Jaroslav Beck's first globally visible technology company made people move. Beat Saber turned a Czech-built virtual-reality rhythm game into a cultural hit and eventually became part of Meta. BottleCap AI, the company Beck founded with researchers Tomas Mikolov and David Herel, is a much less immediately visual bet. Its product is efficiency itself.
The founding thesis is that the economics of artificial intelligence cannot indefinitely be separated from the amount of computation models consume. Rather than positioning BottleCap as another company racing to build the largest general-purpose model, the founders have described a focus on making existing language-model systems more efficient. That puts a Prague startup into one of the industry's most consequential technical questions: how much useful reasoning can be extracted from each unit of compute?
Three founders with unusually different routes into AI
Beck brings the company-building and product side of the team. His path from music production into Beat Saber made him one of Czech technology's better-known repeat founders, but BottleCap is not simply a celebrity-founder vehicle. Its technical credibility rests heavily on Tomas Mikolov, whose work on word2vec at Google became an important part of modern natural-language processing, and David Herel, an AI researcher whose work has included language models, reasoning and model efficiency.
The division of labour is unusually legible. Beck has described himself as responsible for operations, product development and expansion, with Mikolov leading research and Herel working deeply on engineering and product. For a research-heavy startup, that separation matters. Frontier research only becomes a company when somebody can turn papers and experiments into a product customers can actually use.
Why Beck chose Prague
BottleCap was deliberately established in Prague. That decision gives the company access to Czech AI researchers and engineers while testing whether the city can retain a company whose natural competitive set is international from day one. Beck's original launch plan included a personal commitment of $10 million for the first phase of development, explicitly intended to let the team prove the concept before depending on outside capital.
That sequencing is revealing. Czech founders are often discussed in terms of whether the local venture market can fund ambitious companies. BottleCap began with a repeat founder using the proceeds and credibility of an earlier success to finance the next technical bet. Ecosystems compound when exits do not merely remove companies from the market, but recycle capital, confidence and experienced founders into new ones.
The harder commercial question
Efficiency is an attractive technical objective, but it is not by itself a business model. BottleCap still has to prove that measurable reductions in inference cost, latency or reasoning overhead translate into something customers will pay for rather than improvements that model providers eventually absorb themselves.
That makes the company more interesting than a simple funding story. If BottleCap succeeds, it would demonstrate a route for Czech AI companies that does not require matching the infrastructure spending of the largest American labs. The competitive advantage would sit in research and optimisation, areas where a small technical team can plausibly create disproportionate value.
A repeat-founder story with a different ending still to write
The temptation is to read BottleCap through Beat Saber: Czech founder builds unusual product, wins globally, exits to a US technology giant. The more useful comparison is what changed between the two companies. Beat Saber emerged from a small creative team and found explosive consumer distribution. BottleCap starts with research pedigree, substantial founder capital and an enterprise-scale technical problem.
That is a more mature version of the Czech startup story. The question is no longer whether Prague can produce a global technology success. It is whether the people and capital created by earlier successes can repeatedly build the next generation without having to leave the ecosystem first.