"There is a trend towards solopreneurship in conjunction with vibecoding," says Philip Karmanov from Black Snowflake on the main developments of 2025

We continue to wrap up the year 2025 with gaming (or gaming sector-related) teams and experts. In the sixth conclusive interview, we talked with the founder and CEO of Black Snowflake, Filipp Karmanov.

How did 2025 turn out for your team? What were you able to achieve, what are you proud of, and what, on the other hand, didn’t you manage to accomplish?

Filipp Karmanov, Black Snowflake: This year was thematic, and the theme was AI. While in 2023 it was primarily used for generating art references, now we can confidently talk about the reality of creating a quality (or at least playable) game entirely based on neural networks. Code, game design, graphics, sound, testing, integrating everything through MCP servers—everything that can be subject to any algorithm can truly be shifted onto neural shoulders. And even though the main rule remains "don’t trust, verify," the progress in just a couple of years is impressive. All of this is commendable, provided one can navigate through the tons of informational noise in this area.

Regarding games, several web projects were launched under NDA. One bold idea fell through due to legal risks, but this did not dampen our enthusiasm.

What conclusions did you draw as a development studio by the end of 2025?

Filipp: “If it works, don’t touch it.” This refers to the development pipeline: integrating the same AI agents can take much more time than it seems. Of course, this rule is not set in stone, but still.

Have the practices for interacting with publishers/investors changed? Has it become easier or harder to work with them?

Filipp: Investor money has become expensive again, especially against the backdrop of last year’s crypto boom. I’d like to particularly note the trend towards solopreneurship, especially in conjunction with vibecoding. If just a few years ago, many representatives of venture capital funds stated that companies with one founder were inadequate and needed at least two, now these same figures might be actively promoting the ideology of launching products cheaply and quickly alone.

How did the year turn out for the niche/genre you work in?

Filipp: For web games, platforms and infrastructure continue to develop. New solutions are emerging: I’d like to highlight the launch of Playgama. There is a noticeable decline in the Telegram audience that moved away from the play-to-earn games popular last year, but the most engaged audience remains, with which such companies work successfully.

Which strengthening or emerging trends in your niche/genre do you expect in 2026?

Filipp: The general trend, regardless of the niche, is AI development. Neural networks are becoming exponentially more powerful and accessible. What justified investing in proprietary hardware at the beginning of the year is now sometimes provided for free by market giants. There is an intense R&D and budget race among them, the outcome of which I dare not predict.

Speaking of web games, a surprise came closer to the end of the year. In November, Apple announced the Mini Apps Partner Program—a program for apps with mini-app functionality, under which the App Store commission on payments within them via In-App Purchase is halved to 15%. Currently, apps can decide whether to participate in the program. However, participation conditions include several intricate requirements. For instance, the requirement for full integration with Apple's infrastructure (Advanced Commerce API, Age Rating API) and complete control over all mini-apps through review and manifest. Certain platforms, like Telegram, strive to distance themselves from external control, so it is unknown whether the commission reduction will outweigh the introduced requirements.

What plans does the team have for the coming year?

Filipp: We are working on a service related to the gaming market, which we planned to launch this December, but operational challenges make the beginning of next year a more likely launch time.