One aspect of rationality is acting optimally (reasonably) on given data. In startups, you often need to act on little data, because when you wait for perfect information, the opportunity is gone. Unknowns can still be handled rationally, but need to be treated probabilistically. From this point of view, I really like the Lean Startup method, because it provides a framework to addressing the unknowns: lay out your plan, identify the greatest unknown, test it efficiently, rinse and repeat.

Re: biases and startups: - bandwagon and VCs - loss aversion - founders being averse to pivoting, changing strategy after working on something for so long - planning fallacy - general issue with software development - projection bias - founders often assume things of others, and build the wrong produce instead of talking to potential customers

On the flipside, biases can also be taken advantage of, for example, when selling to people, using anchoring, framing, etc.