> Like if an Intel manager comes up with an idea and asks an engineer to work up a prototype, surely it'd be illegal for the engineer to quit and work up that same prototype for a new chip-design startup instead?
Yet this is how many Silicon Valley (back when the valley was actually about Silicon) start-ups began. Generally, however, the pattern was engineer and manager would identify a business need that real customers have. They would attempt to build it as a "pet project", be stiffed by conservative and hunch-driven product organization and would then both quit and take a large chunk of the engineers along with them.
The reason this succeeded is that NDAs and non-competes are unenforceable in California. Law suits would still come (due to alleged theft of IP, which generally never happened) and be settled out of court for only a small proportion of the the equity the "idea" produced.
Personally, I will stand by my words. If you can't build your idea it's worthless (you can't estimate whether others can either). If everyone else can (welcome to web 2.0), it's worthless as well. Sure you may be able to sue somebody over your idea (but you generally have to have something more, like a claim of IP theft), but your lawsuit proceeds will be a fraction of what the person who "stole" your idea has made out of it. That $65mm doesn't change Zuckerberg's life _a single bit_ (not having another $65mm
might only make life easier for him, as that's one less bodyguard to hire to protect it).
On the other hand, if you have a genuinely new idea i.e., you have a new CPU architecture you invented but don't have the ability to build the prototype chip _all by yourself_ then one engineer quitting and competing with you won't succeed unless he _actually_ steals your design (this is real IP theft) and even then you'll likely outdo him (as he'll have a bug for bug copy of your design, whereas you actually understand the design choices).
If management wants to be rich, they have to treat engineers with utmost respect. There's difference between engineers mere _coders_ who turn business requirements into code (usually by gluing software others have written together) and engineers. Spot the engineers early on, challenge them, give them autonomy and responsibility. Let them work on pet projects (Google calls it "20% time", but it's merely a codification of something successful companies had been doing before), tune your product vision with their data.
Thanks for correction. I treat NDAs as enforceable, I stick by my word. Just to be clear: I am entirely against IP theft (taking actual source code from a company).
Nonetheless the idea of "social network for X" is not unique and deserves no protection similar to what a chip architecture or a machine learning algorithm may receive. Not that there aren't ambiguous examples e.g., using SVMs to categorize users of a social network or using an emulated RISC architecture to profile the GUI front-end of a web application (such as a social network) but in the cases the idea is not very useful without the engineering chops to put it into use.