Or you do both. Start from the point of known technologies and reverse engineering and improve from there. China bootstrapped their high-tech industry with not just spying, but also licensing technology, join-ventures, training students/employees abroad and then bringing them back as experienced researchers/engineers. No need to start from first-principles. Plus even when you do have tech-specs on something, that doesn't mean you know how to manufacture it, or it may need modification to address broader system tradeoffs.