Small by Design
In 2007, three engineers at Google sat in a room and created Go — one of the most influential programming languages of the past decade. Not a team of hundreds. Three people. Each one the best in their field. That story has stayed with me for a long time. Not because of Go, but because of the logic behind it: when talent density is high enough, a small group can do things that an entire army cannot.
Cyberk was built on that belief. About 30 people. Small on purpose. Not because we cannot afford to hire more — but because I believe more people does not always mean more value.
I do not post job listings. I go to universities — FPT, RMIT, Hanoi University of Science and Technology — find the best students, and invite them to intern while they are still in school. Real pay, real projects, real responsibility. In Vietnam, this is rare. Most companies wait until graduation. I do not.
By the time they graduate, they have been at Cyberk for two or three years. They already understand the culture. They have already shipped production code. They are not juniors starting from zero — they are battle-tested engineers who happen to have just received a diploma.
There is something I think about often: great developers do not want to work with mediocre ones. Not out of arrogance — out of efficiency. Nobody wants to spend an entire morning fixing a bug that should not have existed. Nobody wants to explain the same thing for the third time.
What they want is to sit next to someone who challenges them. Someone who reviews their code and catches the race condition they missed. Someone who asks the right question at the hardest part of the problem.
Cyberk tries to create that environment. And there is something interesting: when talent density is high enough, it becomes self-attracting. Good people pull in more good people. No advertising needed, no employer branding — just an environment good enough that they want to stay and introduce their friends. When the baseline is already high, the ceiling rises on its own. And I can drop a lot of processes — performance reviews, review cycles, training pipelines — because those things were designed for a context that Cyberk does not have. What remains is time spent creating.
In 2026, "vibe coding" became a trend. Everyone got excited about letting AI write code, and I watched from the side. Sounds great, but I do not believe it produces production-ready software. Code that nobody understands at a fundamental level will sooner or later become tech debt.
Cyberk went a different direction. We developed our own philosophy — we call it AI-driven development. Not vibe coding, but using AI with intention to accelerate. We build our own tools, write our own prompts, design our own agent workflows. Each engineer operates three or four AI agents — the agents execute, the engineer commands and makes decisions. AI does not replace thinking. AI frees up time to think deeper.
That freed-up time, I do not use it to stuff more tickets into a sprint. I give it back to the engineers. For research. For thinking carefully about architecture. For solving the kinds of problems that AI cannot solve — ambiguous ones, ones that require judgment, ones that need a real conversation.
I get asked a lot: why not build your own product?
Because I choose to focus on one thing: engineering. I do not pretend to be an entrepreneur. I need the business brain — yours. You bring the vision and the market. I bring the technical depth. That division is intentional.
I could hire 100 engineers tomorrow. I choose not to. Because the moment you cross the talent density threshold, everything starts to degrade. Communication slows down. Standards drop. Good people leave because they are surrounded by average.