This article first appeared in Term Sheet, Fortune’s daily newsletter about deals and dealmakers. Sign up here.

First Round Capital released its annual survey of more than 700 startup founders this morning. It paints a picture of a complicated, sometimes contradictory crew of people. Which isn’t surprising – you have to be at least a little irrational to start a startup. Case in point:

  • Nearly 20% are certain their companies will be worth $1 billion, but just as many had to lay off staff this year.
  • The number one reason those surveyed believe their company could fail is an inability to raise follow-on capital. Not bad execution, not a lack of product-market fit, not burn rate. And yet! Only 39% are optimizing for profitability, versus 61% optimizing for growth. That explains why a majority (51%) said their burn rate is higher than it was a year ago. Only 13% are already profitable.
  • 57% believe we’re in a bubble (down from 73% last year). A majority think it’ll get harder to raise capital in the coming months, and two thirds think investors will have the upper hand going forward. And yet! There’s that pesky optimism, with only 15% admitting their company is not likely to be worth $1 billion or more.

Irrationality can manifest itself in frustrating ways:

  • On average, respondents think gender and ethnic parity in the startup world is just 14 years away, despite the fact that 11% of them currently run companies with companies with zero women employees and 61% of them have all-male boards. Late stage companies are the worst offenders, with a three times higher likelihood of not having a single woman on their board.
  • Regarding the gender issue, half of the men surveyed believe it’s a pipeline problem. The women surveyed know it’s an unconscious bias problem. (One womansplain-y thought: There are lots of women in tech; perhaps try expanding your network. For board members, try TheBoardlist.) The lack of female role models hurts, too. The top tech heroes cited are Elon Musk, Steve Jobs and Mark Zuckerberg.

There’s a lot more data where that came from. Read the rest here.