, investment, and technology trends. This will not be a conversation focused only on problems or only on optimism. We need an honest discussion of where the industry stands, where new growth opportunities have emerged, and where systemic barriers remain.

Want to know which genres are gaining momentum, and which teams and business models have a chance to grow over the next three years? We will discuss all of this—come along!

📍 Venue: Moscow Video Games and Animation Cluster, main hall
Address: 40 Bolshoy Boulevard, Skolkovo
⏰ Time: 11:00–12:00

Bring the whole family, as they say—but do not forget to register 🤩

Last week, I had the opportunity to speak in Khanty-Mansiysk at the 8th International Conference “Information and Communication in the Digital Age,” held as part of the IT Forum and UNESCO’s Information for All Programme.

My thanks to the Russian Committee for UNESCO’s Information for All Programme for the invitation. It was encouraging to see video games discussed not only inside the industry, but on a major international stage alongside AI, digital culture, and the future of the creative industries.

I took part in a session on games, AR/VR, and the digital economy—and once again found myself thinking about how often games are underestimated.

🎮 First, games are quite literally at the frontier of technology. Curiously, today’s AI boom rests to a considerable extent on game development. The NVIDIA GPUs now used to train LLMs and generative models grew out of a decades-long race for better graphics, higher frame rates, and more advanced rendering. In that sense, modern AI runs on the legacy of the game industry.

🎮 Second, games are a powerful instrument of soft power and cultural promotion. One slightly provocative idea emerged during the session: if we want Russian culture to appear more often in games, we need to make it easier for developers to use. Like everyone else, developers often choose the path of least resistance. If high-quality assets, references, architecture, ornament, and cultural artefacts from Russia are readily available, they are much more likely to appear in international projects.

🎮 Another important point: games have long been about more than entertainment.

Game mechanics thrive far beyond Steam and consoles: in corporate training, HR, education, simulations, and the gamification of user experience. For game teams, working with corporations is sometimes not a compromise but a perfectly viable route to a sustainable business.

We also discussed a more controversial idea: using games to show young people the consequences of their decisions. I approach this cautiously, but there is some truth in it. Interactivity can do something lectures and presentations cannot: let people experience the consequences of their own actions instead of merely hearing about them 😩

Once again, I came away convinced that when we talk about games, we are not talking only about entertainment. We are talking about technology, culture, education, economics, and the way digital society will be structured in the years ahead. So let’s keep diving.

And if you are currently considering or developing a video game project from a corporate perspective, let me know 🤫

Pre-orders for the new Steam Machine have opened in Russia. The price, as usual, is eye-watering—almost RUB 200,000. To be honest, though, official Russian retailers are unlikely to see much demand. It will probably play out as it did with the Steam Deck: most of the audience will buy through grey-market imports, marketplaces, or friends abroad. Official pre-orders are mainly for the most impatient geeks.

Specifications:

• Processor: AMD Zen 4, up to 4,800 MHz
• Graphics: discrete AMD RDNA 3 chip with 8 GB of GDDR6 VRAM
• Memory: 16 GB DDR5, with the number of sticks being anyone’s guess
• OS: SteamOS

💛 Verdict: a sad mid-range gaming rig

The Steam Machine concept is interesting in itself: Valve is again trying to bring Steam into the living room and offer an alternative to traditional consoles. But there are still more questions than answers. One less obvious issue is VR support. Judging by the specifications, the VR experience may be roughly comparable to PlayStation VR2 on PlayStation 5: adequate at a basic level, but with little performance headroom for demanding PCVR projects. For the hardcore VR community, this is a questionable device.

And here is my main question: why the hell do we need a Steam Machine at all? 🤔

Seriously. Instead of another box under the television, I want something entirely different. Valve, please give us a proper Steam Frame—a full-fledged VR/MR headset that has been rumoured for years. After the Steam Deck’s success, it would make more sense to move toward a market that genuinely wants a strong new player rather than yet another console form factor.

For now, it feels as though everyone is waiting for one product, and for some reason Valve keeps showing up with something else 😄

Private investment in the game industry is extremely difficult to secure today—especially on reasonable terms.

The reason is straightforward. When the central bank’s key rate is high, money becomes expensive, and an investor has to compare every risky project with the return that can be earned easily from a bank deposit.

The situation is even more difficult for game projects without revenue. In my experience, private capital at this stage is now provided mainly through a startup- or venture-studio model: the investor does not merely finance the team, but takes the studio under its wing and becomes deeply involved in product development, business formation, and oversight.

Government support is not entirely straightforward either. A significant share of the available programs require private co-financing, and finding it for a project without revenue brings us back to the previous paragraph.

For operating companies, however, the situation is noticeably better.

💰 A group of investors approached me because they are looking for game companies with an established business and at least a modest profit. They are considering the purchase of a roughly 10% stake at a reasonable 3–4× valuation multiple. The target transaction size is about $1 million, adjusted where appropriate to the size of the business.

Jurisdiction is not critical, provided it is clear and acceptable for the transaction.

The investors have worked in the market for years. They do not intend to seize control, install their own people on boards, or tell developers how to make games. Their objective is simpler: invest in an operating business, benefit from its growth, and—where there are synergies with other assets—help the company scale faster.

In such cases, my role on behalf of Skolkovo is to keep the process constructive and transparent: align expectations in advance and reduce the risk of unpleasant surprises. In this particular case, I am confident enough in the investors and the absence of hidden pitfalls to publish this request openly.

So, I am looking for game companies that need capital to scale an existing business. If that describes you, or you know a suitable team, send me a DM ✍️

One more reminder: if you already hold Skolkovo participant status or plan to apply for it, there is a business angel investment reimbursement program designed to attract private capital. I discussed it in a previous post.

Remember when half the internet was admiring a gopnik flying on a magic carpet among Russian apartment blocks? Not even a year has passed since we discussed the capabilities of Genie 3. And now a new hero has entered the scene.

Enthusiasts are experimenting with Anthropic’s Claude Fable model. A working copy of Minecraft was recently generated with version 4.8 from one f*#@$&!ing text prompt. The newest flagship, Claude Fable 5, has raised the bar even higher by making a copy of The Elder Scrolls V: Skyrim. Once again, it took a single prompt: “make Skyrim.”

This case confirms that an advanced user is becoming almost omnipotent. At the moment, the main constraint is the safety restrictions built into public language models. It is frightening to imagine what holders of US government security credentials might do with Claude Mythos 5—the intelligence-services counterpart to Fable 5 😮

Where is the world heading? Toward a future in which everything that can be automated with genuine benefit will be automated. A business hypothesis with only a moderate potential impact used to be unworthy of testing because validation was expensive. Now prototyping costs are falling dramatically, and even modest ideas will make it to MVP. Incidentally, an MVP no longer necessarily requires a programmer. With an LLM at the level of Codex 5.5, and especially Claude Fable 5, you can approach a task like making a wish to a genie: solve a problem in almost any field, including making yourself rich 🍸 Sounds like magic? Remember Arthur C. Clarke’s Third Law:

Any sufficiently advanced technology is indistinguishable from magic.

It sounds like cyberpunk, but surveillance cameras have long been a routine part of city life. In Moscow today, for example, it is difficult to find a place beyond Big Brother’s watchful gaze. Now the camera and access to databases are moving directly onto an officer’s face.

China plans to use smart glasses for street patrols, traffic management, and checking people in crowds. The work is done not through a monitor in a control room or a remote data centre, but on the spot: the officer looks at someone and the system immediately displays a prompt. It amounts to a live identity and background check. Interestingly, both the software and the devices themselves are developed entirely by Chinese companies.

Similar scenarios have already appeared in Russia. Astra Consulting, for example, demonstrated AR glasses for Rostransnadzor inspectors with access to Ministry of Internal Affairs and traffic police databases. Skolkovo startup Digital Inspector is working in the same direction, developing solutions for video analytics, monitoring, violation recording, and real-time prompts for security personnel.

This is not yet an augmented-reality revolution. On most of these devices, the AR interface is modest: a short text prompt, status, number, or simple marker in the user’s field of view. Sometimes information is displayed to only one eye 👁

The important conclusion is different.

The scenario of “looking someone up in real time” is becoming increasingly common. It used to rely on infrastructure: large systems operating somewhere in the background.

Now that logic is becoming closer, more local, and more personal. We are gradually moving from city-wide surveillance to devices carried by individual officers—from an abstract “Comrade Major” in a server room to an interface directly in front of the eyes of the person standing opposite you 🚨

Welcome to the brave new world!

We predicted Roblox’s return six months ago. However, press releases and public discussion still contain no information about moving its infrastructure to Russia.

Fabula Rasa: Dead Man Talking is a VR experience in which you are locked in a cage in a medieval square and prepared for execution. The king is about to pass judgment, and you must persuade a witch, bard, executioner, and other characters to speak in your defense—rather like that childhood coin game millennials may remember 🫤. Instead of choosing one of three prepared lines, you have to argue your case aloud, talk your way out, improvise, lie convincingly, or appeal to sympathy. The project does not look like a basement demo made by a lone enthusiast: it has already won the Audience Award at the SXSW 2026 XR Experience.

The key is that its NPCs do not simply recite pre-written lines. They listen, remember who you claim to be, and react to your profession, accusation, and speaking style. Heise reports that conversations developed differently across two 30-minute playthroughs, and the experience felt more like improvisational theatre than a conventional game 🎭

What interests me here is not that “AI can now write dialogue”—that is the least interesting part. It is that a game can suddenly test your real ability to communicate instead of putting you through the usual stress test of reaction time, aim, or character grinding. Persuade them and you live. Fail and you are fed to the monster. True to life, in a way.

This is an important shift for the industry. A writer used to build a dialogue tree. Now they design the framework: the world, roles, conflict, behavioral boundaries, and dramatic structure of the situation. The LLM under the hood becomes an engine for live communication and, to some extent, a co-author of the product ✏️

If we set token costs aside, the business potential is compelling too. These mechanics could be useful not only in games, but in negotiation and sales training, soft-skills development, legal simulations, corporate training, and VR onboarding. I can imagine a future without dull “choose the correct answer” tests: instead, a client is angry, an official applies pressure, an investor has doubts, and the employee must talk their way through the scenario.

I am honestly surprised by how few games have an LLM under the hood. We have already written about different VR experiences, and the next great leap in VR realism may not come from graphics. It may come when virtual characters finally stop being furniture and new interaction mechanics emerge.

If you are building a game, simulator, or corporate training product—or simply wondering how to add AI to an interactive experience without creating AI slop—this is exactly the kind of case I can help you analyze. Get in touch for a consultation ✨

Every time I see a headline about layoffs at Zuckerberg’s company, I instinctively expect the long-suffering VR/metaverse division to have been put out in the cold again. But who cares about my childhood trauma and phantom pains? This time the cuts are spread across the board: 8,000 people will be laid off, 7,000 reassigned to serve artificial intelligence, and 6,000 positions left unfilled.

“As previously announced, we have decided to reduce headcount... to offset other investments we are making,” says a memo leaked to the media. Bravo, Meta. The company’s brave new bet on AI is becoming clearer by the day. The remaining 20% of staff will somehow muddle through 💁‍♂️

I started wondering: what share of Meta’s workforce will now work in AI? According to another memo leaked earlier, the company’s new Applied AI and Engineering division employs 2,000 people. Add the 7,000 being transferred, and the total comes to 9,000.

But is Meta on the right path? Progress in LLM development clearly depends not only on the number of specialists involved, but also on the effectiveness of their intellectual work. Computing power matters too, as does a company’s willingness to spend enormous amounts of time and money without any guarantee of success.

I believe AI adoption is moving from “let’s just try it” to “which role—or part of a role—can we replace right now?”

What programmers saw first—help with routine work, code, and rapid prototypes—is now spreading into marketing, analytics, sales, support, and content production. A production-ready prototype can now be assembled not only by a developer, but by an advanced user.

If your company wants to adopt AI, I would approach it like this:

• ✅ Map the roles and processes inside the business.
• ✅ Identify where AI genuinely reduces manual work, speeds up production, or improves outcomes.
• ✅ Find scenarios where you can build a virtual employee rather than just a “pipeline with a neural network.”
• ✅ Test it quickly in a pilot and calculate the impact.
• ✅ Scale only where the business case works—not where there is merely an attractive AI demo.

I write this so confidently that it sounds as though I am about to sell you an “AI in three days” course and access to a private club. But the point is different: I am watching this happen in real businesses right now. Often on a small scale, yes. But once it gains momentum in your field or role, catching up will be much harder.

When a routine role seems impossible to replace, the problem is often not its uniqueness but the fact that the role itself is poorly defined. For now.

This is exactly where I can help: map the business, identify roles and processes suited to automation, build a realistic AI pilot, and separate working scenarios from fashionable imitations of progress. Send me a DM and let’s discuss it 😎

Meta is designated an extremist organization in Russia.

Congratulations to the team of brave young developers on their long-awaited release! 😏

Why is the game worth your attention? First, because homegrown developers deserve our support 😉 The team has spent four years on the project, starting while they were still at school. These are not merely their first steps in game development—the game is their life and a dream made real. That alone makes me proud of this new generation of Russian developers. But beyond sentiment, the game has plenty of substance.

The Ballad of a True Thug is an atmospheric top-down shooter set in crime-ridden 1990s Europe. It is a dark story of revenge and friendship in which you rise from petty thief to criminal legend. The plot has an unexpected ending, and friends and enemies are not always what they seem. In mood and action, it sits somewhere between Hotline Miami and Katana Zero.

Key features:

• ✅ Combo system + slow motion: the faster and more varied your kills, the more you slow down time.
• ✅ Risk-reward mechanics: buffs that change your play style.
• ✅ Freedom of approach: complete levels aggressively, stealthily, or with a mix of both.
• ✅ Arsenal: grenades, slides, and plenty of guns.

Six months ago, we interviewed the game’s creators and are still proud of the result. Be sure to read it if you missed it. If you would like to know what has changed for the team and what they are planning next, tell us in the comments. We will talk to them again and share their story in detail ✨

• ➡️ The Ballad of a True Thug on Steam
• ➡️ Demo on itch.io

One of the sessions I moderated at CIPR focused on AdTech and autonomous demand. Here are several figures that are exciting—and may be shocking to some.

The age of SEO is not over, but a new era is arriving: GEO, or Generative Engine Optimization. Like it or not, AI has come to transform marketing too, and generative optimization is growing at extraordinary rates:

• Traffic from GEO has grown 3.3× since autumn 2025
• Revenue from GEO traffic has grown 6×
• Its conversion rate is 2.5× higher than organic search

GEO is currently the fastest-growing marketing channel.

Consumer demand is also changing: it is becoming autonomous.

• 35% of buyers use AI to choose products (RBC + Yandex, December 2025)
• 50% of internet users regularly use AI assistants
• 10% of requests to Alice are direct product searches

Across industries, AI is involved in roughly 10% of decisions on average. Depending on the category, the figure ranges from 6.4% to 14.4%. In absolute terms, traffic from LLM tools is still modest, but its growth rate suggests that once this flywheel gets going, it will be off to the moon 🚀

This new reality presents the industry with serious challenges.

Models need user data to give accurate recommendations. We surrender some of it ourselves by accepting terms and conditions without reading them. Other information—such as a person’s purchasing history or content-consumption profile—is available only to retailers, telecom operators, banks, and similar companies, which are not eager to share it. Solving this problem may require a third player in the market: a data aggregator.

But there is another question to which I do not yet have an answer. Perhaps we can think it through together 🤔 If LLMs can weigh data objectively, how important is brand equity? Do we still need brands in the traditional sense, or should a brand become clear, machine-readable information backed by arguments? It is as if consumers were no longer looking at labels, but choosing products by their attributes—as they already do with marketplace filters 🛍

The conclusion seems obvious: whoever controls consumer attention controls the cash and the margins of everyone further down the value chain.

How do you search for products? Do you still Google them, or do you go straight to ChatGPT or Gemini? ❔

P.S. The polished slides with all the figures are below.

Friends, in recent weeks all my attention has been focused on the Digitalization of Industrial Russia conference. Every spare minute—and quite a few that were not spare—has gone into preparing for this important cross-industry event. Yes, I will be there. And I will not just attend: I will moderate two excellent sessions. Clear some space in your calendars 😏

✅ “AdTech. Autonomous demand”

We recently published a post arguing that AI agents may soon replace the apps on our phones. That future is already knocking at the door. When it enters, the rules of the advertising market will change dramatically. The winners will be those who prepare in advance for what I will boldly call a revolution. That is the subject of my first session.

📍 Hall 12
🕒 May 19, 10:00–11:00

✅ “Components for the video game industry”

Have you heard about RAMmageddon? Perhaps you have even read our first and second posts about it ✨ When computer hardware is in short supply, users may appear to be the main victims, but every part of the video game industry suffers—including developers. They have to adapt to new constraints, perhaps by reducing visual fidelity for optimization. Otherwise, only three members of the PC master race will be able to run their hardware-hungry title; the game will not pay for itself, and the studio will close. What can be done? Options do exist. I invite you to the second session, which I will moderate, to learn more :)

📍 Hall 3
🕒 May 20, 13:45–14:45

Technical details, directions, and the full conference program are available on the CIPR website.

See you in Nizhny Novgorod! 🌟

Industrial VR has long had an awkward problem. Everyone imagined an engineer putting on a headset, entering a digital factory, inspecting a machine, finding a fault, and fixing it—saving a great deal of money. It sounds wonderful. But first, someone had to build that factory in 3D by hand: walls, machines, stairs, passages, utilities, and everything else. That took weeks, sometimes months, and a budget large enough to send any sensible production manager back to Excel, an A3 printout, and the phrase “let’s just talk it through for now.”

Then Gaussian Splatting crashes into this familiar pain point 💣

At Hannover Messe, a team demonstrated a workflow in which a real location can be captured in about ten minutes instead of being modelled manually. The data is then processed automatically in three to four hours to produce a photorealistic XR environment. It is not a perfect engineering model, not millimetre-accurate CAD, and not the burned-out shell that classic lidar scans could resemble. It is a space you can enter, explore, show to a customer, use for training, or overlay with guidance.

And that may matter more than yet another headset with a better display.

The main bottleneck for industrial XR was not just hardware; it was content. Building an environment was expensive, slow, and tedious. When that process changes from “a separate three-month project” to “captured in the morning, reviewed in the afternoon,” the economics change. The workflow becomes almost routine.

You start with a real facility. Capture it quickly in 3D. Produce a photorealistic XR scene. Add the necessary information. And voilà 💁‍♂️ Industrial XR stops being a “look, we have a metaverse too” attraction and becomes a practical working environment. Engineers from around the world can meet inside the same space and discuss not an abstract drawing, but a specific factory floor—right down to the pipe someone will hit their head on. Employees can train inside a recognizable facility instead of staring at a PDF with a dreary arrow saying “turn the valve.”

There is, however, a fly in the ointment: these models are still mostly static. You can show the factory floor, inspect a machine, and discuss a production area, but you cannot yet show workflows in motion. That is a whole new level of pain.

This is where Gracia AI comes to mind.

Dynamic, video-based versions of Gaussian Splatting already exist. They are not yet mainstream industrial tools, but the direction is clear: scan a space quickly, then add live scenarios. Gaussian Splats do not have to replace other technologies, either. They can be combined with 360° video, conventional 3D objects, CAD/BIM models, interactive labels, avatars, voice assistants, and enterprise databases.

The result is not “a new religion replacing old-school engineering,” but a useful hybrid. CAD and BIM provide precision. Gaussian Splatting provides realistic visual context. XR provides shared presence and a sense of scale. Now this starts to look like a mature market—not a fantasy about shop-floor workers suddenly spending every day in VR headsets, but a real opportunity to reduce the costliest part of the XR pipeline: creating the environment.

Industrial VR used to be sold with “wow, you can walk around a virtual factory.” Now there is a more rigorous business case: coordinate faster, train at lower cost, sell more clearly, and prevent mistakes before they are literally set in concrete and signed off.

When XR content can be assembled in a single working day rather than several months, it stops being a toy for an innovation showcase. It becomes a tool whose economics finally add up 🔟

DeepMind alumnus David Silver has raised $1.1 billion to launch a new AI startup, Ineffable Intelligence. That impressive sum will go toward an enormous goal: creating artificial intelligence capable of learning without human-generated data.

For context: DeepMind is an AI research laboratory founded in 2010 by Demis Hassabis, Shane Legg, and Mustafa Suleyman. It became known for developing AI systems that learn independently through reinforcement learning—learning by interacting with their environment. In 2014, the Evil Corporation bought it for roughly half a billion dollars.

DeepMind’s best-known projects include:

• AlphaGo—the first AI system to defeat a world champion at Go.
• Gemini—Google’s current multimodal AI model.

Headlines like this can make AI seem like rich black soil: stick in a chair leg and an evergreen tree covered in golden apples will grow. That impression is deceptive. You can slap an “AI” label on any toilet, but that will not turn it into a work of art or guarantee successful fundraising. You need a genuinely deep-tech solution with a serious engineering and scientific foundation. And one other small detail: access to the computing infrastructure needed to run it.

It is worth remembering that AI capable of training itself indefinitely is one of the ingredients of the intelligence explosion associated with the technological singularity. Imagine switching on a server with such an AI and letting it improve until it becomes smarter than its creators. What did Jewish folklore warn us about with the Golem? 🤔

With that context in mind, it is even more interesting to watch an eight-year-old film about DeepMind’s early days.

Documentary AlphaGo:

• 💛 Russian
• 💛 English

And the follow-up documentary, The Thinking Game:

• 💛 Russian
• 💛 English

Bloomberg reported a tragic incident that occurred in July 2024 during the production of an immersive video for Apple Vision Pro. British Paralympian and pilot Claire Lomas died when her ultralight aircraft crashed over the Jordanian desert. A 3D camera was mounted on the aircraft, and the footage was being shot for Apple’s Adventure series—the videos the company uses to demonstrate why Vision Pro should exist at all.

What is striking about this story is not only the tragedy itself—remember Woland’s observation about death—but the logic of the genre.

VR sells safe risk. You put on a headset at home and a moment later you are soaring over the desert in a glider, hanging by one hand over a precipice, jumping into the water from a 27-metre cliff, or watching a volcanic eruption from the front row. You do not spend a fortune or years of your life on training, travel, logistics, insurance, and everything else. Most importantly, you risk nothing except the cleanliness of the space around you if motion sickness strikes. Technology delivers an extreme experience safely to your sofa 😌

But this magic has an uncomfortable downside: only the viewer gets the sofa. You feel as though you are almost flying, almost falling, almost standing at the edge of an abyss. The brain believes it, the body tenses, and the wow effect works. Yet for that “almost” to happen, someone had to do everything for real: take off, follow the route, lug the equipment, work in the desert heat, or freeze on a 7,000-metre peak—with deadlines and a production crew behind them.

The aviation angle is especially close to my heart. I love flying, gliders, simulators, and that feeling of air and space. So I understand perfectly why Apple wanted this particular episode. Flight is an almost ideal demonstration of immersive video. You do not need to explain why 3D and 180-degree footage matter: people put on the headset and their bodies understand immediately.

For businesses, this leads to an uncomfortable conclusion. The more platforms demand authentic, on-location footage, the more often content creators will have to enter genuinely dangerous situations. Mountains, deserts, aviation, diving, extreme sports—everything VR sells well tends to look terrible in a strict risk spreadsheet.

AI-generated content can be a lifesaver for such spectacular scenarios. Generate a desert, an aircraft, and a beautiful camera move: nobody crashes, the insurer does not weep, and the lawyers do not turn grey. There is only one small problem. As soon as viewers realize it is synthetic, the magic of “I was there” quickly becomes “that was beautifully rendered.”

It is a strange fork in the road for immersive media. Reality delivers a genuine wow effect but can create an unjustifiably high risk, sometimes measured in human lives. Generation is safe but weakens the feeling of authenticity. The most valuable content may emerge where technology enhances real-world experience without turning people into disposable props for a beautiful shot. The golden mean. Whoever reaches it first will win 😑