|
It's a lot harder to break into solving real, hard problems. For instance, consider the market I know: neutron physics modeling of nuclear reactors. The existing solutions represent a ton of accumulated domain knowledge, but they're none of them very user-friendly or well-integrated into either the reload design process or the document creation process, nor were they designed with certain improvements in technology infrastructure (distributed computing, GPU computing) in mind. Getting into this type of technology is hard, though - you need to pay for access to nuclear data files, and vast reserves of historical data are essential for validating your code's predictions, as well as reams of experimental thermal-hydraulic data for building empirical correlations. Information about modern neutron transport algorithms is scattered throughout academic journals, and unbiased comparisons, to say nothing of sample implementations, are so rare as to be nonexistent. Once you've done all of this work, then you need to produce all of the documentation required to convince the NRC that your algorithm is accurate enough for design calculations. And after all of that, chances are the utilities will just buy the fuel vendors' codes because why deal with two companies, and the fuel vendors will keep using their in-house codes, because fuck you, that's why. To compete in a market like that, you need connections, and you need experience, probably way more than 4 or 5 people's worth of either. A minimum viable product is probably a man-decade or so of labor. |